Picture the ideal interview process. The interview was warm, the answers were sharp, the take-home came back polished. They make the offer.
Three weeks in, the work looks nothing like what was submitted. Not worse, exactly. Just different, as if it came from someone else.
That gap is the problem the take-home test used to solve. It has stopped solving it now.
The test that stopped testing
One hiring assessment company, Fabric, reviewed more than 50,000 candidates and reported that AI assistance on assessments more than doubled in six months, from 15% in June 2025 to 35% by December. Fifty-nine percent of hiring managers in the same data said they now suspect candidates use AI during assessments. Treat the exact numbers with care, since Fabric sells assessment tools. The direction is hard to argue with.
Take-homes had other problems before AI arrived. A task billed as two hours often took four or five. The strongest candidates, the ones with other offers, simply declined. And the people who finished were often the ones with the most free evenings, not the most ability.
So the unpaid take-home now filters for time and tooling. Neither of those is what you are hiring for.
What is replacing it
The first way to evolve the take-home is to hold a meeting afterwards where the candidate presents their solution, without sending it over beforehand. In that meeting you ask questions and work together on challenges related to the take-home. It gives you plenty of information on how people work through problems. Using AI tools is normal in these meetings, and welcomed. You are not checking whether they avoided it. You are watching how they use it, and whether they can explain what it produced.
Another approach that works well is working through a task together with the manager who owns the craft. Again, you get plenty of data points on how people approach things, and you see their real thinking process. It also helps filter out the candidates who interview brilliantly but are less brilliant at the craft itself.
Of course, if trade secrets could come up in these sessions, you need to sort out the NDA side first. It is a minor complication next to the benefits.
Why this suits a small company
A large company can afford a long process, a panel, and a recruiter to run it. You usually have a founder, a calendar full of other things, and a gut feeling.
Working through a challenge for one to two hours gives that gut feeling something to look at. You see how someone reads a brief. Whether they ask a clarifying question or guess. What they do when the task is slightly messy, which at your size it always is. How they take feedback when you push back on their first pass.
None of that comes out of an interview answer. Interviews reward people who talk well about work. Working through a real challenge shows you the work.
It is also fairer than it sounds. Everyone gets the same brief and the same time. Nobody has to give up a weekend, because the work happens in the room, not at 11pm.
Making it worth the hour or two
It only helps if you decide what you are looking for before it starts. Pick three things that matter for this role, and score every finalist against the same three. Without that, you are back to a gut feeling with extra steps.
Use a challenge that mirrors the job but is not live company work you intend to ship.
Then ask why they made the choices they made. This also settles the AI question without a lecture. Someone who used AI well can explain every decision. Someone who did not cannot.
The shift worth making
Most founders already know the interview is a weak signal. They keep running it because it is familiar, and because the alternative felt like too much to ask of a candidate.
Sitting down to work through a challenge together is not too much. It takes an hour or two, it is the same for everyone, and it is the part of the process closest to the actual job.
The best evidence you can get about a hire is watching them do a small piece of the work, before it becomes your whole problem.