ContactBook a call

Explainer

Agentic AI vs AI agents vs generative AI

Clear definitions of generative AI, AI agents and agentic AI, a side-by-side comparison, examples, and when each one fits the work in a business.

Cover for “Agentic AI vs AI agents vs generative AI”, with the line: Three terms for how much an AI system does on its own.
Three terms for how much an AI system does on its own. Image: SBD.

Generative AI is a model that creates content, such as text, images or code, when someone asks for it. An AI agent is a piece of software that uses a generative model to decide what to do next and then does it, by calling tools such as your CRM, inbox or calendar, until a task is finished. Agentic AI is the broader term for systems built this way, from a fixed workflow with one AI step to a set of agents working towards a goal with little supervision. The three build on each other, and the real question for a business is how much a system may do without a person checking it.

This explainer is for owners and operations leads who hear the three terms used interchangeably by vendors and want to know which one they are being sold. It gives a definition of each, a comparison table, examples from ordinary business work, the most common confusions, and a way to decide which one fits a given job.

The short answer

Use generative AI when a person needs a draft and will check it. Use a fixed workflow with an AI step when the process is the same every time. Use an AI agent when each case needs a different sequence of lookups and actions, and keep a person approving anything that moves money or contacts a customer. "Agentic AI" covers the last two, so ask a vendor exactly what their system may do without asking you.

Generative AI

Generative AI means models that produce new content from an input. IBM’s definition centres on creating text, images, audio, video or code in response to a prompt. ChatGPT answering a question, Claude drafting an email and an image model producing a product photo are all generative AI.

On its own, a generative model does not do anything in your systems. It reads what you give it and returns content, and a person decides what happens next. Most of the value businesses got from AI in 2023 and 2024 came from this: faster first drafts, summaries of long documents and answers drawn from a knowledge base.

The model at the centre is usually a large language model. It is the same engine inside an agent, which is one reason the terms blur.

AI agents

An AI agent wraps a generative model in a loop. It is given a goal, a set of tools it may use and rules about what needs approval. On each turn the model decides which tool to call, reads the result and decides again, until the task is done or it hands over to a person.

A support agent, for example, might read an incoming ticket, look up the customer in the CRM, check the order in the shop system, issue a replacement within a set limit and draft a reply for a person to approve. No one wrote the exact sequence in advance. The model chose the steps for that ticket.

Anthropic’s engineering team draws the line in a way we find useful. In their description, workflows run language models and tools along “predefined code paths”, while agents “dynamically direct their own processes and tool usage”. Both are agentic systems, but they behave very differently when something unexpected happens.

Agentic AI

Agentic AI is the umbrella term. IBM describes it as goal-oriented: the system works out the steps, takes actions and uses tools, with varying levels of independence. In practice, vendors use it for anything from a single agent to a group of agents passing work to each other, and sometimes for a fixed workflow with one AI step in it.

That breadth is the source of most confusion. “Agentic” describes a property, how much the system decides and does by itself, and it comes in degrees. A useful way to picture it is as a dial:

  1. Generative AI. The model drafts and a person does everything else.
  2. Workflow with an AI step. Software follows a fixed path, and the model handles one step, such as reading an invoice. Our glossary calls this an agentic workflow.
  3. A single agent. The model chooses its own steps from a set of tools, within limits.
  4. Several agents. Agents hand work to each other, for example one that researches and one that writes, with a coordinating agent above them.

Each step up the dial adds flexibility and also adds cost, delay and ways to go wrong. Anthropic’s advice is to start with the simplest design that works and add complexity only when it clearly improves the result, and that matches what we see in client work.

Side by side

Generative AIAI agentAgentic AI
What it isA model that creates content from a promptSoftware that uses a model to choose and take actions with toolsThe wider category of AI systems that act towards a goal
What it does in your systemsNothing on its ownReads and writes through the tools it is givenRanges from one AI step in a workflow to several agents working together
Who decides the next stepA personThe model, within rules and limitsDepends on the design
Typical exampleDrafting a proposal from notesTriaging a support ticket and issuing a replacement within a limitA set of agents that research, draft and send a weekly report
Main riskWrong or invented contentWrong action taken in a live systemErrors that compound across steps and agents
Main controlA person reviews the outputScoped access, approval for money and messages, logsAll of the agent controls, plus tests across the whole chain
Running costPer prompt, usually smallSeveral model calls per task, plus tool and hosting costsGrows with the number of agents and steps

Which one fits the job

Use generative AI when a person will read and act on the result anyway. Drafting job adverts, summarising meeting notes and answering staff questions from a policy handbook all fit. It is the cheapest option to try and the easiest to stop.

Use a fixed workflow with an AI step when the process is the same each time but one step needs judgement on unstructured input. Reading supplier invoices into the accounts system, sorting inbound emails into queues or extracting fields from forms are good examples. The workflow tool handles the sequence and the model reads the document. Our post on automation compared with AI systems covers this choice in more depth. For the step-by-step version, see how to automate data entry.

Use an AI agent when each case needs a different sequence of lookups and actions, and a person would otherwise spend their day clicking between systems. Customer support triage, lead qualification and chasing missing documents fit well. Give the agent a narrow job, the fewest tools that will do it and approval steps for anything risky. Our post on moving from zaps to agents shows when the switch is worth it. For sales outreach in particular, our explainer on AI SDR vs human SDR sets out where an agent helps and where a person should stay in charge.

Use several agents only when one agent’s job has grown too large to test. Most businesses with fewer than a few hundred people do not need this for some time.

A question that settles many cases: could you write the steps down as a flowchart that covers nearly every case? If yes, build a workflow. If the flowchart would need a branch for every customer, an agent may earn its extra cost.

Common confusions

“Our chatbot is an agent.” A chatbot that only answers questions from a knowledge base is generative AI with retrieval, even if it is labelled an agent. It becomes an agent when it can take actions, such as booking an appointment or updating a record.

“Agentic AI means no humans.” The systems that work in production keep people approving the decisions that carry risk. How much an agent may do alone is a design choice you make, and our post on how much an AI agent should be allowed to do sets out the levels.

“Agents replace automation.” Fixed workflows are cheaper, faster and easier to test. Agents often sit on top of them and call them as tools.

“Generative AI is old news.” Every agent runs on a generative model. Better models make agents more reliable, so the two move together.

What it costs

Generative AI costs a subscription or a small amount per prompt. Agents cost more because each task makes several model calls and needs hosting, logging and testing, and our guide on how much an AI agent costs works through the sums. At SBD, workflow builds start from $1,000 and care plans from $50 a month, and every project starts with a $500 audit that is credited in full to the build. The audit tells you which parts of your work suit a workflow and which suit an AI agent. Our Automation team at $999 a month suits businesses that want a steady flow of builds.

If you want to see how agents connect to your existing software, our explainer on MCP for business owners covers the standard most of them now use.

Questions we get asked

What is the difference between agentic AI and generative AI?

Generative AI creates content when asked. Agentic AI uses generative models to pursue a goal by taking actions in other systems, with some degree of independence.

What is the difference between AI agents and agentic AI?

An AI agent is one piece of software that chooses and takes actions with tools. Agentic AI is the broader category, covering single agents, several agents working together and workflows with AI steps.

Is ChatGPT an AI agent?

Used as a chat assistant, ChatGPT is generative AI. When it is connected to tools and allowed to take actions in other systems, it acts as an agent.

Do I need agentic AI for my business?

Many businesses get most of the value from generative AI and fixed workflows first. Agents earn their place where each case needs different lookups and actions across several systems.

Are AI agents safe to use with customer data?

They can be, with scoped access, approval for anything that moves money or contacts a customer, logs of every action and testing before go-live.

What is an example of agentic AI in a small business?

A support agent that reads a ticket, checks the customer and order, offers a replacement within a set limit and drafts a reply for a person to approve.

Is agentic AI more expensive than generative AI?

Usually, because each task makes several model calls and the system needs hosting, monitoring and testing. A fixed workflow with one AI step sits in between.

Will agentic AI replace staff?

In our experience it takes over repetitive lookups and data entry, and people spend more time on exceptions, customers and decisions. Results depend on the process and how well it is defined before the build.

Sources and further reading

  • AI agents
  • Agentic AI
  • Generative AI
  • Automation

Keep reading

Cover for “When to replace a pile of Zaps with an agent”, with the line: From single-track automations to agents that hold state and route exceptions.

Systems

How to move a Zap-based process to an AI agent

A step-by-step guide to replacing a chain of Zaps with an agent: inventory, data model, approvals, side-by-side running, switch-off and pitfalls.

Cover for “AI voice agent pricing per minute in 2026”, with the line: Five parts make up the price of every minute.

Explainer

AI voice agent pricing per minute in 2026

What an AI phone agent costs to run per minute: platform fee, phone line, transcription, model and voice, with Retell, Vapi, Bland and ElevenLabs compared.

Cover for “MCP for business owners, in plain language”, with the line: One standard plug between an AI assistant and your software.

Explainer

MCP for business owners, in plain language

What the Model Context Protocol is, what it changes for a business, official MCP servers from Stripe and HubSpot, and the risks and controls.

Find out what is worth automating

About 95% of AI pilots never reach the profit line. We start with the work your team still does by hand and measure each build by the hours it saves.

See how we work

Cookie preferences

Choose what this site may use on your device. You can change this at any time from “Cookie settings” in the footer.

Strictly necessary

Keep the site working and remember the choices you make here. These cannot be switched off, and none of them track you.

Always on
What this includes (4)
NameProvider PurposeDuration
sbd-consentThis site Remembers your cookie choices12 months
Vercel Web AnalyticsVercel Inc. Counts page views in aggregate. Sets no cookie and stores nothing on your deviceNone
ticker-dismissed, sbd-note-offThis site Remembers that you closed a notice or language tipUntil the tab closes
sbd-voice-resume, sbd-voice-prefillThis site Carries a voice conversation across pages, only if you start oneUntil the tab closes

Analytics

Tell us which pages are read and where visitors leave, so we can improve them. IP addresses are anonymised and nothing is used for advertising.

What this includes (2)
NameProvider PurposeDuration
_gaGoogle Analytics (Google LLC) Distinguishes visitors with a random ID13 months
_ga_FPXDMQPRV1Google Analytics (Google LLC) Keeps the session state for this site13 months

Functional

Embedded services that set their own cookies. Without them the site still works; you get a link to open the service instead.

What this includes (1)
NameProvider PurposeDuration
Booking calendarCal.com Inc. Shows live availability and books the call on the booking pageSet by Cal.com

Advertising

Not used. This site has no advertising, retargeting or social media tracking, and we never sell data.

Not used

Questions about your data: hello@smallbaydigital.com