Interviewing in 2026 is hard
AI has fundamentally changed the way we can and should interview.
10/31/2026
For decades we've interviewed software engineering and other technology roles based on how well they can produce the hands on work required of the job. Sure lots of companies have used (dumb) engineering challenges designed to stump people rather than represent the actual skills required for success in a given role, but by and large you would do a series of sessions to test how well something can actually write code, troubleshoot systems, draft product briefs, etc.
Well today AI is largely better at doing those things than people are. If you don't believe me you need to spend more time with the best models of 2026, seriously. AI can write code, create incredibly robust test suites, debug your production containers and servers, and manage the product management lifecycle with an incredible level of quality. Our jobs shift more towards engineering and product managers managing a team of technical resources in a box (or herdr) that complete work very quickly. If its not good enough they can rework it equally quickly. It's not 2023 where AI is rewriting entire codebases to change a text label. Teams are shrinking and moving more senior as a result.
So how do you hire and interview for people who will essentially be very skilled AI? This is the challenge we are all grappling with.
In 2023 and even 2024 it was pretty common to go to extraordinary lengths to prove people weren't using AI to answer coding challenges and coding questions. Well this doesn't make sense. People aren't writing code by hand in the future outside of very very niche problems (which you probably don't have) so why interview for a skill that isn't critical to the success of the role? Plenty of people who write great code are really struggling to get velocity improvements from AI. You could very well hire someone who will actually not be successful with this tactic.
But AI does make producing software or troubleshooting bugs almost effortless, so giving someone a problem to solve that is a few prompts away hardly makes sense either.
Instead what I've seen work really well is come up with a case. The case should be complex enough such that a human operator will have to make trade off or other branching point style decisions while solving it. Think about reliability vs cost trade offs, or safety trade offs. Detail the case well enough and give it to the candidate, either live or a soft time boxed take home, with the specific understanding that you want to see the solution a day before an interview will be scheduled to review it.
That interview session is the real magic. Use this to understand that they did make smart informed decisions and weighed the alternatives. Ask for expansions that weren't in the original case and ask how they'd think through solving them. Add new constraints and talk it through. You want to test that they have the critical thinking skills required to navigate the technical process while someone (in this case, an agent) produces the actual code.
Couple this with a session on general safe AI tool use, another on your usual architectural live discussion / white-boarding sessions, and my favorite session on how they self direct impact (good for all senior and leadership roles) and you should be in a good spot to decide if this person is actually going to be successful in the role.
The hiring landscape is evolving extremely quickly, and we as leaders are going to have to continue to evolve with it to build successful teams in this new future.