Most firms hire accountants the same way they have for 30 years. You screen the resume, talk through experience, and send a take-home exercise: a messy set of books, a reconciliation, or a scenario to write up. The point is to see how someone solves a problem.
AI broke that. A candidate can hand your exercise to an AI tool and send back something clean, correct, and completely unrevealing. You have no way to tell which parts are theirs and which came from a prompt you never saw.
The instinct is to ban AI outright, with a signed statement that the work is theirs or a proctored, timed test. But then you miss one of the skills you most need to screen for: how someone works with AI. Your clients already have their own AI subscriptions. The people serving them need to be better with these tools than the clients are.
At Pilot, we care about how someone thinks through a problem they've never seen, how they handle a client who's unhappy, and what they do with an AI tool when it hands them an answer. So we built a live exercise into our interview process that tests all three.
It's a little cheesy, but we've found it works best as a role-play. Someone on our team plays an upset client with a real problem: books that don't tie, six months of expenses categorized their own way, or something equally tangled. The candidate gets it on a screen share with no prep time and has to work through it with the client in the room.
We tell candidates up front that they're welcome to use AI tools and it won't count against them. Then we watch how they use them. What do they ask it to do? Which answers do they take at face value, and which do they check? What do they do when it gives them something that looks right and isn't?
The client pressure is what makes the rest visible. Plenty of people can reason through a reconciliation in a quiet room. Far fewer can do it while someone is frustrated with them. You learn whether they get defensive, whether they listen before they explain, and whether they can say "I don't know yet, here's how I'll find out" without losing the room.
Not every accountant's job will center on the client relationship. Some of your best people work across a large portfolio. For those roles, keep the live format but drop the angry client, and test for what the work actually depends on: problem solving, creativity, and a head for process.
Start with a problem that doesn't have an obvious answer, like a balance sheet account that won't tie or a client whose business doesn't fit a standard chart of accounts. Watch how they break it down: where they look first, what they rule out, and how they use AI to move faster without letting it do the thinking for them.
Then push on it. Ask what they'd do if their first approach weren't available, or change the client's situation halfway through. You're looking for someone who can find a second path, not just repeat the first one.
Finally, ask them to make it repeatable. If this came up across fifty clients, how would they catch it every time? How would they set things up so it doesn't happen again? The best people for this work think past the fix to the process that prevents the next one.
The resume and the credentials used to carry most of the weight. This was a fair proxy for whether someone could do the accounting work. But it tells you less now. Two people with the same license will do this job very differently, and the difference shows up in the exercise, not on the page.
An interview is supposed to show you how someone will do the job, and nobody does this job without AI anymore. Banning it tests candidates under conditions they'll never actually work in, like asking them to close the books without Excel. You'd learn something, just not what you need to know. It also means you miss one of the skills you most need to screen for: how someone works with AI.
Accounting can be taught. Creativity, flexibility, problem-solving skills, and a nose for an excellent client experience are much harder to build into someone who does not already have them. These candidate qualities have always mattered, but they matter more than ever when everyone has a subscription to their AI model of choice.
Sponsored Content: This article is generously brought to you by one of our valued sponsors. Their support enables us to continue delivering expert insights and the latest industry trends to our dedicated community of accounting professionals.