Less noise. More useful answers.
Practical guides to choosing your first AI workflow, keeping human control and measuring what actually changes.
Your first AI workflow
A practical 30-day preparation plan
Nine pages to help you choose one workflow, set review rules and measure whether a pilot is worth exploring. With accounting and logistics examples, worksheets and a brief for your first conversation.
A preparation plan for your team, separate from the paid 45-day pilot.
One useful move this week.
Choose the situation closest to yours. Each path gives you a practical first week and a worksheet. No email required.
01Make the process clearThe work does not yet have clear rules or a named owner.
What you will have: One written process, a named reviewer and a clear exception boundary.
- Days 1 and 2
Describe one task
Write down its trigger, information sources and expected output. Include the manual process your team uses today.
- Days 3 to 5
Define review and exceptions
Name the reviewer. Collect routine examples and cases that need human judgement. Use synthetic or appropriately redacted data for preparation.
- Days 6 and 7
Test the instructions
Ask a colleague to perform the task using only the written steps. Each additional question reveals a rule that is still missing.
Move forward when the reviewer can explain the correct result and the exception route. If that is still unclear, improve the process first.
02Measure what is worth fixingThe workload is small or you do not yet know how much time it takes.
What you will have: A weekly baseline for workload, handling time and corrections.
- Day 1
Define a consistent measure
Choose one task type. Record case count, handling time, corrections and waiting. Keep active effort separate from elapsed response time.
- Days 2 to 6
Capture a normal week
Include everyone involved and the difficult exceptions. Extend the measurement if the sample is too small or the week is unusual.
- Day 7
Check the business case
Add up active handling time. Allow for future review, maintenance and running costs. A small workload may only need a template or a clearer rule.
Discuss a pilot when you have a repeatable problem and a credible baseline. Any saving still needs to be measured in practice.
03Prepare one focused pilotThe process repeats and someone can review the results.
What you will have: A brief for discussing one workflow, its controls and acceptance criteria.
- Days 1 and 2
Define the boundary
Choose one task and input source. Write down what is included and what AI must not do. Keep the existing manual process as the fallback.
- Days 3 to 5
Agree the measures
Define a useful result with the reviewer. Include review time, accuracy, exception handling and an acceptable running cost.
- Days 6 and 7
Check readiness
List the tools and permissions required. Bring the brief to a fit call; implementation starts only after agreeing the scope and terms.
Start a pilot after agreeing scope, access, review and acceptance criteria. These seven days are preparation, not a promised implementation timeline or an already purchased service.
Not sure which path fits? Find your starting point in five questions.
How to choose your first AI workflow
Five criteria for choosing a repeatable, measurable and controlled first workflow.
Read the guideWhat AI should do and where it should stop
Practical rules for approvals, exceptions and responsibility in everyday AI workflows.
Read the guideHow to measure whether a pilot is worthwhile
A simple framework for comparing time, quality and the review work AI still requires.
Read the guideFive questions. Your next step.
Use the assessment to turn the ideas into a practical starting point for your business.
Find your first AI workflow