Pilot Planning Canvas
Your answers are saved in this browser. Creating your completed PDF report sends them to us, and we keep a copy. Use Start Over on the results page to clear the answers saved in this browser.
Plan a small AI pilot you can learn from
Choose one workflow and one question the pilot needs to answer. The canvas uses a two-week example to help you plan the work; your actual timing depends on access, scope and the people involved.
Start your answers ↓What you are testing
A good test starts small and clear. Pick one job from the drawing you made, the one where AI closes the biggest gap. Say who will use it and when. Then nobody can argue later about what you were testing.
How you will know it worked
Pick two or three numbers and write them down before you build anything. Write down where each one sits today, too. If you do not say what good looks like now, you will talk yourself into calling any result a win.
Week 1, build it
Week one takes you from an idea to something that runs. On days one and two, go back to the people who do the job and check you have the problem right. On days three to five, build one thing, for one job, and make it good. Do not build a finished system. Build the smallest thing someone would be glad to use.
Try it out, then week 2
On days six and seven, try it yourself, then show one or two people and fix what they trip on. On days eight to ten, give it to the whole team to use on real work. Be in the room. Watch them use it, do not just ask them afterwards. What people do and what people say are two different things.
Measure it, then decide
This is where the test earns its keep. Measure against the numbers you wrote down, not against how it feels. All three answers are good answers. The only bad one is never trying. If it works, go and show the people who hold the budget. If it half works, run another two weeks. If it does not work, you found out in two weeks instead of six months.
A worked example
Illustrative example, not a client result.
Imagine testing draft client reports using approved sample data. Record how long preparation and checking take today. Give a reviewer the same kinds of reports from the pilot, then compare the total time, missing information and corrections before deciding what to build next.
What you leave with
You leave with the test, success measures, work plan and next decision in one place. Keep the pilot's results separate from forecasts for a wider rollout.
Complete the form to create a PDF of your answers. No email is required. Creating the PDF sends your answers to us and keeps a copy, as explained above.
What you are testing
A good test starts small and clear. Pick one job from the drawing you made, the one where AI closes the biggest gap. Say who will use it and when. Then nobody can argue later about what you were testing.
Create a plan for a useful AI pilot
The Pilot Planning Canvas helps you choose one job, name the people involved and decide what to measure. Use it to turn an AI idea into a small test with a clear review date.
Gather examples of the work, agree who can access the records and name someone who can judge the result. Count the time spent checking and fixing as well as the time spent producing a draft. The two-week trial is a starting point for planning; choose a window that fits your work.
Read the AI proof-of-concept plan and sample test cases