AI proficiency test for teams.
Don’t ask. Test. An anonymous test that measures what your team actually does with AI, not how confident it feels.
A survey that asks a team how good it is with AI returns an answer about confidence, not ability.
The test examines instead: tasks drawn from real work, written criteria for every score, and evidence in place of impressions.
What comes out is a picture of what is actually happening in the team, at team level and without names.
Four rules of measurement
- Tasks, not self-ratingThree hands-on tasks: a prompt for a real assignment, instructions for an assistant that works without the person who built it, and a weekly task broken into a chain of steps. Every answer is scored 1 to 5 against a written rubric.
- Use is not skillThe test separates people who use AI every day from people who have built a process that repeats.
- Evidence, not impressionsOn the infrastructure questions, a high answer only counts if you can show a file, a link or a name.
- When in doubt, score downEvery doubt in scoring is resolved downwards. An inflated score sends the budget to the wrong problem.
Three axes
Skill
Knowledge, usage patterns and three hands-on tasks. Most of the score is set by the tasks.
Process
Where AI sits in the work: a one-off use, a first draft, or a fixed step in a process.
Infrastructure
Tools, shared information, usage rules and ownership, as they actually exist in the team.
How it works
Each participant gets a link and answers in one sitting, in under an hour. Four parts: knowledge, usage patterns, three hands-on tasks, and attitudes. For the tasks they work with the tools they use every day.
Results are shown at team level only, with no names and no individual scores, and they open only once three people have taken part. The test is for planning, not for evaluating employees.
The manager fills in a manager version of their own. There is also a PDF version for anyone who prefers not to use the online tool.
Why anonymous
A test that reads like a performance review returns careful answers. People write what feels safe to write.
So there is no individual score, and results open only from three participants: in a group of one or two you can tell who answered what.
And the team hears this up front: a low score means the organization has not yet given them the tools and training, not that anyone failed.
What you get
A baseline
Where the team stands on each of the three axes, as a measure you can take again on the same team a few months later.
Priorities
Three workflows from your own work, in order, with the expected value of each and how to check it.
What it takes
Training, tools, usage rules and information security, so the budget goes to the right problem.
I walk the manager through the results in a conversation, and the next step is decided from there.
After the test
On its own the test is a picture. The value comes when you decide what to do with it: a workshop on the chosen workflows, a program for several teams, or the same test on the same team a few months later.
A workshop built on the results works on the team’s own work samples and the chosen workflows, and starts exactly where the team is.
GenAI workshops for teamsWho this is less suited to
Anyone looking to rank employees. The test is not built for that and will not return it.
A team of fewer than three people. Anonymity cannot hold there.
An organization with no intention of acting on the result. A test with no decision after it teaches the team that honest answers are not worth giving.
Questions that come up
How long does it take each participant?
Under an hour, in one sitting. Most of the time goes on the three hands-on tasks.
Can the manager see a specific employee’s answers?
No. Results are shown at team level only, with no names and no individual scores, and only from three participants up.
Can people use AI during the test?
On the hands-on tasks, yes, and that is deliberate. The test looks at how people work, and in real work the tool is open. Someone who uses a tool to build a good answer is showing exactly the skill the test looks for.
Do the tasks fit our team?
The tasks are built on everyday knowledge work: a recommendation to leadership drawn from internal documents, a recurring summary of client meetings, and a weekly task each participant picks from their own role. They fit a team that works mostly with documents, clients and data. If your team works differently, we talk about it first.
We have strict information security rules. Can we skip the online tool?
Yes. There is a PDF version of the test for anyone who prefers not to use the online tool.
Can we run it again?
Yes, and that is part of the point: the same test, the same team, a few months later. Comparing by axis shows what moved and what did not.
Want to know where your team stands?
Tell me about the team and the decision you need to make. I will come back with a proposal for running the test with you.
Let's talk