Free tool
Find out what to automate first.
List the processes your teams run. Answer eight questions about each. Get every process placed on an impact and ease matrix, a ranked roadmap, and the right approach for each one, whatever tools you use.
Start mappingFree, no sign-up Updated Method and sources Markdown version
- Volume900 a month × 6 min
- Decided byMostly rules
- Arrives asPDFs and emails
- SystemsTwo or three
- ChangesRarely
- Exceptions10 to 30%
- API accessYes, all
- Cost of a mistakeMoney or compliance
Quick win AI extraction, then workflow, with review of exceptions
Summary and how to use it
In short
Automate first the processes that score high on both impact and ease. Impact is mostly the hours the work takes each month, plus the cost of a mistake and whether it touches customers or cash. Ease comes from fixed rules, structured input, few systems, stable steps, few exceptions and API access. Standardise any process that changes monthly or has more than about one exception in five before automating it.
Worked example
Supplier invoice entry: 900 invoices a month at 6 minutes each is 90 hours. Decisions mostly follow rules, invoices arrive as PDFs and emails, two or three systems are involved, the steps rarely change, 10 to 30% are exceptions, every system has an API and a mistake costs money. It scores 65 on impact and 67 on ease, a quick win. The suggested approach is AI extraction into a rules-based workflow, with a person reviewing the exceptions.
How to use it
- 1List the processes
Add your own, or start from the typical processes for finance, sales, support, operations, people or IT.
- 2Answer eight questions each
Volume and minutes, then rules, input, systems, stability, exceptions, API access, cost of a mistake and customer or cash contact.
- 3Read the matrix
Each process is placed by impact and ease. Bubble size shows hours a month.
- 4Follow the roadmap
Start with wave 1 quick wins, design the big bets next, batch the fill-ins and standardise anything unstable first.
- 5Model the return
Open any process in the ROI calculator with its volume and timing filled in.
Mapper
Your processes, scored.
Start from a department’s typical processes or a blank list. Edit anything; the matrix moves as you go.
Roadmap
The order to work in.
Ranked by quadrant, then by combined score. Each row links to the ROI calculator with its volume and timing filled in.
Approaches
Five ways to handle a process.
The mapper picks one per process from its answers. Logos are examples of each category, not recommendations.
- 01 Rules-based workflow
Fixed rules, structured data, systems with APIs. A workflow tool moves the data and handles the standard path end to end.
Chosen whenFixed or mostly rules, structured input
- 02 AI extraction, then workflow
Documents and emails read by a model into structured fields, checked against rules, and passed on. Low-confidence cases go to a person.
Chosen whenRules apply, but the input is a document or email
- 03 AI drafts, a person decides
The model gathers context and drafts the answer, quote or decision. A person approves before anything leaves. Time saved is lower but real.
Chosen whenMostly judgement
- 04 Integration or browser automation
One or more systems have no usable API. Budget for an integration layer or screen automation, and for more upkeep.
Chosen whenScreens only
- 05 Standardise first
Steps change monthly or too many cases are exceptions. Automating now locks in the mess. Write the process down, cut the variants, then score it again.
Chosen whenUnstable or over 30% exceptions
Process mapping, not tooling
Method
How each process is scored.
Both scores run from 0 to 100. A process at 50 or above on both is a quick win. The cut-off at 50 is a convention; read the matrix, not just the label.
Impact score
| Criterion | Weight | Scoring |
|---|---|---|
| Hours a month | 60% | Runs × minutes ÷ 60, on a log scale; 1,000 hours a month scores 100, 30 hours about 50 |
| Cost of a mistake | 25% | Little 10, rework 50, money or compliance 90 |
| Customers or cash | 15% | Yes 80, no 20 |
Ease score
| Criterion | Weight | Scoring |
|---|---|---|
| Rules or judgement | 25% | Fixed 100, mostly 55, judgement 15 |
| Exceptions | 20% | Under 10% 100, 10 to 30% 50, over 30% 10 |
| Input type | 15% | Structured 100, documents 55, free text 25 |
| Systems touched | 15% | One 100, two or three 65, four or more 30 |
| Stability | 15% | Rarely 100, a few times a year 60, monthly 20 |
| API access | 10% | All 100, some 60, screens only 20 |
- Quick winsImpact 50+, ease 50+
Do now. Proves the approach and frees time for the next wave.
- Big betsImpact 50+, ease under 50
Worth it, but needs design: AI for messy input, integrations, or a process fix.
- Fill-insImpact under 50, ease 50+
Cheap and safe. Batch several into one piece of work.
- ParkImpact and ease under 50
Leave for now. Check again if volume grows.
Sources
Where the numbers come from.
Each default and rule is traced to a published source, or labelled as practitioner guidance where no audited benchmark exists.
- McKinsey Global Institute A future that works: automation, employment and productivity, 2017Used forWeighting toward data collection and processing work
Technical automation potential by activity: predictable physical work 81%, processing data 69%, collecting data 64%. Fewer than 5% of occupations could be fully automated, so score tasks, not jobs.
- ICAEW Where could RPA work for your businessUsed forRules, volume, structured input and stability as criteria
Guidance for finance teams on choosing processes: rule-based, high volume, digital and structured inputs, and stable steps.
- Practitioner frameworks Process selection criteria cataloguesUsed forException bands and the standardise-first rule
Published selection guides agree on the criteria but not the cut-offs. Most treat exception rates above 15 to 20% as a warning, and favour processes that run the same way in 80 to 90% of cases.
- NIST AI Risk Management Framework 1.0, 2023Used forHuman review where mistakes are costly
A voluntary framework (Govern, Map, Measure, Manage) that ties the level of human oversight to the impact of an AI system’s decisions.
Terms
What the terms mean.
- Impact score
- A 0 to 100 score from hours a month (60%), the cost of a mistake (25%) and contact with customers or cash (15%).
- Ease score
- A 0 to 100 score from rules (25%), exceptions (20%), input type, systems touched and stability (15% each) and API access (10%).
- Quick win
- A process scoring 50 or more on both impact and ease.
- Big bet
- High impact but harder: messy input, many systems or judgement. Worth doing once the approach is proven.
- Exception rate
- The share of cases that cannot follow the standard path. Above about one in five, standardise before automating.
- Structured input
- Data that arrives in fixed fields, such as a form or a system record, rather than a document, email or call.
- Human review
- A person checks or approves the output before it takes effect. Used where mistakes are costly or judgement is involved.
- Standardise first
- Write the process down and cut its variants before automating, so the automation does not lock in the mess.
FAQ
Questions about choosing what to automate.
How do you decide which processes to automate first?
Score each process on impact (hours a month, the cost of a mistake, whether it touches customers or cash) and on ease (fixed rules, structured input, few systems, stable steps, few exceptions, API access). Start with processes high on both. Large, hard processes come next, once the first wins have proved the approach.
What makes a process a good candidate for automation?
High volume, decided by rules, structured or at least consistent input, steps that rarely change, fewer than about one case in five needing special handling, and systems that can be reached by API. A process that fails on stability or exceptions should be standardised before it is automated.
Has AI changed which processes can be automated?
Yes, mainly on input. Documents, emails and free text that once ruled a process out can now be read by a model into structured data, with low-confidence cases sent to a person. Work that needs judgement is better suited to AI drafting with a person approving than to full automation.
Should we automate a process that changes often?
Not yet. Automating an unstable process locks in its variations and raises upkeep. Document the current steps, cut the variants and exceptions, run it the new way for a few weeks, then score it again.
When does an automated process need human review?
When a mistake costs money, breaks a compliance rule or reaches a customer, and whenever a model makes a judgement. Frameworks such as NIST’s AI Risk Management Framework tie the level of oversight to the impact of the decision.
Is my data sent anywhere?
No. The mapper runs in your browser. Your list is saved in this browser only so it is there when you come back, and you can clear it at any time.
More free tools
More free tools
- Automation and AI ROI calculatorFree, vendor-neutral automation and AI ROI calculator. Any pricing model, loaded labour cost, risk adjustment, NPV, IRR, discounted payback, scenarios and sensitivity, with the method and sources shown.Open the tool
- Build vs buy decision toolFree, vendor-neutral build vs buy tool. Compare buying software, configuring a low-code platform and building custom on a weighted scorecard and a discounted three-year total cost of ownership, with the method and sources shown.Open the tool
- AI readiness assessmentFree AI readiness self-assessment across strategy, infrastructure, data, governance, talent and culture, the six pillars of the Cisco AI Readiness Index. 24 questions, a radar chart, a readiness stage and the next steps for your weakest pillars.Open the tool
- AI agent risk and controls plannerFree planner for AI agents. Describe what the agent reads and can do, and get its risk profile, the OWASP LLM risks that apply, a recommended starting autonomy level and the controls to put in place before launch.Open the tool
- All free toolsEvery calculator and planner, with its method and sources.See all