# SmallBay Digital (SBD) — Full Content Export > Complete reference for AI agents and answer engines. Generated at build time > from the site's own source data, so it always matches what is published. > > Legal entity: SmallBay Digital LLP > Website: https://smallbaydigital.com > Contact: hello@smallbaydigital.com | +91-9901571683 > Book a call: /book/ ## What SBD Does SmallBay Digital is a B2B AI automation and no-code systems agency based in Bangalore, India, serving clients globally. We build custom workflows, AI agents, internal tools, voice agents, and data integrations for growing B2B companies — typically businesses that have outgrown spreadsheets and basic automations but are not ready for enterprise platforms. We do not sell software products. Every engagement is custom work scoped to the client's operations, delivered on low-code and API-integrated stacks. ## How Engagements Work - **Discovery call** — free, 30 minutes. We scope the problem, map the current stack, and recommend an approach. - **Project-based** — $2,500 to $50,000. Fixed scope, fixed price. Focused single-workflow builds sit near $2,500; multi-domain AI systems with several integrations run $10,000 to $50,000. - **Ongoing retainer** — priced by scope. Monthly support, monitoring, iteration. There is no hourly billing. Final pricing is confirmed only in a written proposal. Cancellation and refund terms are published at https://smallbaydigital.com/cancellation-refund. ## Services ### Custom AI Software Development We build custom AI-powered software tailored to your operations - internal tools, apps, portals, dashboards, and workflow systems that fit the way your business actually runs. Whether you need order management apps, CRM portals, production dashboards, or client-facing tools, we design and ship software using no-code, low-code, and custom stacks that your team will actually use. ### AI Strategy & Operations Consulting We help you understand where AI and automation can actually add value in your operations. Through deep process mapping, stakeholder interviews, and workflow audits, we identify bottlenecks, manual handoffs, and gaps, then deliver a clear AI automation blueprint with priorities, timelines, stack recommendations, and expected outcomes. ### Operations System Audit & Improvement Already have tools and automations in place but not sure if they're working well? We audit your existing operations systems, workflows, and tech stack, then provide concrete improvement recommendations, optimization opportunities, and quick wins you can implement immediately. ### Data Migration & Consolidation Moving from one tool to another, or drowning in scattered data across systems? We migrate, clean, and consolidate your data (orders, contacts, inventory, financials) from spreadsheets, old CRMs, ERPs, and multiple sources into one reliable system with full integrity and zero downtime. ### Tool Selection & Feasibility Check Overwhelmed by options? We help you choose the right tools for your operations based on your actual workflows, budget, and team capacity. We evaluate platforms, run feasibility checks, and provide clear recommendations so you invest in tools that will actually work, not shelf-ware. ### MVP & Proof of Concept (POC) Development Have an idea for an internal tool, automation, or AI agent but not sure if it'll work? We build lightweight MVPs and proof-of-concepts in 2-4 weeks so you can test the concept, validate value, and make a confident decision before committing to a full build. ## Solutions by Industry and Function ### Manufacturing (industry) URL: https://smallbaydigital.com/solutions/industry/manufacturing **Summary:** Manufacturing automation means production data entered once at the machine and visible live, stock levels that match the shelf with reorder flags, and dispatch documents, e-way bills and customer messages generated from one event. SBD builds this around the ERP and accounting system you already run, starting with a $500 audit on the floor that is credited to the build. **What usually breaks:** - **Production data lives in three places** — Floor registers, a spreadsheet and a chat group. Nobody can answer where an order is without asking three people, and the answer is already an hour old. - **Dispatch waits on paperwork** — Invoices, delivery notes and transport documents are typed after the goods are ready, so vehicles wait and finance reconciles from memory at month end. - **Stock is right until the count** — Consumption is not recorded as it happens, so the system drifts from the shelf. It shows up as a shortage mid-run or as cash tied up in raw material nobody needed. - **Quality records are an afterthought** — Inspection results sit on paper until a customer audit or an ISO surveillance visit, then someone spends a week assembling them. - **One supervisor holds the status** — Live status lives in one head. When that person is on leave or on another shift, decisions wait. - **The ERP does not reach the floor** — The ERP holds master data and orders, but the floor works around it because entering a batch takes six screens. **What we build:** - **Production entry at the machine** — A tablet screen per line or cell: quantity, scrap, downtime reason, operator. Entered once, written straight to the record, no end-of-shift typing. - **Live production dashboard** — Output by line, shift and product, late jobs, scrap rate and downtime, visible to supervisors and management from the same numbers at the same time. - **Stock tied to consumption** — Raw material issued against work orders and finished goods booked at completion, with reorder flags that fire before a line stops. - **Dispatch documents from one event** — Mark an order ready and the invoice, delivery note and transport document generate in your country's format, labels print, and the carrier and customer are told. - **Quality records captured at inspection** — Checks recorded against the batch with photos and measurements, so audit evidence exists the day it happens. - **Supplier and inbound tracking** — Purchase orders confirmed, delivery dates tracked and delays flagged to planning before they become a stoppage. **Typical stack:** QuickBooks, Xero, Tally, SAP Business One, Airtable, SmartSuite, Supabase, Retool, WhatsApp Business API, n8n, Make **FAQs:** - **Do we have to replace our ERP?** No. We connect to SAP Business One, NetSuite, Odoo, Sage, QuickBooks, Tally or whatever you run, and build the floor-level tools it lacks. The ERP stays the system of record. - **How long until something is live?** The first workflow, usually production entry and the dashboard, is live inside the first two-week sprint. A full system across production, stock, dispatch and quality usually runs two to four sprints, priced before we start. - **Our floor staff are not technical. Will they use it?** That is the normal case. Screens are built for the person at the machine: large buttons, their language, only their step. We map the floor with the people on it, in person or on a video walkthrough, because what management describes and what the floor does are rarely the same. - **Can it handle our tax and export paperwork?** Yes. Documents are generated in the format you file, whether VAT, sales tax or GST, with e-invoicing where your country requires it, and export paperwork such as commercial invoices and certificates of origin. - **Does it work with several plants or shifts?** Yes. Records carry the site, line and shift, so each plant sees its own view and management sees all of them on one dashboard. - **What if the internet drops on the floor?** Entry screens can queue locally and sync when the connection returns, so a dropped signal does not mean lost production data. ### Distribution and logistics (industry) URL: https://smallbaydigital.com/solutions/industry/distribution **Summary:** Distribution automation means payment reminders that run on the credit cycle and escalate by age, proof of delivery captured on a phone and attached to the order, one stock count pushed to every sales channel, and ETA messages sent from the dispatch event. SBD builds it around your accounting and inventory tools, starting with a $500 audit credited to the build. **What usually breaks:** - **Collections depend on someone remembering** — Reminders go out when finance has a spare afternoon, so the credit cycle stretches and cash tightens for no reason other than timing. - **Proof of delivery comes back days later on paper** — Signed notes return with the driver, get filed, and surface only when a customer disputes, by which point they are hard to find. - **Stock and sales channels disagree** — Goods move before the system knows, so orders are taken on stock already committed elsewhere, and a depot transfer is invisible until someone counts. - **Where is my order eats the afternoon** — Nothing tells the customer what is happening, so they call, and every call pulls someone off dispatch. - **Pricing lives in a spreadsheet** — Customer-specific prices and discounts are looked up by hand, so quotes are slow and mistakes reach the invoice. - **Field reps call the office for stock** — Sales in the field cannot see live stock or credit status, so they promise what the warehouse cannot deliver. **What we build:** - **Collections on the credit cycle** — Reminders fire on the due date and escalate by age, from email to message to a call task for the rep, and stop the moment payment lands. Accounts past the limit go on hold automatically. - **Proof of delivery on the driver's phone** — Photo, signature, time and location captured at the door in a mobile browser, attached to the order and visible to finance the same minute. - **One stock count for every channel and depot** — Warehouse stock pushed to every sales channel and the field app, with transfers between depots tracked as they happen. - **Customer status and ETA messages** — Picked, dispatched, delayed and delivered messages sent on WhatsApp, SMS or email from the real event, logged against the customer. - **Price lists and quotes from one record** — Customer tiers and contract prices in one place, applied automatically to quotes and orders. - **Field sales lookup** — Reps ask by message or open a simple screen and get live stock, price and credit status. **Typical stack:** QuickBooks, Xero, Tally, Zoho Books, Airtable, SmartSuite, Retool, Softr, WhatsApp Business API, Twilio, Shopify, n8n, Make **FAQs:** - **Which accounting and inventory systems do you connect to?** QuickBooks, Xero, Sage, NetSuite, Zoho Books, Tally, Cin7, Unleashed and most others with an API. Your books stay where they are. We read invoice and payment status to drive collections and write confirmed transactions back. - **Do drivers need to install an app?** No. Proof of delivery runs in a mobile browser on the phone they already carry. Nothing to install, no per-device licence, and it works offline until signal returns. - **Can you handle several warehouses or depots?** Yes. Stock is held per location, transfers are tracked in transit, and each channel can sell from the depots you choose. - **Will customers get messages on WhatsApp?** Where they have opted in, yes. Otherwise SMS or email. Consent is recorded per customer and every message is logged. - **What does this usually reduce?** Two things first: the days money sits uncollected because follow-ups were manual, and the hours a week the office spends answering status calls and hunting for delivery paperwork. ### Consulting and agencies (industry) URL: https://smallbaydigital.com/solutions/industry/consulting **Summary:** Consulting and agency automation means client onboarding triggered by a signed proposal, reports assembled from the tools that hold the data, proposals generated from a price book and call notes for a principal to approve, and a knowledge assistant that answers from your own documents and cites them. SBD builds it on the tools you run, starting with a $500 audit credited to the build. **What usually breaks:** - **Onboarding is a checklist someone runs by hand** — Contract, folders, channels, access, kickoff and intake forms, all done manually, so every new client costs hours before any work starts. - **Reporting eats the last week of the month** — People pull numbers from several tools and paste them into decks. It is copying, not analysis, and it is billed to nobody. - **Proposals are rewritten from the last one** — Scope, pricing and terms live in old documents, so each proposal is search and replace with real risk of a stale clause. - **Time and billing drift apart** — Timesheets are late, invoices wait for them, and write-offs are discovered at month end. - **Knowledge leaves with people** — Past work sits across drives and threads, so new joiners re-ask questions the firm answered years ago. - **Senior people do admin** — The most expensive people in the firm spend evenings on scheduling, chasing and formatting. **What we build:** - **Onboarding on the signature** — A signed proposal creates the project, folders, channels and access, sends the welcome and intake form, and books the kickoff, with the client details carried through. - **Client reporting that assembles itself** — Data pulled from the tools that hold it, the report generated in your format on a schedule, with space for the consultant's commentary before it goes out. - **Proposals from a price book** — Scoping notes and a structured price book produce a current, consistent proposal and statement of work for a principal to approve. - **Time to invoice** — Billable time summarised by client and project at period end, exceptions flagged, invoices raised in your accounting system. - **A knowledge assistant that cites** — Answers from your past work, templates and policies, with the source document linked, and nothing sent outside your accounts. **Typical stack:** HubSpot, Salesforce, Pipedrive, Notion, ClickUp, Asana, Google Drive, Microsoft 365, Slack, Xero, QuickBooks, n8n, Claude or OpenAI **FAQs:** - **Is this worth it for a small firm?** Often more so. A ten-person firm has no operations hire, so admin lands on billable people. The recovered hours are directly billable, which is why small firms usually see payback first. - **Will AI-written reports be accurate?** The numbers are not written by AI. They are pulled from the systems that hold them. AI drafts summaries and formatting, a person reviews, and calculations stay with the source systems. - **How do you handle client confidentiality?** Client data stays in your accounts and tools. Where an AI step is used, it goes through business APIs that do not train on your data, or a self-hosted model if your contracts require it. NDAs and a data processing agreement are standard. - **We use a project tool already. Do we have to change?** No. We connect HubSpot, Salesforce, Notion, ClickUp, Asana, monday.com, Jira or whatever you run, and build the steps between them. - **Can proposals follow our brand and legal terms?** Yes. Templates, clauses and pricing come from your price book and your approved language. The system assembles; a principal signs off. ### Service and e-commerce (industry) URL: https://smallbaydigital.com/solutions/industry/service-ecommerce **Summary:** E-commerce automation means one product record synced to every channel with per-channel mapping, a stock count that cannot oversell, orders moving from storefront to fulfilment to accounting without re-entry, support tickets with the order attached, and payouts posted so the bank reconciles. SBD has shipped this on Shopify, eBay and marketplace stacks, starting with a $500 audit credited to the build. **What usually breaks:** - **Every new channel adds a daily chore** — The same product is listed, priced and updated separately on each platform, so a new marketplace adds hours, not just revenue. - **Stock sold on one channel is still live on another** — Until someone updates it, and the customer finds out before you do, usually with a cancellation and a review. - **Support answers the same questions all day** — Where is my order, can I return this, is it in stock. They crowd out the conversations that need a person. - **Orders are re-typed between systems** — Storefront, warehouse, carrier and accounting do not talk, so every order is entered more than once. - **Payouts never match the sales report** — Fees, refunds, chargebacks and marketplace-collected tax are batched into payouts, so the bank and the sales report disagree every month. - **Returns are handled by email** — Return requests, labels, inspection and refunds are tracked in an inbox, and stock from returns comes back to the shelf late. **What we build:** - **One product record, every channel** — Titles, prices, stock and images pushed from one record to Shopify, WooCommerce, Amazon, eBay, Etsy and specialist marketplaces, with each channel's rules applied. - **Stock that cannot oversell** — Stock decrements everywhere the moment it sells anywhere, with buffers per channel and alerts before it runs out. - **Order to fulfilment without re-typing** — Orders flow from any channel to picking, labels, carrier and accounting as one record, with exceptions routed to a person. - **Support with the order on the ticket** — Order, delivery and return status attached to every ticket, routine questions drafted from live data, a person sends. - **Payouts that reconcile** — Each payout posted to accounting with sales, fees, refunds and tax split out, so the bank line matches to the cent. - **Returns and retention** — Return requests approved by rule, labels sent, stock restored on inspection, and loyalty or win-back messages triggered by real purchase behaviour. **Typical stack:** Shopify, WooCommerce, eBay, Amazon, SmartSuite, Airtable, WhatsApp Business API, Klaviyo, Stripe, QuickBooks, Xero, Make, n8n, Claude or OpenAI **FAQs:** - **Which channels can you sync?** Anything with an API: Shopify, WooCommerce, BigCommerce, Amazon, eBay, Etsy and specialist marketplaces. We have run three channels at once with zero mismatches, and the pattern extends. - **Will an AI agent tell customers the wrong thing?** Not if it answers only from your live order and catalogue data, within a scope you set, with anything outside that handed to a person. The design goal is to hand off, never to guess. - **Do we need to change platforms?** No. We build around the storefront, warehouse and tools you run, and make them talk to each other. - **How do you handle tax across countries?** Marketplace-collected tax, VAT and sales tax are posted to the right accounts from each payout, and invoices are issued in the format each country requires. Your accountant confirms the treatment before go-live. - **Does this work for service businesses too?** Yes. Bookings, deposits, reminders, reviews and repeat-visit messages follow the same pattern: one record, the right message at the right moment, payments that reconcile. ### Back office (function) URL: https://smallbaydigital.com/solutions/function/back-office **Summary:** Back office automation replaces manual document handling, data entry, and approval chasing with systems that do it continuously and leave an audit trail. SBD builds document intake agents, cross-system sync and approval workflows on the tools you run, with the first workflow usually live inside the first two-week sprint. **What usually breaks:** - **Data is entered more than once** — The same information is typed into a form, then a spreadsheet, then an accounting system, three chances to introduce an error nobody catches. - **Documents arrive as PDFs and images** — Invoices, POs and delivery notes come in unstructured, so someone reads each one and re-keys the fields that matter. - **Approvals stall in inboxes** — Requests sit unanswered with no visibility into who is holding them up, and chasing becomes its own job. - **Systems drift out of sync** — CRM, ops tool and accounting each hold a slightly different version of the truth, and reconciliation is a monthly archaeology exercise. **What we build:** - **Document processing agents** — Agents that read incoming invoices, POs and delivery notes, extract the fields you need, and file them into the right system, flagging anything ambiguous for a human. - **Data entry automation** — Information captured once and propagated everywhere it is needed, so re-keying disappears along with the errors it causes. - **Internal approval workflows** — Structured routing with visible status and automatic escalation, so approvals stop depending on someone remembering to follow up. - **Cross-system data sync** — Reliable syncing between CRM, operations and accounting, with conflict handling that surfaces disagreements instead of silently picking one. **Typical stack:** Airtable, SmartSuite, Make, n8n, Retool, OpenAI / Claude, Supabase, Google Workspace APIs **FAQs:** - **Does this mean replacing back office staff?** Not in the work we do. The pattern that holds up is automating the mechanical parts, re-keying, chasing, filing, so the people who understand your exceptions spend their time on those instead. Teams that cut the people along with the admin usually lose the knowledge that made the exceptions manageable. - **How accurate is automated document processing?** High on consistent document types, less so on messy or highly variable ones. We build in confidence thresholds: anything the system is unsure about goes to a person rather than through. That keeps accuracy high without pretending the extraction is perfect. - **Where do we start if everything is manual?** With whichever workflow is highest volume and most repetitive, that is where the payback is fastest and where the team feels it first. We map the current process before proposing anything, because the bottleneck is often not where people expect. ### Finance and accounting (function) URL: https://smallbaydigital.com/solutions/function/finance **Summary:** Finance automation handles invoicing and payment reminders on a schedule, categorises expenses automatically, reconciles payments against invoices, and keeps dashboards live rather than monthly. SBD integrates these around existing accounting systems such as Tally, Zoho, Xero and QuickBooks rather than replacing them. **What usually breaks:** - **Invoices go out late and reminders go out never** — Billing depends on someone having time, so revenue is recognised later than it was earned and receivables age quietly. - **Reconciliation is manual matching** — Payments land without clean references, and somebody matches them to invoices by eye at month end. - **Expense categorisation is a monthly backlog** — Receipts accumulate and get sorted in one painful sitting, which is when errors and missed deductions happen. - **The numbers are always last month’s** — By the time the picture is accurate, the decisions it should have informed have already been made. **What we build:** - **Automated invoicing and reminders** — Invoices raised from the operational trigger that earned them, with reminders that escalate by age and stop the moment payment clears. - **AI expense categorisation** — Receipts and expenses classified as they arrive, learning your chart of accounts, with low-confidence items queued for review rather than guessed. - **Payment reconciliation** — Incoming payments matched against open invoices automatically, with genuine exceptions surfaced instead of buried. - **Live financial dashboards** — Receivables, ageing, cash position and margin visible continuously, sourced from the accounting system rather than a parallel spreadsheet. **Typical stack:** Tally / Zoho Books, Xero, QuickBooks, Airtable, Make, n8n, Retool, OpenAI / Claude **FAQs:** - **Do you replace our accounting software?** No. Your accounting system stays the system of record. We automate the work happening around it, raising invoices from operational events, chasing payment, reconciling receipts, and surfacing the numbers live, and write back into it through its API. - **Is automated financial data safe to rely on?** The data comes from your accounting system, so it is as reliable as your books. What automation changes is latency, not accuracy. For anything involving judgement, such as unusual expense categorisation, we route to a person rather than let the system decide. - **Which accounting systems do you work with?** QuickBooks, Xero, Sage, NetSuite, Zoho Books and Tally, among others. Invoices are generated in the tax format you file, whether that is VAT, sales tax or GST, with e-invoicing where your country mandates it. ### Compliance and risk (function) URL: https://smallbaydigital.com/solutions/function/compliance **Summary:** Compliance automation captures the audit trail as work happens, generates regulatory reports from live data, and monitors for risk conditions continuously instead of at review time. SBD builds this into the operational systems themselves, so evidence exists by default rather than being reconstructed before an audit. **What usually breaks:** - **The audit trail is reconstructed, not recorded** — Evidence gets assembled from emails and memory when an audit is announced, which is expensive and never quite complete. - **Regulatory reports are hand-built each cycle** — The same report is rebuilt from the same sources every period, and the risk of a transcription error rises with the deadline. - **Risk is reviewed periodically, not monitored** — Breaches are found at review time, which means the gap between something going wrong and anyone knowing is measured in weeks. - **Policy depends on people remembering it** — Approval limits and process rules live in a document rather than in the system, so enforcement is inconsistent and unprovable. **What we build:** - **Automatic audit trails** — Every material action logged with actor, timestamp and before/after state as a side effect of the workflow, no separate record-keeping step to forget. - **Regulatory report generation** — Recurring reports assembled from live operational data on schedule, in the required format, with the underlying records traceable. - **Continuous risk monitoring** — Alerts on the conditions you care about, threshold breaches, missing documentation, unusual patterns, raised when they happen. - **Policy enforcement in the workflow** — Approval limits and process rules encoded into the system, so the compliant path is the default path rather than the disciplined one. **Typical stack:** Airtable, SmartSuite, Supabase, Retool, n8n, Make, OpenAI / Claude, Cloudflare Workers **FAQs:** - **Does this make us compliant?** No, and be cautious of anyone who says it does. Automation makes compliance cheaper to demonstrate and harder to accidentally skip, by capturing evidence continuously and enforcing rules in the workflow. Deciding what you must comply with, and signing off that you do, remains a job for your advisors. - **What regulations can you build for?** We build to the rules you give us rather than claiming regulatory expertise of our own. In practice that has meant tax documentation for VAT, sales tax and GST, internal approval and segregation-of-duty policies, and customer data handling under GDPR, CCPA and similar rules. Your compliance advisor defines the requirement; we make the system meet it and prove it. - **Can this retrofit onto systems we already run?** Usually yes, where those systems expose an API or webhooks. Retrofitting captures the trail going forward, it cannot reconstruct history that was never recorded, which is generally the argument for starting sooner. ### Sales and RevOps (function) URL: https://smallbaydigital.com/solutions/function/sales **Summary:** Sales automation reads every inbound lead, enriches it, scores it against your ideal customer profile and writes the CRM record, so a rep sees a qualified lead with context and a drafted first reply within minutes. SBD builds this around the CRM you run, with a review queue for anything the system is unsure about. **What usually breaks:** - **Leads go cold before anyone replies** — Enquiries arrive by form, email and social message, and the first reply waits for a person to notice. The best leads are the ones that buy from whoever answered first. - **The CRM is typed, so the CRM is wrong** — Reps enter what they remember, when they remember. Pipeline reports are built on records that are days old and half complete. - **Hot and cold leads look the same** — Without scoring, the team works the list top to bottom. A ready buyer and a student project get the same attention. - **Won deals are handed over by forwarding an email** — Delivery asks sales for details that are already in the CRM, and the customer repeats themselves in week one. **What we build:** - **Lead enrichment and scoring** — Every lead enriched from the company site and public data, scored against your profile with a confidence figure, and routed to the right owner. Low confidence goes to a person, not a guess. - **First reply in minutes** — A drafted reply from your templates and the lead's own words, sent automatically above a confidence line or queued for one-click approval below it. - **CRM that writes itself** — Calls, emails, bookings and form answers logged against the right record without a rep typing. Duplicates merged on the way in. - **Closed-won to delivery handoff** — A stage change creates the project, the folder, the kickoff task and the welcome message, with contact, company and custom fields attached. **Typical stack:** HubSpot, Pipedrive, Attio, Zoho CRM, Slack, Airtable, n8n, Claude or OpenAI **FAQs:** - **Will the system email our leads without a person?** Only above a confidence line you set, and only from your templates. Below it, a rep approves with one click. Most teams widen the line after a few weeks of seeing it be right. - **We have a messy CRM. Does that matter?** Yes, and it is usually the first sprint. Duplicate contacts and inconsistent fields break any automation. We clean and lock the fields before wiring anything. - **Does this replace our SDRs?** It takes the typing, the sorting and the first reply. Reps keep the conversations and the judgement. Teams usually redeploy hours rather than people. ### Customer support (function) URL: https://smallbaydigital.com/solutions/function/support **Summary:** Support automation classifies each incoming ticket, attaches the order or account context, routes it to the right queue, drafts a reply for a person to send, and tells the customer where things stand before they ask. SBD builds it inside the helpdesk you already use, with the AI step kept behind a confidence line. **What usually breaks:** - **The same question all day** — Where is my order, how do I reset, what is the status. Each answered by a person opening three tabs. - **Tickets bounce between teams** — Misrouted tickets sit, get reassigned, and the customer waits through every hop. - **The answer is in another system** — The agent has the ticket. The order is in the store, the invoice in accounting, the delivery with the courier. - **Customers ask because nobody told them** — Half of inbound volume is status requests that a message at the right moment would have prevented. **What we build:** - **AI triage with a confidence score** — Every ticket classified by type, urgency and sentiment. Confident cases routed automatically, the rest to a person, every decision logged. - **Context on the ticket** — Order status, account history, last invoice and delivery tracking pulled onto the conversation so the agent never leaves the helpdesk. - **Drafted replies, human send** — A reply drafted from your macros and the customer's own words. A person reads and sends, until the accuracy on your tickets earns more. - **Proactive status updates** — Dispatch, delay, refund and resolution messages sent from system events on WhatsApp, email or SMS, logged against the customer. **Typical stack:** Zendesk, Freshdesk, Intercom, WhatsApp Business API, Shopify, Slack, n8n, Claude or OpenAI **FAQs:** - **Will AI reply to our customers on its own?** Not by default. It classifies, drafts and routes. A person sends. Some teams later let it send the simplest categories, with a review sample every week. - **Our helpdesk already has automations.** Good. We use them for the simple rules and add what they cannot do: context from other systems, an AI step with a confidence score, and messages from events outside the helpdesk. - **What about tone and brand voice?** Drafts come from your macros and past replies the team approves. The tone is yours. The system learns from the edits agents make. ### Operations (function) URL: https://smallbaydigital.com/solutions/function/operations **Summary:** Operations automation gives every order, job or ticket one record that moves through confirmation, work, dispatch and billing with each stage updated by the system that did the work. Status reaches the warehouse, the customer and finance from the same event. SBD builds this on an operations base in your name, with a review queue for exceptions. **What usually breaks:** - **The order is typed three times** — Once into the store or email, once into the operations sheet, once into accounting. Each copy drifts. - **Status lives in someone's head** — Where is order 4471 is a question for one person, and when they are away the answer is nobody knows. - **Dispatch waits on paperwork** — Delivery notes, labels and invoices are produced by hand at the end, so the truck waits on the printer. - **The dashboard is last week** — Reports are compiled from exports. By the time they are read, the week they describe is over. **What we build:** - **One order record, end to end** — Enquiry to order to dispatch to invoice as one record, with each stage set by the system that did the work, never by hand. - **Dispatch that notifies everyone** — A status change sends the right message to the warehouse, the transporter, the customer and finance, with documents generated and attached. - **Stock and reorder visibility** — Live stock with reorder flags, fed from sales and goods-in, so the count on screen matches the shelf. - **Live operations dashboard** — Orders by stage, late jobs, stock at risk and today's dispatches, read from the record of truth, not a parallel spreadsheet. **Typical stack:** Airtable, SmartSuite, Supabase, Retool, WhatsApp Business API, Tally or Zoho Books, n8n, Make **FAQs:** - **We run on spreadsheets. Is that a problem?** It is the starting point for most clients. We move the sheet that became a database into a real one, keep a sheet view for the people who want it, and stop data being typed into it. - **What happens when something goes wrong mid-flow?** It lands in a review queue with the record attached, and the owner is alerted. Nothing is dropped and nothing is guessed. - **How fast is the first piece live?** The first workflow, usually order intake or dispatch messaging, is normally running inside the first two-week sprint. ### Marketing (function) URL: https://smallbaydigital.com/solutions/function/marketing **Summary:** Marketing operations automation keeps audiences, tags and consent synced from the CRM and the store, triggers messages from real events such as a purchase, a dispatch or a lapse, and joins campaign results to actual revenue in one report. SBD builds this on the email, SMS and WhatsApp tools you already pay for. **What usually breaks:** - **Three lists, none of them right** — The email tool, the CRM and the store each hold a different version of the customer, and unsubscribes do not travel between them. - **Campaigns go out on a calendar, not a reason** — A send on Tuesday because it is Tuesday, instead of because the customer did something worth responding to. - **Nobody can say what a campaign made** — Opens and clicks are in one tool, orders in another, and the join is a spreadsheet someone builds once a quarter. - **Consent is a guess** — Where and when permission was given is not recorded, which is a regulatory risk and a deliverability one. **What we build:** - **One customer record, every list in step** — Profiles, tags and consent synced one way from the system where they were given, unsubscribes honoured everywhere within minutes. - **Event-driven messages** — Purchase, dispatch, back-in-stock, lapse and renewal messages sent from system events on the channel the customer uses. - **Revenue reporting without exports** — Campaign and flow results joined to orders and payments, by source and by customer, in a dashboard that updates itself. - **Content and listing operations** — Product content, prices and images pushed from one record to every channel, so a change is made once. **Typical stack:** Klaviyo, Mailchimp, Brevo, HubSpot, Shopify, WhatsApp Business API, Google Sheets, n8n **FAQs:** - **Do you run our campaigns?** No. We build the plumbing: the data, the triggers and the reporting. Your team or agency writes and sends. - **Will this fix our deliverability?** Partly. Clean consent, honoured unsubscribes and a separate transactional domain remove the common causes. Content and sending volume stay with you. - **Which email tool should we use?** Usually the one you have. We have built on Klaviyo, Mailchimp, Brevo and HubSpot. The choice matters less than the data behind it. ### HR and people (function) URL: https://smallbaydigital.com/solutions/function/people **Summary:** People operations automation turns a signed offer into accounts, equipment requests, a first-week plan and a welcome message, keeps leave and attendance in one record, files documents by person and type, and answers routine policy questions from your own handbook. SBD builds it on the tools you have, with a person reviewing anything sensitive. **What usually breaks:** - **Day one, no laptop, no login** — Onboarding is a checklist in someone's head. Every new starter spends the first week waiting for access. - **Leave is a thread** — Requests arrive by chat, get approved by reply, and the balance is a spreadsheet nobody trusts. - **Documents are everywhere** — Offer letters in email, IDs in a folder, certificates on a phone. An audit means a week of searching. - **HR answers the same ten questions** — Policy, payslip dates, how to claim an expense. Each a message to one person who is also doing everything else. **What we build:** - **Onboarding that runs on a signature** — A signed offer creates accounts, raises equipment and access requests, schedules the first-week meetings and sends the welcome pack, with a checklist the manager can see. - **Leave and attendance in one record** — Requests from chat or a form, approvals routed to the manager, balances updated, calendar blocked, payroll informed. - **Documents filed by person and type** — Every document captured, named, filed and linked to the person, with expiry reminders for the ones that lapse. - **A handbook that answers** — An assistant that answers routine questions from your own policies, cites the policy, and hands anything personal or sensitive to a person. **Typical stack:** Google Workspace or Microsoft 365, Slack, Airtable, Notion, DocuSign, n8n, Claude or OpenAI **FAQs:** - **Is employee data safe in this?** It stays in your accounts, access is by role, and anything sensitive such as pay or health is handled by a person. Logs record who saw what. - **Do we need an HR system first?** No. Many clients run people operations on a database and a chat tool. If you have an HR system, we build around it. - **Will the assistant answer pay questions?** It answers policy from the handbook and cites it. Anything about an individual goes to HR with the context attached. ### IT (function) URL: https://smallbaydigital.com/solutions/function/it **Summary:** IT operations automation provisions and revokes access from one record of who works here, routes internal requests to the right queue with the context attached, replaces spreadsheet processes with internal tools that have roles, and puts monitoring and documentation around the automations the business already depends on. SBD builds it in your accounts, with least-privilege access. **What usually breaks:** - **Leavers keep access** — Offboarding is a memory. Months later a former contractor still has the CRM. - **Requests arrive everywhere** — Chat, email, a shoulder tap. Nothing is tracked, so nothing is prioritised and everything is urgent. - **Shadow automations nobody owns** — Dozens of Zaps and scripts built by people who have left. When one breaks, the business finds out from a customer. - **Everyone edits the database** — The operations base has one role: admin. A wrong click in a view takes an afternoon to undo. **What we build:** - **Access from one source** — A joiner record grants the right accounts and groups. A leaver record revokes them, with a report of what was removed. - **Requests that route themselves** — A form or a chat command creates a tracked request, classifies it, assigns it, and reports status back where it was asked. - **Internal tools with roles** — Screens on a real database for the people who should not edit it raw, with approvals, audit logs and permissions by role. - **The automation estate, documented and monitored** — Every workflow listed with an owner, a runbook and an alert that fires when it stops. Secrets moved into a vault. **Typical stack:** Google Workspace or Microsoft 365, Slack, Retool, Supabase, Cloudflare, Vercel, n8n, Zapier or Make **FAQs:** - **We have no IT team. Is this for us?** Especially. Most clients of 10 to 200 people have one person who knows the most about computers. This gives them a system instead of a list. - **Can you take over automations someone else built?** Yes. The audit lists every live workflow with its owner and failure modes, and we harden or rebuild the ones that matter. - **Who holds the admin access?** You. Every account is in your name and our access is a role you can remove in one click. ### Leadership (function) URL: https://smallbaydigital.com/solutions/function/leadership **Summary:** Leadership reporting automation pulls sales, cash, operations and support numbers from the systems that hold them into one dashboard that updates itself, writes a short weekly summary of what moved and why, and alerts when a number drifts past a line you set. SBD builds it from your record of truth, not from exports. **What usually breaks:** - **The report takes two people a day** — Exports from four tools, pasted into a sheet, formatted, and out of date by the time it is read. - **Every number has two versions** — Sales says one figure, finance another. The meeting is spent agreeing which is right. - **Problems surface at month end** — A drifting margin or a slowing pipeline is visible in the data for weeks before anyone looks. - **Decisions wait for data** — Hiring, pricing and stock calls are made on instinct because the numbers arrive after the moment. **What we build:** - **One dashboard, from the record of truth** — Revenue, cash, pipeline, orders, margin and support load read directly from the systems that hold them, refreshed continuously. - **The weekly note, written from the numbers** — A short summary each Monday: what moved, why, and what needs a decision, drafted by an AI step from the data and reviewed by a person. - **Alerts on drift** — A line per metric. Cross it and the owner hears, with the records behind the number attached. - **Board and investor packs** — The same numbers, formatted for the people who need the summary, generated on a schedule. **Typical stack:** Google Sheets, Airtable, Supabase, Retool, Slack, Tally, Xero, QuickBooks or Zoho Books, n8n, Claude or OpenAI **FAQs:** - **Can the AI summary be trusted?** It is written from the numbers in the dashboard, cites them, and a person reviews it before it goes out. It saves the writing, not the reading. - **We have a BI tool already.** Keep it. The gap is usually the data behind it, which is exported and stale. We fix the feed and the alerts; the charts can stay where they are. - **How long before the first dashboard?** Usually inside the first two-week sprint, with the sources you already have. Alerts and the weekly note follow. ### Healthcare (industry) URL: https://smallbaydigital.com/solutions/industry/healthcare **Summary:** Healthcare practice automation means patient intake forms that write to the record, bookings with reminders and waitlist backfill that reduce no-shows, referrals and documents captured and filed by patient, and invoices or claims raised from the completed appointment. SBD builds the administrative layer around your practice management or EHR system, in your accounts, and never makes clinical decisions. **What usually breaks:** - **The front desk lives on the phone** — Bookings, reschedules, directions and the same five questions, all by phone, while patients wait at the desk. - **No-shows leave empty chairs** — Reminders are manual or missing, cancellations are not backfilled, and every empty slot is lost revenue and a longer wait for someone else. - **Intake is typed twice** — Patients fill a paper or PDF form, then staff type it into the system, and errors creep into the record. - **Referrals and documents go missing** — Referral letters, results and consent forms arrive by fax, email and post, and filing them against the right patient is a daily backlog. - **Billing waits for someone to notice** — Invoices and insurance claims depend on staff remembering after the appointment, so revenue arrives late and denials are found weeks on. - **Recalls and follow-ups slip** — Patients due for review, repeat prescriptions or annual checks are tracked on a list someone forgets to check. **What we build:** - **Digital intake that fills the record** — A secure form sent with the booking, completed on the patient's phone, validated and written to the practice system before they arrive. - **Bookings, reminders and backfill** — Confirmations and reminders by SMS, WhatsApp or email at the intervals you set, two-way replies to confirm or cancel, and the waitlist offered the freed slot automatically. - **Referral and document intake** — Letters, results and forms captured from email or a scanning inbox, matched to the patient, filed by type, with anything uncertain queued for staff. - **Billing from the appointment** — A completed appointment raises the invoice or claim with the right codes, payment links sent, and denials and overdue balances followed up on schedule. - **Recall and follow-up lists** — Patients due for review flagged from the record and invited to book, with responses tracked. - **Front desk assistant** — Common questions answered from your own policies by message or web chat, with anything clinical or personal handed to staff. **Typical stack:** Google Workspace, Microsoft 365, Twilio, WhatsApp Business API, Supabase, Retool, Stripe, QuickBooks, Xero, n8n **FAQs:** - **Do you work with our practice management system or EHR?** We connect to systems with an API or a supported integration route, and build around them. The PMS or EHR stays the record. Where a system has no API, we work from its exports and the forms that feed it. - **How do you handle patient data?** Everything runs in accounts you own, with access by role and an audit log. Where you are subject to HIPAA, GDPR or similar rules, we use vendors that sign a business associate or data processing agreement, and we sign one with you. Your compliance lead sets the requirement; we build to it. - **Will AI give medical advice?** No. Any assistant answers administrative questions from your own policies, such as opening hours, preparation instructions and billing, and hands anything clinical to staff. - **Does this work for dental, physio, mental health and specialist clinics?** Yes. The administrative pattern is the same: intake, booking, reminders, documents, billing and recalls. The forms, intervals and wording are yours. - **How long until the first part is live?** Reminders and digital intake are usually live inside the first two-week sprint, because they sit beside the practice system rather than changing it. ### Real estate (industry) URL: https://smallbaydigital.com/solutions/industry/real-estate **Summary:** Real estate automation means enquiries from portals, the website and social messages answered and qualified in minutes, viewings booked against agent calendars with reminders, each deal tracked through offer, checks and completion with documents and deadlines, and for property managers, rent collection, maintenance requests and renewals run on schedule. SBD builds it around your CRM, in your accounts. **What usually breaks:** - **Portal leads wait hours** — Enquiries arrive from several portals and the website, and the agent who replies first wins. Usually that is not you. - **Viewings are phone tag** — Booking, confirming and rescheduling viewings takes more calls than the viewing itself, and no-shows waste an afternoon. - **Deals stall between offer and completion** — Checks, documents, solicitors and deadlines are tracked in an inbox, so a missing form holds up completion for a week. - **The CRM is out of date** — Agents log what they remember, so matching buyers to new listings is guesswork. - **Maintenance requests get lost** — For property managers, requests arrive by phone, text and email, contractors are chased by hand, and tenants chase you. - **Rent and renewals rely on memory** — Arrears are noticed late, renewals are raised at the last minute, and compliance certificates expire unnoticed. **What we build:** - **Lead response in minutes** — Enquiries from every portal and the website enriched, qualified on budget, area and timing, and answered with a tailored reply and viewing slots. - **Viewing booking and reminders** — Buyers and tenants pick a slot against the agent's calendar, reminders go out, feedback is requested after, and the CRM is updated. - **Deal progression tracker** — Each deal tracked from offer through identity checks, surveys, contracts and completion, with every party updated and overdue steps flagged. - **Buyer and listing matching** — New listings matched to registered buyers and sent the same day, on the channel they chose. - **Maintenance request routing** — Requests logged with photos, triaged, assigned to a contractor, tracked to completion, and the tenant updated at each step. - **Rent, arrears and renewals** — Rent reconciled against the bank, arrears followed up by stage, renewals and certificate expiries raised in time. **Typical stack:** HubSpot, Salesforce, Pipedrive, Google Workspace, Microsoft 365, WhatsApp Business API, Twilio, DocuSign, Stripe, Xero, QuickBooks, Airtable, n8n **FAQs:** - **Which property portals and CRMs do you connect?** Any portal that sends enquiries by email feed or API, and the CRMs agencies use, including HubSpot, Salesforce, Pipedrive and property-specific systems with an API. Where a system has no API, we work from its email notifications and exports. - **Can it handle lettings and sales?** Yes. Sales follow the offer-to-completion flow; lettings add referencing, tenancy agreements, deposits, rent and renewals. - **Does it work for property managers and developers?** Yes. Property managers use the maintenance, rent and renewal flows. Developers use lead response, reservations, payment schedules and buyer updates through construction. - **How do you handle identity and anti-money laundering checks?** We connect to the checking provider you use, track each check as a step in the deal, and file the evidence. The provider does the check; we make sure it happens on time and is recorded. - **Will buyers know they are talking to automation?** The first reply is written from your templates and signed by the agent. Anything beyond scheduling and standard questions goes to the agent. ### SaaS (industry) URL: https://smallbaydigital.com/solutions/industry/saas **Summary:** SaaS operations automation joins product usage, billing, CRM and support so trial users get the right nudge at the right moment, new customers are onboarded by a sequence that tracks progress, failed payments are retried and chased, support sees plan and usage on every ticket, and revenue and churn are reported from live data. SBD builds it around Stripe, your CRM and your helpdesk. **What usually breaks:** - **Trials end without anyone noticing** — Product usage lives in one tool, the CRM in another, so the team cannot see which trial is about to convert and which went quiet on day two. - **Onboarding is a spreadsheet** — Customer success tracks setup steps by hand, so some accounts get three calls and others none. - **Failed payments become churn** — Cards expire and payments fail, and the retry email is the only follow-up. - **Support cannot see the account** — Agents ask for the email, then look up the plan, usage and billing in three tools before they can answer. - **Revenue numbers disagree** — MRR from billing, pipeline from the CRM and churn from a spreadsheet, so the board deck is assembled by hand. - **Expansion signals are missed** — Accounts hitting plan limits or adding users are not flagged to sales until they complain. **What we build:** - **Usage-triggered lifecycle messages** — Product events and usage scores drive in-app, email and sales tasks: activation nudges, quiet-account alerts and upgrade prompts. - **Onboarding that tracks itself** — Each new customer gets a plan of steps, progress read from product events, and customer success alerted when an account stalls. - **Dunning and payment recovery** — Failed payments retried on a smart schedule, customers told before cards expire, and accounts at risk flagged to a person. - **Support with account context** — Plan, usage, billing status and recent activity on every ticket, with bugs escalated to engineering and status fed back. - **Live revenue reporting** — MRR, churn, expansion and cohort retention calculated from billing and CRM data, refreshed daily. - **Expansion alerts** — Accounts nearing limits or adding seats flagged to the owner with the context to start the conversation. **Typical stack:** Stripe, HubSpot, Salesforce, Attio, Intercom, Zendesk, Linear, Jira, Slack, Supabase, PostgreSQL, n8n, Claude or OpenAI **FAQs:** - **Do you need access to our product database?** Read access to the events or tables that matter, usually through a replica, an analytics tool or a webhook, never write access to production. - **We already use a lifecycle or customer success tool. Why you?** Keep it. We feed it the data it is missing, such as billing status, CRM stage and support history, and build the steps between tools it cannot reach. - **Can you work with our engineering team?** Yes. We work in your repos and your cloud accounts, follow your review process, and keep business automation out of your product code unless you want it there. - **How do you calculate MRR and churn?** From billing data, with your definitions for discounts, annual plans, pauses and refunds written down in the blueprint, so the number matches what finance reports. - **Is this for product-led or sales-led SaaS?** Both. Product-led teams lean on usage triggers and onboarding; sales-led teams lean on handoffs, renewals and expansion alerts. ### Professional services (industry) URL: https://smallbaydigital.com/solutions/industry/professional-services **Summary:** Professional services automation means new clients onboarded from an intake form to a signed engagement letter and identity check, documents requested and chased automatically, every deadline tracked across matters with reminders, and time captured and turned into invoices at period end. SBD builds it around your practice management, document and accounting systems, in your accounts. **What usually breaks:** - **Client documents are chased by hand** — Staff send the same request three times by email before the records arrive, and the work waits. - **Intake and engagement letters take days** — A new client means forms, identity checks, conflict checks and an engagement letter, each done manually. - **Deadlines live in calendars and heads** — Filing dates, limitation periods and renewals are tracked by individuals, and a missed one is a liability. - **Time is captured late or not at all** — Timesheets are filled on Friday from memory, so billable work leaks and invoices go out late. - **Work in progress is invisible** — Principals cannot see which matters are stalled or over budget without asking. - **Client updates are ad hoc** — Clients ask for status because nobody told them, and each answer takes a search. **What we build:** - **Intake to engagement letter** — An intake form starts identity and conflict checks, generates the engagement letter from your template, sends it for signature, and opens the matter when signed. - **Document collection that chases itself** — A portal or secure link lists what each client owes, sends reminders on schedule, and files what arrives against the matter. - **Deadline tracking across matters** — Every statutory and contractual date held against its matter, with reminders to the owner and escalation as it approaches. - **Time capture to invoice** — Time captured from calendars and activity, reviewed by the fee earner, and turned into invoices or interim bills at period end. - **Matter dashboard** — Status, budget used, outstanding items and next deadline for every matter, visible to the principals. - **Client status updates** — Updates sent at each stage of a matter, so clients hear before they ask. **Typical stack:** Microsoft 365, Google Workspace, DocuSign, Xero, QuickBooks, Sage, HubSpot, Notion, SharePoint, Airtable, n8n, Claude or OpenAI **FAQs:** - **Do you work with our practice management system?** We connect to systems with an API and build around them, whether that is a legal practice management tool, an accounting practice suite or a project system. Where there is no API, we work from exports and email. - **How do you handle client confidentiality and privilege?** Everything stays in your accounts. Access is by role and logged. AI steps, where used, go through business APIs that do not train on your data, or a self-hosted model if your obligations require it. - **Can you run identity and anti-money laundering checks?** We connect to the checking provider you use, run the check as part of intake, and file the evidence against the client. The provider does the check; we make sure it is never skipped. - **Is this for accountants, lawyers or architects?** All three, and other advisory firms. The pattern is the same: intake, documents, deadlines, time, billing and client updates. The templates and deadlines are yours. - **Will it change how fee earners work?** As little as possible. Time capture reads from where they already work, document requests go out without them, and reviews happen in one screen. ## Integrations ### Shopify + Airtable URL: https://smallbaydigital.com/integrations/shopify-and-airtable-automation Category: E-commerce **Summary:** Connecting Shopify to Airtable means mirroring orders, inventory and customers into an operational base your team can work in. The hard part is not the initial push, it is keeping records correct when orders are edited, partially fulfilled, or refunded. SBD builds this with reconciliation rather than fire-and-forget triggers. **Why the obvious approach breaks:** - **A "new order" trigger only fires once** — Zapier and most native connectors push a record when the order is created and never look at it again. Shopify orders are mutable, they get edited, partially fulfilled, cancelled and refunded, so your Airtable base is accurate for about an hour and then quietly drifts. - **Retries create duplicate rows** — If a run fails midway and re-fires, a plain "create record" step makes a second copy. Without an idempotency key matched on Shopify order ID, you end up deduplicating by hand every few weeks. - **Line items do not fit one row** — An order with five line items is one Shopify object and five Airtable records in a linked table. Flat one-row-per-order sync loses per-SKU visibility, which is exactly what ops needs for picking and stock. - **Per-task pricing punishes volume** — Task-billed tools charge per operation, so cost scales with order count. At a few thousand orders a month, a purpose-built sync is usually cheaper than the automation subscription it replaces. **Data flows we implement:** - When: Order created, updated, cancelled or refunded in Shopify → Then: Upsert the order record in Airtable by Shopify order ID, and rewrite its linked line items so the base always matches current state - When: Inventory level changes in Shopify → Then: Update the SKU record and raise a reorder flag if it crosses threshold - When: Fulfilment status changes → Then: Move the order through your ops pipeline and notify the assigned owner - When: Record edited in Airtable → Then: Write the permitted fields back to Shopify, guarded so the write does not re-trigger the inbound sync **Common use cases:** Ops team working orders in Airtable instead of the Shopify admin; Per-SKU inventory visibility with reorder alerts; Fulfilment pipelines with owner assignment and SLA tracking; Reporting joined against data Shopify does not hold; Shopify order sync to Airtable; Airtable inventory management for Shopify; Shopify refunds and edits reflected in Airtable; Picking and packing lists from Airtable; Shopify customer records in Airtable CRM; Reorder alerts from live Shopify stock **Typical stack:** Shopify Admin API, Shopify webhooks, Airtable API, Make.com, n8n, Supabase **FAQs:** - **Why not just use Zapier for this?** Zapier is fine for a genuinely one-way, low-volume push. It struggles here because Shopify orders change after creation and Zapier steps are stateless, there is no built-in upsert, so edits and refunds either duplicate or go missing. It is also billed per task, so cost rises with order volume. We often keep Zapier for simple notifications and build the order sync properly alongside it. - **Can data flow both ways?** Yes, with care. Two-way sync creates feedback loops unless writes are guarded, a change in Airtable triggers a Shopify update, which fires the Shopify webhook, which updates Airtable again. We scope which fields are writable from which side and add loop protection so that cycle cannot start. - **How long does this take to build?** A one-way order and inventory sync with proper upsert handling is typically one to two weeks. Adding two-way writes, line-item tables and reorder logic usually takes it to two to three weeks. ### Shopify + eBay URL: https://smallbaydigital.com/integrations/shopify-and-ebay-automation Category: E-commerce **Summary:** Syncing Shopify and eBay means holding one master catalogue and pushing stock changes to both channels the moment anything sells. Without it, the same unit stays listed on the second channel until someone updates it manually, which is how oversells and cancelled orders happen. SBD has run this pattern across three channels with zero mismatches. **Why the obvious approach breaks:** - **Neither platform will be the master** — Point-to-point sync forces you to pick a winner, and both platforms accept orders independently. The moment two sales land close together, whichever direction you chose loses one of them. A separate master catalogue is the only arrangement that holds. - **Listing structures do not map cleanly** — eBay item specifics, category taxonomy and variation structure do not correspond one-to-one with Shopify variants. A naive field-to-field copy produces listings that are either rejected or badly formed. - **Polling is too slow for stock** — Off-the-shelf connectors often poll every 15-30 minutes. For single-unit or low-stock inventory that window is exactly when you oversell. Stock decrements have to be event-driven. - **eBay rate limits will bite at scale** — The eBay API caps calls per day per application. A sync that updates every listing on every change hits the ceiling and starts silently failing, so updates need batching and prioritisation by what changed. **Data flows we implement:** - When: Sale on either channel → Then: Decrement master stock and push the new level to every other channel immediately - When: New product added to the master catalogue → Then: Create the listing on both channels with channel-correct fields and taxonomy - When: Price or content updated centrally → Then: Propagate to both channels, respecting per-channel pricing rules - When: Stock reaches zero → Then: End or hide the listing on every channel rather than leaving it orderable **Common use cases:** Single-unit and low-stock inventory where oversell is unacceptable; Launching products to multiple channels without duplicate data entry; Channel-specific pricing from one catalogue; Adding a third marketplace without adding manual work; Shopify to eBay inventory sync; Multi-channel listing management; eBay orders into Shopify; Prevent oversells across marketplaces; Per-channel pricing from one product record; Marketplace listing automation **Typical stack:** Shopify Admin API, eBay Sell API, Airtable, Make.com, n8n, Supabase **FAQs:** - **Can you add marketplaces beyond Shopify and eBay?** Yes, the master-catalogue design is what makes that cheap. We have run Shopify, eBay and Chrono24 together, and adding a fourth channel is mostly a mapping exercise rather than a rebuild, because every channel talks to the master rather than to each other. - **What happens if a channel API goes down?** Changes queue rather than being lost, and replay when the API recovers. We also alert on sustained failure, because the dangerous failure mode is a sync that stops quietly while listings stay live. - **Do we need to change how we list products?** You list once in the master catalogue instead of separately per channel. Channel-specific fields, eBay item specifics, for example, are held as part of the mapping, so the listing is still correct on each platform. ### HubSpot + Airtable URL: https://smallbaydigital.com/integrations/hubspot-and-airtable-automation Category: CRM & Sales **Summary:** Connecting HubSpot to Airtable lets sales keep working in the CRM while delivery runs in a base built for operations. The difficulty is two-way sync: without loop protection and stable record matching, updates ping-pong between systems and duplicates accumulate. SBD builds this with a defined field ownership model on each side. **Why the obvious approach breaks:** - **Two-way sync loops without a guard** — An Airtable edit updates HubSpot, which fires a HubSpot webhook, which updates Airtable, which fires again. Without marking system-originated writes and skipping them on the return trip, this runs until something rate-limits. - **HubSpot associations have no Airtable equivalent** — Contacts, companies and deals are associated objects in HubSpot. Flattening them into one Airtable table loses the relationships; recreating them needs linked tables and a matching strategy that survives records being merged in HubSpot. - **Email is not a reliable primary key** — Most quick integrations match on email. People change addresses and HubSpot merges duplicates, so matching on email eventually attaches the wrong activity to the wrong contact. HubSpot object IDs are the only stable key. - **Nobody defines who owns which field** — When both systems can write deal stage, the last writer wins and the pipeline becomes untrustworthy. Field ownership has to be an explicit decision, not an emergent one. **Data flows we implement:** - When: Deal reaches closed-won in HubSpot → Then: Create the delivery project in Airtable with owner, scope and dates populated - When: Delivery milestone updated in Airtable → Then: Write status back to the HubSpot deal so sales can see progress without asking - When: Contact or company updated in HubSpot → Then: Update the linked Airtable records, matched on HubSpot object ID - When: Project completed in Airtable → Then: Log the activity against the deal and trigger the renewal or upsell sequence **Common use cases:** Sales-to-delivery handoff without a re-keying step; Delivery status visible in the CRM without giving delivery CRM seats; Reporting that joins pipeline data with delivery reality; Renewal triggers driven by delivery milestones; HubSpot closed-won to Airtable project; Sales to delivery handoff automation; HubSpot contacts in Airtable; Delivery status back to HubSpot; Customer onboarding from a HubSpot deal; HubSpot and Airtable two-way sync **Typical stack:** HubSpot API, HubSpot webhooks, Airtable API, Make.com, n8n **FAQs:** - **Will this work on HubSpot’s free tier?** Partly. The CRM object APIs are available on free and Starter, which covers most of the sync. Workflow-based triggers and some webhook features need higher tiers, so on lower plans we poll on a schedule instead. We confirm what your tier exposes before scoping. - **Which system should be the source of truth?** Usually HubSpot for anything customer-facing, contacts, companies, deal stage, and Airtable for delivery data HubSpot has no concept of. What matters more than the split is that it is decided explicitly per field, so no field is writable from both sides. - **What about existing duplicate records?** We deduplicate as part of the initial migration rather than syncing the mess across. That usually means a matching pass on HubSpot object IDs, plus a manual review list for genuine ambiguity, before the ongoing sync is switched on. ### Tally + Airtable URL: https://smallbaydigital.com/integrations/tally-and-airtable-automation Category: Finance & Accounting **Summary:** Connecting Tally to Airtable means extracting ledgers, vouchers and outstanding balances into a base the whole team can see, while Tally remains the accounting system of record. Tally has no cloud API, so this is done over its local XML/ODBC interface with a bridge, which is precisely why generic connectors do not offer it. **Why the obvious approach breaks:** - **Tally has no cloud API** — This is the whole problem. Tally Prime exposes an XML interface over HTTP on the local machine and ODBC on a local port, both reachable only on the LAN where Tally is running. There is no hosted endpoint for Zapier or Make to call, which is why you will not find a ready-made connector. - **It only responds while Tally is open** — The XML gateway serves requests only when the Tally application is running with the right company loaded. Any sync has to tolerate the machine being off, and must not silently report success when it simply got nothing. - **Vouchers get edited and deleted in place** — Tally permits altering and deleting past vouchers. An append-only export drifts out of step with the books within a month, so the sync needs to detect changes and deletions, not just additions. - **Master data is matched by name, not ID** — Ledger and item identity in Tally is effectively the name string. Renaming a ledger breaks naive matching, so the bridge has to maintain its own stable mapping. **Data flows we implement:** - When: Scheduled pull from the Tally XML gateway → Then: Extract ledgers, outstanding balances and vouchers for the period into staging - When: Changed or deleted voucher detected → Then: Update or tombstone the matching Airtable record so the base mirrors the books - When: Receivable crosses its credit period → Then: Raise a follow-up task and trigger the payment reminder sequence - When: Sync fails or returns nothing → Then: Alert immediately, a quiet failure that looks like "no new data" is the dangerous case **Common use cases:** Live receivables and ageing visible to sales and ops, not just accounts; Automated payment follow-ups driven by real outstanding balances; Dashboards without waiting for the month-end close; Cross-referencing dispatch records against booked invoices; Tally to Airtable sync; Tally outstanding balances in Airtable; Sales vouchers from Airtable to Tally; Tally stock levels in a dashboard; Collections follow-up from Tally receivables; Tally ERP automation for operations **Typical stack:** Tally XML/ODBC gateway, Local bridge agent, Airtable API, n8n, Supabase **FAQs:** - **Does this change anything in Tally?** No. The default setup is read-only: we pull data out and leave your books untouched. Write-back into Tally is technically possible but we generally advise against it, because the audit and correctness risk sits with your accountant, not with the integration. - **Tally runs on a machine in our office. Does that work?** Yes, and that is the normal case. We run a small bridge agent on the same network that talks to the Tally gateway locally and pushes outward over HTTPS. Nothing about Tally needs to be exposed to the internet. - **Does this work with Tally Prime and older versions?** Tally Prime is what we usually see and it is well supported. Tally.ERP 9 also exposes the XML interface, though request formats differ in places. We confirm the version and the specific reports you need during scoping. ### Shopify + QuickBooks URL: https://smallbaydigital.com/integrations/shopify-and-quickbooks-automation Category: E-commerce **Summary:** Connecting Shopify to QuickBooks means posting sales, fees, refunds and payouts so the ledger reconciles against money received. Most integrations post the order total and ignore payment processing fees and payout batching, which is why the bank line never matches. SBD builds it around the payout, not the order. **Why the obvious approach breaks:** - **The order total is not what hits the bank** — Shopify Payments deducts fees and batches many orders into one payout. If you post gross order totals, every payout line has to be reconciled by hand forever. The payout is the unit that has to be modelled. - **Refunds and partial refunds post as new events** — A refund is not a deleted order. It needs its own credit entry linked to the original, and partial refunds need to allocate correctly across line items and tax. - **Tax handling is not a single field** — Multi-jurisdiction tax, and GST in India, do not map to one QuickBooks tax code. Getting this wrong is worse than no integration, because the books look reconciled while being wrong. - **High order volume needs summarisation** — Thousands of individual invoices will bloat QuickBooks and slow it down. Most businesses want daily or payout-level summary journals, with detail retained elsewhere for drill-down. **Data flows we implement:** - When: Shopify payout batch closes → Then: Post a journal splitting gross sales, processing fees and net deposit - When: Order paid → Then: Record the sale against the correct income account and tax code - When: Refund issued → Then: Post a linked credit note allocated across the original line items - When: Reconciliation mismatch detected → Then: Flag for review rather than forcing a balancing entry **Common use cases:** Bank reconciliation that matches without manual adjustment; Accurate margin reporting net of processing fees; Refunds and chargebacks reflected correctly in the ledger; Summarised journals instead of thousands of individual invoices; Shopify payouts to QuickBooks; Shopify to QuickBooks reconciliation; E-commerce bookkeeping automation; Shopify fees and refunds in QuickBooks; Summary journals per payout; Shopify sales tax in QuickBooks **Typical stack:** Shopify Admin API, Shopify Payments payouts, QuickBooks Online API, n8n, Make.com **FAQs:** - **Should every order become an invoice in QuickBooks?** Usually not. Past a few hundred orders a month, per-order invoices bloat the file and slow everything down without adding insight. We typically post summary journals per payout or per day, and keep line-level detail in the operational system where drill-down happens. - **Can you handle tax correctly?** Yes. Tax mapping, whether GST, VAT or sales tax, is part of scoping rather than an afterthought, because tax posted to the wrong code is worse than not posting it at all. We confirm the treatment with your accountant before anything goes live. - **What about historical orders?** We usually start the sync from an agreed cutover date rather than backfilling years of history, since backfilled entries can conflict with periods already closed and filed. Historical backfill is possible if your accountant wants it. ### Zoho Books + Airtable URL: https://smallbaydigital.com/integrations/zoho-books-and-airtable-automation Category: Finance & Accounting **Summary:** Connecting Zoho Books to Airtable lets invoices be raised automatically from operational events such as dispatch or project completion, and pulls receivables back so the team can see what is outstanding. Zoho Books rate-limits by plan and scopes tokens per organisation, which is where most quick integrations come apart. **Why the obvious approach breaks:** - **API call limits are per plan, not generous** — Zoho Books enforces a daily API cap tied to your subscription. A naive integration that re-reads every invoice on a schedule burns the quota before lunch and then fails for the rest of the day. - **OAuth tokens are scoped per organisation** — Multi-entity setups need a token per organisation ID, and refresh tokens expire on re-authorisation. Integrations that assume one static credential break the first time somebody re-auths. - **Invoice numbering is controlled by Zoho** — Sequences are managed inside Books and often have statutory formatting. Generating numbers on the Airtable side produces gaps and duplicates that create real compliance problems. - **Payments arrive without clean references** — Bank payments rarely carry the invoice number, so matching is fuzzy by nature. Auto-applying every payment to the nearest invoice creates errors that surface at audit. **Data flows we implement:** - When: Dispatch or project milestone marked complete in Airtable → Then: Create the invoice in Zoho Books and write the returned number back to the record - When: Invoice status changes in Zoho Books → Then: Update the Airtable record so ops sees payment state without asking finance - When: Invoice passes its due date → Then: Trigger an escalating reminder sequence that stops when payment is recorded - When: Payment received with no clear reference → Then: Queue for human matching instead of guessing **Common use cases:** Invoicing triggered by real operational events, not a weekly billing run; Receivables and ageing visible to the whole team; Automated, escalating payment follow-ups; Fewer disputes because ops and finance read the same record; Zoho Books to Airtable sync; Airtable orders to Zoho invoices; Zoho Books payments in Airtable; Overdue invoice follow-up automation; Zoho Books customer sync; Approval workflow before invoicing **Typical stack:** Zoho Books API, Zoho OAuth, Airtable API, Make.com, n8n **FAQs:** - **Can this create tax-compliant invoices?** Yes, because Zoho Books creates them, not us. We pass the line items, tax treatment and place of supply, and Books applies your configured tax settings, whether GST, VAT or sales tax, and your numbering. Keeping generation inside Books is deliberate: that is where the statutory formatting and sequence live. - **Will we hit API limits?** Not if the integration is built for them. We use webhooks where Zoho offers them and incremental queries rather than full re-reads, so call volume tracks actual changes instead of table size. - **Does this work with Zoho CRM as well?** Yes, and it is a common pairing, CRM for the pipeline, Books for billing, Airtable for delivery. The same credential and rate-limit design applies across the Zoho suite. ### WhatsApp + Airtable URL: https://smallbaydigital.com/integrations/whatsapp-and-airtable-automation Category: Operations & Internal Tools **Summary:** Connecting WhatsApp to Airtable lets you send order updates, payment reminders and delivery notifications from your operational base, and log replies back against the right record. It has to run on the official WhatsApp Business API, where templates need pre-approval and business-initiated messages fall outside the 24-hour window. **Why the obvious approach breaks:** - **Unofficial bridges get numbers banned** — Tools that automate WhatsApp Web violate the terms of service. They work until the number is blocked, which tends to happen once volume is high enough to matter, taking your primary business number with it. - **The 24-hour window governs everything** — Outside 24 hours of the customer’s last message, you can only send pre-approved template messages. Any design that assumes you can send free-form text whenever you like will fail in production. - **Templates need approval before they exist** — Every business-initiated message must be submitted to Meta and approved, and variables are positional placeholders. You cannot compose arbitrary copy at send time, so message design has to happen up front. - **Replies arrive with no record context** — An inbound message carries a phone number and nothing else. Matching it to the right order or contact, and to the right one when a number has several open orders, is the actual engineering work. **Data flows we implement:** - When: Order status changes in Airtable → Then: Send the approved template with order details filled in - When: Invoice becomes overdue → Then: Send a payment reminder and log the attempt against the invoice record - When: Customer replies on WhatsApp → Then: Match to the open record by phone number and append the message to its thread - When: Reply needs a human → Then: Assign to an owner and notify, within the 24-hour free-form window **Common use cases:** Dispatch and delivery notifications customers read; Payment reminders on a channel with far higher response than email; Conversation history attached to the order, not stuck on one person’s phone; Order status requests handled without a phone call; WhatsApp order status notifications; WhatsApp Business API automation; WhatsApp messages logged in Airtable; Automated WhatsApp reminders from Airtable; WhatsApp customer support with a record; WhatsApp opt-in and opt-out tracking **Typical stack:** WhatsApp Business API, Meta Cloud API, Twilio, Airtable API, Make.com, n8n **FAQs:** - **Do we need the official WhatsApp Business API?** For automated sending, yes. The consumer and Business app are not built for it, and unofficial automation risks the number being banned. The official API is accessed through Meta Cloud API or a provider such as Twilio, and involves business verification plus per-conversation pricing. - **What does it cost to run?** Meta bills per conversation window rather than per message, with rates varying by country and category, utility messages like order updates are cheaper than marketing. For most operational use the cost is modest, but it is a real per-message line item rather than free. - **Can customers reply and reach a person?** Yes. Replies land in the record and can be routed to an owner. Within 24 hours of a customer message you can respond in free-form text, which is usually enough for operational conversations. ### Airtable + Retool URL: https://smallbaydigital.com/integrations/airtable-and-retool-automation Category: Operations & Internal Tools **Summary:** Pairing Airtable with Retool keeps Airtable as the operational database while Retool provides the interface, role-scoped screens, validated forms and bulk actions that a grid view cannot do. The main constraints are Airtable’s API rate limit and record ceiling, which shape how the Retool app should query. **Why the obvious approach breaks:** - **Five requests per second, per base** — Airtable rate-limits at 5 requests/second/base and returns 429s beyond it. A Retool app where each component queries independently trips this with only a handful of concurrent users, so queries have to be consolidated and cached. - **Airtable permissions are per base, not per row** — You cannot grant a user access to only their own records natively. Anyone with base access sees everything. Row-level scoping has to be enforced in the Retool layer, driven by the logged-in user. - **Record limits arrive sooner than expected** — Per-base record caps depend on plan, and performance degrades before the hard ceiling. High-volume tables such as event logs usually belong in Postgres or Supabase, with Airtable holding the records humans edit. - **No transactions across tables** — Airtable has no multi-record transaction. A Retool action updating several linked tables can half-succeed, so writes need ordering and a compensating path when a later step fails. **Data flows we implement:** - When: User opens a Retool screen → Then: Query only their permitted records via a consolidated, cached request - When: Form submitted in Retool → Then: Validate, then write to Airtable with the acting user recorded for audit - When: Bulk action run → Then: Batch into chunked writes that respect the rate limit rather than firing in parallel - When: Write fails mid-sequence → Then: Roll back or flag the partial state instead of leaving records inconsistent **Common use cases:** Role-scoped views so each team sees only its own records; Validated data entry instead of free typing into a grid; Bulk approvals and status changes in one action; Customer or vendor portals over Airtable data; Retool internal tool on Airtable; Airtable approval workflow in Retool; Role-based access for Airtable data; Field app on top of Airtable; Airtable to Supabase for high volume; Internal portal for operations **Typical stack:** Airtable API, Retool, Supabase, Postgres, n8n **FAQs:** - **When should we move off Airtable entirely?** When record volume, rate limits or the lack of row-level permissions start dictating your design rather than supporting it. A common middle path is keeping Airtable for the records humans edit and moving high-volume or sensitive tables to Supabase, with Retool querying both. - **Can external users access a Retool app?** Retool is licensed and priced for internal users. For customer or vendor portals we normally build on Softr, Bubble or a custom front end instead, and keep Retool for the internal side. - **Does this replace Airtable automations?** It complements them. Airtable automations are fine for simple in-base triggers. Retool covers the interface, and anything crossing systems generally belongs in n8n or Make where retries and error handling are properly supported. ### Pipedrive + Airtable URL: https://smallbaydigital.com/integrations/pipedrive-and-airtable-automation Category: CRM & Sales **Summary:** Connecting Pipedrive to Airtable automates the handoff from a won deal to delivery, and reports progress back into the CRM. The awkward part is Pipedrive’s custom fields: they are addressed by generated hash keys rather than names, so integrations built against field labels break the moment a field is edited. **Why the obvious approach breaks:** - **Custom fields are hash keys, not names** — Pipedrive exposes custom fields as generated hash identifiers. Integrations mapped against display names break silently when a field is renamed or recreated, and the failure looks like empty data rather than an error. - **Pipeline stages are IDs that get reordered** — Stage IDs are stable but the pipeline they belong to is not. Teams reorder and rename stages regularly, so logic keyed to stage position rather than ID misfires after any pipeline edit. - **Deal, person and organisation are separate objects** — A handoff needs all three plus their relationships. Fetching only the deal produces a delivery record missing the contact details the delivery team needs. - **Won is not always the right trigger** — Deals get marked won prematurely, then reopened. Firing delivery setup on the status change alone creates projects that should not exist, so a confirmation step or delay is usually warranted. **Data flows we implement:** - When: Deal marked won in Pipedrive → Then: Create the delivery project in Airtable with contact, organisation and deal value - When: Delivery status updated in Airtable → Then: Write a note or custom field back to the Pipedrive deal - When: Person or organisation updated in Pipedrive → Then: Sync to the linked Airtable records, matched on Pipedrive object ID - When: Project completed → Then: Log the activity and create the renewal opportunity **Common use cases:** Automatic delivery setup on deal close; Sales visibility into delivery without CRM seats for the delivery team; Reporting joined across pipeline and delivery; Renewal opportunities created from delivery milestones; Pipedrive won deal to Airtable project; Pipedrive to Airtable sync; Sales to delivery handoff from Pipedrive; Pipedrive activities in Airtable; Delivery status back to Pipedrive; Pipedrive CRM automation **Typical stack:** Pipedrive API, Pipedrive webhooks, Airtable API, Make.com, n8n **FAQs:** - **How do you handle Pipedrive custom fields safely?** We resolve field hash keys at build time and store the mapping explicitly, rather than matching on labels. We also validate the mapping on each run, so a renamed or deleted field raises an alert instead of quietly syncing blanks. - **Can we trigger on something other than deal won?** Yes. Any stage change, activity, or custom field update can be the trigger. In practice a short delay or an explicit confirmation field is safer than firing on won, because deals do get reopened. - **Is this different from Pipedrive’s own automations?** Pipedrive automations work well inside Pipedrive. They cannot create structured, linked records in Airtable or handle retries across systems, which is where a dedicated integration layer earns its place. ### HubSpot + Slack URL: https://smallbaydigital.com/integrations/hubspot-and-slack-automation Category: CRM & Sales **Summary:** Connecting HubSpot to Slack routes meaningful CRM events, a deal stage change, a high-value form fill, a stalled opportunity, to the right channel or person with enough context to act on. The failure mode is volume: notify on everything and the channel gets muted, which is worse than no integration. **Why the obvious approach breaks:** - **The default is a firehose** — Native and templated setups notify on every deal and every form. Within two weeks the channel is muted, and the one notification that mattered is lost with the rest. Filtering is the entire design problem. - **A notification without context is useless** — "Deal updated" forces someone into HubSpot to find out what changed. A useful message carries the value, owner, previous stage and the reason, which means enriching from several objects before posting. - **Threading gets lost** — Posting each update as a new message scatters one deal’s history across the channel. Keeping updates threaded to the original requires storing the Slack message timestamp against the HubSpot record. - **Slack rate limits apply per workspace** — Bulk pipeline updates can generate a burst of posts that hit Slack’s per-method limits, so messages need queueing and coalescing rather than one post per event. **Data flows we implement:** - When: Deal crosses a value threshold or reaches a key stage → Then: Post an enriched message to the right channel with owner and next step - When: High-intent form submitted → Then: Notify the assigned rep directly with contact context and source - When: Deal stalls beyond its stage SLA → Then: Nudge the owner in a thread rather than broadcasting to the channel - When: Reply or action taken in Slack → Then: Log the activity back against the HubSpot record **Common use cases:** Alerts for deals that warrant attention; Instant routing of high-intent inbound leads; Stalled-deal nudges tied to stage SLAs; Sales activity logged without leaving Slack; HubSpot deal alerts in Slack; HubSpot to Slack notifications with context; Stalled deal nudges; High-intent lead alerts to the owner; Slack replies logged to HubSpot; Sales pipeline updates in Slack threads **Typical stack:** HubSpot API, HubSpot webhooks, Slack Web API, Slack Block Kit, n8n, Make.com **FAQs:** - **How is this better than the official HubSpot Slack app?** The official app is a reasonable starting point for simple notifications. It is limited on conditional routing, on enriching a message from several objects before posting, and on writing actions back to HubSpot. If the native app covers your case, use it, we build this when the filtering logic is the point. - **Can people act on records from Slack?** Yes. Block Kit supports buttons and modals, so updating a stage or logging a note from the message is straightforward, with the write going back through the HubSpot API. - **How do we stop notification fatigue?** By deciding up front what genuinely deserves an interrupt, usually value thresholds, specific stages and SLA breaches, and sending everything else to a digest or nowhere. We tune this after a couple of weeks of real traffic rather than guessing at the start. ## Case Studies ### AI outbound sales platform — IntellReach AI URL: https://smallbaydigital.com/case-studies/intellreach-ai-outbound-platform Tags: AI Agent, SaaS, Sales Results: Product areas 11, Tenant isolation Every row, Scheduled jobs 3 An outbound sales platform that ingests contacts, writes personalised emails from a knowledge base, runs follow-up sequences, handles replies and bounces, and hands interested prospects to a person. ### Part finder and product catalogue — SBEE Cables URL: https://smallbaydigital.com/case-studies/sbee-cables-part-finder Tags: Manufacturing, Internal Tool, Catalogue Results: Article numbers 6,000+, Product categories 39, Lookup tables 24 An internal platform for a cable manufacturer that decodes positional article numbers across the catalogue, finds the right part in four steps, and turns enquiries into documents. ### Property operations automations and agent fleet — Ruyi Flux URL: https://smallbaydigital.com/case-studies/rental-portfolio-automations Tags: Real Estate, n8n, AI Agents Results: Automations live 6, Properties ~120, Agents in build 12 Six fixed-scope automations for a US residential rental portfolio of around 120 properties, followed by an agent fleet with a control plane, decision queue and policy engine now in build. ### Enquiry and quotation platform on SAP — Revotec URL: https://smallbaydigital.com/case-studies/rfq-to-quotation-on-sap Tags: Distribution, AI, SAP Results: RFQs Read by AI, Parts Matched to master, ERP SAP An enquiry platform for an industrial parts distributor that reads emailed RFQs with AI, matches parts against the item master, and works with SAP from enquiry to order. ### Internal operations tool replacing SmartSuite — OneMagnate Consulting URL: https://smallbaydigital.com/case-studies/consulting-internal-tool Tags: Professional Services, Internal Tool, Modernise Results: Modules 9, Roles Server-enforced, Shared logins 0 A purpose-built operations tool for a consulting firm, replacing a SmartSuite workspace with quotations, brands, architect firms, contracts, projects and roles enforced on the server. ### Multi-user production dashboard — Copper wire manufacturer URL: https://smallbaydigital.com/case-studies/manufacturer-dashboard Tags: Manufacturing, Web App, API Integration Results: Processing Time -45%, Data Accuracy +60%, Modules 10+ Built a central dashboard for a copper wire manufacturer to unify data tracking, order movement insights, and production stage monitoring, replacing scattered manual systems. #### The Challenge A copper wire manufacturer had a complex and manual process for tracking production data, resulting in inefficiencies and delays in order fulfilment. Disjointed data systems led to errors, manual tracking caused delays in the dispatch process, and there was high dependency on team members for real-time updates. #### What We Built We built a multi-user dashboard on Bubble.io that gives the team live visibility into every production stage. Cloud integration through Cloudflare Workers handles report generation, and PrintNode API handles direct printing, no manual steps in between. The system includes modules for production data, input/output stages, drawing stages, order management, cashback processing, and dispatch tracking, all accessible through role-based secured access. #### Solutions Delivered - Workflow Automation across production stages - Custom web app with secured multi-user access - Real-time data dashboard with production metrics - Automated report generation and direct printing - API integrations with Cloudflare Workers, Render, and PrintNode - Business process mapping and database architecture #### The Results 45% reduction in order processing time. 60% improvement in data accuracy. Enhanced traceability across all production stages. The team now has real-time visibility into every stage of production without depending on manual updates. #### Behind the Scenes This project required on-site visits to the factory floor, watching every stage of production happen in real time. We sat with operators, floor managers, and dispatch teams across multiple rounds to map out where things actually broke down, not just where management thought they did. ### Sales channel inventory sync — D2C watch reseller URL: https://smallbaydigital.com/case-studies/ecommerce-product-sync Tags: E-Commerce, Workflow Automation, API Integration Results: Time Saved 80%, Mismatches 0, Launch Speed 2x Unified product data across Shopify, eBay, and Chrono24 for a fast-growing D2C reseller, eliminating manual updates and inventory mismatches across platforms. #### The Challenge A fast-growing D2C watch reseller was managing operations across multiple online sales channels. Each platform had its own format for product attributes (titles, prices, availability), which required separate manual updates, leading to operational overload and delays in launching new products. Key pain points: no single source of truth for inventory and product data, each sales channel had its own formatting rules and catalogue standards, product updates had to be done manually on each platform, high risk of out-of-stock products still being listed, and no visibility into what's live where. #### What We Built A custom backend tool built on SmartSuite that gives the business one control panel for all product operations. Update a listing once, and it pushes to every storefront in real time via Make.com automation workflows. The system includes product CRUD workflows, sales channel sync from Shopify to SmartSuite, and integrations with Shopify APIs, eBay APIs, HubSpot APIs, and Chrono24 APIs via Missive App. #### How the Sync Works SmartSuite is the single source of truth. Every product lives there once, with a unified schema covering title, price, condition, availability and images. When a record changes, a Make.com scenario picks it up and translates it into each channel's format before pushing it out, so a price change or a sale on one storefront reaches every other storefront without anyone touching it. The translation layer is where the real work sits. Each platform enforces its own title lengths, category taxonomies and image rules, so the sync applies per-channel mapping, resizes images and generates platform-specific metadata on the way out. New channels plug into the same schema rather than needing a parallel process. #### Solutions Delivered - Single product database on SmartSuite as the source of truth - Product create, update and delete workflows pushed to every channel - Shopify-to-SmartSuite sales sync so stock reflects real orders - Integrations with Shopify, eBay, Chrono24 and HubSpot APIs - Per-platform field mapping, image resizing and meta generation - A clean, tagged catalogue built from the client's existing listings #### The Results 80% reduction in time spent updating product details across platforms. Zero inventory mismatch cases reported in the last 90 days. 2x faster product launch time. Team can now launch across platforms in under 20 minutes vs. 2-3 hours before. #### Behind the Scenes We mapped every platform's quirks with the client team, eBay's listing rules differ wildly from Shopify's and Chrono24's. We reverse-engineered their existing listings to clean, tag, and unify them under one schema flexible enough for new channels later. Custom logic handles image resizing and meta generation per platform. ### Order and dispatch management system — Building materials supplier URL: https://smallbaydigital.com/case-studies/order-management-app Tags: Distribution, Internal Tool, Retool Results: Time Saved 40%, Stakeholders 5+, Invoices Auto Digitized the entire order, enquiry, and dispatch management process for a building material supplier, replacing Excel tracking with a custom Retool app and automated WhatsApp notifications. #### The Challenge The client managed their orders and enquiries through a manual system with some Excel tracking, leading to inefficiencies and difficulty in managing stakeholders' needs. Multiple stakeholders (transporters, warehouse staff) needed coordination, orders needed accurate tracking to avoid delays, and there was a lack of real-time updates for orders and dispatches. #### What We Built Using Retool, AI Sensy (WhatsApp API), and Zoho Books API, we built a custom order management system around their actual workflow. It handles enquiries, orders, and dispatch end-to-end, with real-time WhatsApp notifications keeping every stakeholder in the loop. The system includes CRUD functionality for all entities, automated invoice generation via Zoho Books API, WhatsApp notifications via AI Sensy for order updates, and role-based views for different stakeholder types. #### How the Workflow Runs An enquiry is logged in Retool the moment it arrives and converts into an order without re-entering anything. As the order moves through confirmation, warehouse picking and dispatch, each status change triggers a WhatsApp message through AI Sensy to whoever needs to act next, the transporter, the warehouse, or the customer. When an order is confirmed, the invoice is generated directly in Zoho Books through its API, so accounts works from the same record as operations instead of a copied spreadsheet. Each stakeholder type sees only the views relevant to their part of the job. #### Solutions Delivered - Custom Retool app covering enquiries, orders and dispatch - Enquiry-to-order conversion without duplicate data entry - Automated WhatsApp status notifications via AI Sensy - Invoice generation through the Zoho Books API - Role-based views for transporters, warehouse and accounts - Full replacement of the Excel tracking sheets #### The Results 40% time savings by automating manual tasks. Improved communication with all stakeholders, reducing delays. Full visibility into each stage of the order and dispatch process. Automated invoice generation eliminated manual accounting errors. #### Behind the Scenes We spent time at the warehouse and office watching the order-to-dispatch lifecycle play out. Talked to transporters, warehouse staff, and the accounts team separately, each had a different version of "the process." The final system reflects how they actually work, not a sanitized version of it. ### Automated cashback and loyalty system — Industrial manufacturer URL: https://smallbaydigital.com/case-studies/customer-loyalty-system Tags: Manufacturing, Rewards, Make.com Results: Sales Lift 5-10%, Engagement 3x, Payouts Auto Built a QR-based customer loyalty and cashback system for an industrial manufacturer to establish direct relationships with end users and increase brand stickiness. #### The Challenge An industrial manufacturer had no direct relationship with end users, all transactions happened via distributors, cutting the manufacturer off from valuable end-user insights. Without customer data, it was impossible to segment, target, or retain actual users. End users often defaulted to distributor preferences, limiting brand loyalty. #### What We Built A lightweight yet powerful campaign infrastructure including: QR-based cashback system deployed in product packaging, mobile-first forms for data capture, automated payout system linked with UPI platforms via Razorpay X, and a backend database to manage and analyze customer data. Built on Make.com with Cloudflare Workers, Razorpay X Payout APIs, Google Sheets, and WP Fluent Forms for the customer-facing interface. #### The Results Improved customer loyalty and increase in sales by 5 to 10%. Direct engagement with new end-user customers that was previously impossible. Automated cashback payouts eliminated manual processing. The system is being expanded across more product lines with tiered cashback for premium SKUs. #### Behind the Scenes We understood their target customer base and built the form in multiple local languages, keeping the UI super simple. We aligned with the expectation of the cashback processing and ensured full visibility into the build and data flow. ### Lead enrichment and sales automation — B2B sales team URL: https://smallbaydigital.com/case-studies/lead-enrichment-sales-automation Tags: Lead Management, AI Agent, CRM Automation Results: Manual CRM Entry -90%, Response Time <3 min, Conversion Rate 3.4x Automated the entire inbound lead path, capture, enrichment, AI qualification, CRM sync, follow-up and meeting booking, for a team drowning in manual context-switching across fragmented lead sources. #### The Challenge Leads arrived from everywhere, website forms, cold email replies, social DMs, and the sales team handled follow-up, qualification and CRM entry by hand. The result was delayed responses, inconsistent lead scoring, and conversions lost simply because nobody got to them in time. There was no unified qualification system, no real-time view of pipeline health, and CRM hygiene depended on whoever remembered to update it. #### What We Built A single trigger layer captures leads from web forms, cold email replies and scraped DM responses via webhook and IMAP. Each lead is then enriched automatically through Apollo, Clearbit, BuiltWith and LinkedIn scraping via Apify. A GPT-4 qualification step matches each lead against the ICP and predicts likely objections before a human ever reads it. Qualified leads sync to the CRM with tagged fields and status, follow-up tasks are created in Notion weighted by lead value, and a personalised follow-up email goes out through the Gmail API. Meeting scheduling runs through Cal.com with confirmation tracking, and Fireflies call summaries sync back to the CRM and Slack, closing the loop between the call and the record. #### Solutions Delivered - Unified lead capture across web forms, email and social - Automated enrichment via Apollo, Clearbit, BuiltWith and Apify - AI qualification scoring leads against the ICP - CRM auto-update with tagged fields and pipeline status - Value-weighted follow-up task creation - Personalised follow-up email and Cal.com meeting booking - Call summary logging back into CRM and Slack #### The Results 90% reduction in manual CRM entry. Average response time under three minutes. AI-matched ICP scoring improved sales conversion rate by 3.4x, and the team booked 40% more qualified calls, with a closed-loop sync between sales, CRM, email and meeting tools. #### What's Next AI-generated proposals based on lead industry and historical wins, auto-classification of churned versus converted leads through CRM tagging, and re-engagement drip workflows for leads that go quiet after seven days. ### Lightweight project management system — E-commerce photo studio URL: https://smallbaydigital.com/case-studies/photo-studio-project-management Tags: Workflow Automation, Task Management, Custom Forms Results: Time Saved 60%, Team Size 5, Software Cost Near-zero Built a full project and task management system on Google Sheets for a five-person photo studio that needed real capability without the cost of enterprise project software. #### The Challenge A studio doing e-commerce product photography and videography ran every lead enquiry, project deliverable and team task offline or from memory. With five people and no digital system, work was tracked by whoever happened to remember it. The constraint that shaped the whole project was budget: as a small early-stage team, they could not justify high-cost project management software. But their requirements were not simple, they needed real task tracking, live status, and client notifications. #### What We Built Rather than push them onto a platform they could not afford, we built the system on Google Sheets and kept operational cost close to zero. A Fillout lead enquiry form feeds project data straight into Sheets. Each team member gets an individual task sheet that syncs automatically to an admin dashboard, so project status updates in real time without anyone reporting it. Make.com handles backend orchestration, with Google Apps Scripts and Cloudflare Workers taking the heavier processing to keep Make.com operation counts, and therefore cost, down. #### Solutions Delivered - Lead enquiry form integrated directly with Google Sheets - Per-member task sheets syncing to a central admin dashboard - Make.com backend automation for workflow orchestration - Google Apps Scripts and Cloudflare Workers for heavier processing - Centralised dashboard covering projects, deliverables and client data #### The Results 60% time saving on previously manual tasks, materially better communication with every stakeholder, and full visibility into each stage of a project, achieved on infrastructure that costs almost nothing to run. #### What's Next A move to SmartSuite or Retool as the studio scales, AI-assisted prediction of delivery timelines, and a client portal for approving deliverables and handling payment. ### Retailer catalogue and order portal — Wholesale jeweller URL: https://smallbaydigital.com/case-studies/jeweller-catalogue-portal Tags: Client Portal, Web App, WhatsApp API Results: Order Processing 30% faster, Client Inquiries -50%, Access Approved only Digitised order processing and catalogue management for a wholesale jeweller, with gated retailer access and automated WhatsApp updates replacing manual customer communication. #### The Challenge A wholesale jeweller processed orders manually and updated customers by hand, which produced both delays and miscommunication. Because they sell to retailers rather than the public, the platform also had to be genuinely gated, only approved accounts could see the catalogue or place an order. Two things made it harder: the inventory changes constantly, so the catalogue could not be static; and clients expected updates in real time rather than when someone got round to sending them. #### What We Built An order management system on Bubble.io with approved-user access control, integrated with the WhatsApp API so order status updates send themselves. The build includes inventory management, order tracking, and a communication path that works for both retailers and back-office staff. Elastic Email handles transactional notifications alongside WhatsApp, and the API layer was tested through Postman before going live. #### Solutions Delivered - Gated client portal restricted to approved retailers - Live catalogue management for changing inventory - Automated WhatsApp order and dispatch notifications - Inventory tracking and order status management - Custom UI/UX designed around how retailers actually order #### The Results 30% faster order processing with direct client communication, and a 50% drop in inbound client enquiries, because customers no longer had to ask where their order was. Inventory tracking now runs against a catalogue that stays current. #### Behind the Scenes We visited the showroom, warehouse and back office to watch the process run in real time, and interviewed team members across departments so the system reflected what every stakeholder actually needed, not just what the brief described. ## General FAQs **What do you build?** Automations, AI agents, internal tools and portals, integrations, and document and finance workflows. Across sales, finance, support, operations and people teams. We also take over vibe-coded apps and make them production-grade. **Which tools do you build on?** The ones you already run where we can: your CRM, accounting software, store and helpdesk. For the system itself we use Zapier, Make, n8n, Airtable, Retool, Supabase, Vercel, and Claude or OpenAI through their business APIs. Every account is in your name. **How long until something is live?** The Automation Audit takes one to two weeks. Builds run in two-week sprints and go live in 2 to 8 weeks, with working software at every second Friday review. Small automations are often running inside the first sprint. **What does it cost?** The Automation Audit is $500 and is credited in full to your build. Builds are fixed scope and fixed price from $2,500, with 60 days of hypercare included. Care plans start at $1,000 a month. Nothing is billed by the hour. **Can you connect to our existing systems?** Yes. We integrate through APIs with tools such as HubSpot, Salesforce, Shopify, QuickBooks, Xero, SAP, Tally, Zendesk, Slack and Microsoft 365, and with in-house systems that have an API or a database we can read. **Who owns what we build?** You do. Admin access, billing, app connections, code and documentation sit in accounts in your name. If you leave, it keeps running. That is written into every statement of work. **What happens after go-live?** Sixty days of hypercare are included: the engineers who built it stay on call, fix what breaks and tune what the team asks for. After that you can take a care plan or run it yourself with the handover documentation. ## Blog Articles ### Does an AI agent still need Zapier? URL: https://smallbaydigital.com/blog/does-an-agent-still-need-zapier Published: 2026-10-20 Tags: AI agents, Zapier, Integrations, Strategy A widely shared post argues Zapier goes to zero because agents click buttons and call APIs themselves. We run both every day. Where the argument is right, where it is wrong, and what we use for each job. ### How much should an AI agent be allowed to do? URL: https://smallbaydigital.com/blog/how-much-should-an-ai-agent-be-allowed-to-do Published: 2026-10-20 Tags: AI agents, Security, Permissions An agent with an unscoped token dropped a production database and its backups in nine seconds. The permission model we use for every agent: read widely, write narrowly, spend with a cap, and escalate when unsure. ### Invoice intake with an agent: the first thirty days URL: https://smallbaydigital.com/blog/invoice-intake-with-an-agent-30-day-log Published: 2026-10-20 Tags: Finance, AI agents, Document processing What the first month of an invoice intake agent looks like from the inside: how the queue behaves, which invoices escalate, what the finance team changes, and the three settings that decide whether it survives month two. ### n8n self-hosted or cloud for a team under 50? URL: https://smallbaydigital.com/blog/n8n-self-hosted-vs-cloud-for-a-team-under-50 Published: 2026-10-20 Tags: n8n, Pricing, Infrastructure Self-hosting n8n costs a few dollars a month for the server and a lot more for the person who keeps it up. The break-even in executions, what you take on when you host, and the answer we give most teams of 1 to 50. ### Six months of AI agents in production: what broke URL: https://smallbaydigital.com/blog/six-months-of-agents-in-production Published: 2026-10-20 Tags: AI agents, Operations, Lessons Not the model. Field renames, a confidence score that drifted, a supplier who changed invoice format, and a Slack channel nobody read. What failed across the agent systems we run for clients, and what we changed. ### The 95% figure, read properly URL: https://smallbaydigital.com/blog/the-95-percent-figure-read-properly Published: 2026-10-20 Tags: Strategy, Research, AI adoption MIT says 95% of generative AI pilots show no profit impact. The study was 52 interviews, 153 survey responses and 300 public deployments. What it measured, what it did not, what the other 5% did, and why we quote it anyway. ### We tried AI and nothing stuck URL: https://smallbaydigital.com/blog/we-tried-ai-and-nothing-stuck Published: 2026-10-20 Tags: AI adoption, Operations, Strategy The sentence we hear on most first calls. It is almost never about the AI. The four reasons tools get abandoned in small companies, how to tell which one you have, and the one change that makes the next attempt stick. ### Who noticed when the workflow stopped? URL: https://smallbaydigital.com/blog/who-noticed-when-the-workflow-stopped Published: 2026-10-20 Tags: Monitoring, n8n, Zapier, Operations Most automation failures are silent. The run shows green, the trigger is armed, and nothing happens for six weeks. How silent failure works in Zapier and n8n, what it costs, and the four checks we put on every build. ### Your Lovable app is probably public URL: https://smallbaydigital.com/blog/your-lovable-app-is-probably-public Published: 2026-10-20 Tags: Vibe coding, Supabase, Security, MVP audit Over 170 Lovable-built apps had open databases in one disclosure. Most vibe-coded apps ship with row-level security off. How the leak works, the three named breaches of 2026, and the twenty-minute check to run before you tell anyone your app exists. ### The Zapier tax in 2026: what a 20-person team pays URL: https://smallbaydigital.com/blog/the-zapier-tax-in-2026 Published: 2026-10-20 Tags: Zapier, n8n, Pricing, Automation Zapier now bills tasks, AI steps by model tier, and Agents in activities. A worked bill for a 20-person company, what the same workflows cost on n8n, and the rule Zapier itself now recommends to cut the bill. ### Stop building Zaps. Start building agents. URL: https://smallbaydigital.com/blog/stop-building-zaps-start-building-agents Published: 2026-01-04 Tags: AI agents, Automation, Operations Fifty disconnected automations are a liability, not a system. How to move from triggers and actions to agents that hold state, handle exceptions and get looked after. ### The 2026 AI readiness checklist URL: https://smallbaydigital.com/blog/ai-readiness-checklist-2026 Published: 2026-01-02 Tags: Strategy, Checklist, AI adoption Ten questions across data, process, tools, team and strategy. Score your business before you spend on AI, and find out whether to build now or fix the foundations first. ### The first eight weeks with SBD URL: https://smallbaydigital.com/blog/ai-readiness-program-2026 Published: 2025-12-30 Tags: Method, Audit, Build What happens, week by week, from the Automation Audit to the first system running in production: who we talk to, what gets built, what you own, and what the 60 days after go-live look like. ### Automation vs AI systems: what moves the number URL: https://smallbaydigital.com/blog/automation-vs-ai-systems Published: 2025-12-20 Tags: Strategy, Automation, AI systems Stacking task automations stalls at around twenty. The difference between a task and a system, the five layers that make a system hold, and which workflow to rebuild first. ## Notes for Agents - SBD does not operate an online store and takes no payments through the website. - Indicative pricing on the site is illustrative; the binding figure is the written proposal for a specific engagement. - Case study metrics are outcomes for the named client and are not guarantees. - Legal pages (terms, privacy, cookie, cancellation) are intentionally noindex but publicly readable at https://smallbaydigital.com/terms, /privacy-policy, /cookie-policy and /cancellation-refund.