The questions every leader should be able to answer about AI.
Sooner or later someone will ask the person in charge a plain question about AI. It might be a board member, a funder, an auditor or a new hire on their second day. What are we using? Who decided? Is it helping? Most leaders we talk to can answer some of these from memory and have to guess at the rest, which is normal for an organization a year or so into using these tools.
Why a self-test is worth an hour.
AI tends to arrive in pieces. One department starts using a chat assistant, someone connects a form to a spreadsheet, a vendor switches on a feature in software you already pay for. Each step is small and reasonable, and nobody draws the whole picture because nobody was asked to.
The result is a leader who is accountable for something they can only partly describe. That gap rarely causes trouble on an ordinary Tuesday. It shows up when something goes wrong, when the budget is reviewed, or when a client asks how their files are handled and the honest answer is that you would need to check.
A short list of questions closes that gap faster than a strategy offsite. You sit down, answer each one in a sentence or two, and mark the ones where you had to guess. Those guesses become your plan.
What you are using, and who is using it.
Start with an inventory, because every other answer depends on it. You should be able to name the AI tools in use across the organization, including the ones built into email, accounting or scheduling software, and roughly who uses each. A school district might find that the front office, two principals and the special education team are each using a different assistant, none of them approved and none of them known to the others.
You do not need a perfect list. You need one that is close enough that nothing important surprises you. If the honest answer is "I'm not sure," the first step is to ask people, without any hint that they are in trouble for telling you.
Who owns it, and what it has changed.
Ownership is the second question. For each tool or workflow that matters, one named person should be able to say how it is set up, what it is allowed to touch and when it was last checked. We have written elsewhere about choosing that person, so here the test is narrower: can you say the name without looking it up?
If you can't describe what AI has changed in your organization, you can't tell whether it is worth what you are paying for it.
Then ask what has actually changed. This is where many leaders stall, because the true answer is usually small and specific. Proposals go out a day sooner. The intake coordinator no longer retypes referral forms. A town clerk answers routine records requests from a drafted reply instead of a blank page. Those are real answers. "People seem more productive" is not one, and if that is all you have, the question has done its job.
What it costs, all in.
The subscription line is the easy part. The full cost includes usage fees on top of seats, the hours someone spends setting things up and fixing them, the training time, and any outside help. A manufacturer might be paying for three overlapping tools because three managers each bought one on a company card, and nobody has added them up.
You should be able to give a rough monthly figure and say what it buys. If the number surprises you once you total it, that is useful to learn now rather than at year end. If you can't produce a number at all, someone needs to own that too.
Where the lines are.
The last group of questions is about limits. What is your team not allowed to put into these tools? Which decisions must a person make, even when the software could draft them? What would you do if a tool produced something wrong that reached a client, a patient or a member of the public?
A clinic will have firmer answers here than a design studio, and that is how it should be. What matters is that the answers exist somewhere other than one person's head, and that your staff would give the same answers you would. If you asked five people and got five different versions, you have found the next thing to write down.
Leaders sometimes treat these questions as a compliance exercise for later. In our experience they are more useful early, while the number of tools is small enough to list on one page and the habits are still forming.
The scorecard.
Print this and fill it in by hand, ideally with one or two colleagues in the room. Score each question honestly: 2 if you could answer it confidently today, 1 if you could answer it with a few phone calls, 0 if you would be guessing.
- Which AI tools are in use across the organization, including features built into software we already have?
- Roughly who uses each one, and for what kind of work?
- For each tool or workflow that matters, who is the named owner?
- What has AI changed in our work that we can point to, in our own terms?
- What does it cost us each month, counting usage, setup time and outside help?
- What are people not allowed to put into these tools, and do they know it?
- Which decisions must stay with a person, and what do we do when the output is wrong?
Total the column. A score of 12 to 14 means you have a working picture and should revisit it each quarter. Anywhere from 7 to 11 means the picture has holes, and the zeros are your to-do list for the next month. Six or below is common and nothing to be embarrassed about. Take the lowest-scoring question, give it to one person with a two-week deadline, and run the scorecard again when they report back.
