Hiring in the AI Era

When anyone can produce plausible code with an agent, the coding screen stops predicting much. Hiring shifts toward what agents don't supply -- judgment, systems thinking, taste -- and toward integrators, not ML researchers. Teams redesign the screen (AI-allowed rounds, workflow observation) and read proxies like open-source work and tool maturity, since no one yet has years of agentic-coding experience.

The Pattern

"Yes, you can use AI in our interviews. In fact, we insist." -- Canva Engineering (source)

When anyone can produce plausible code with an agent, the closed-book coding screen stops predicting much -- it tests the thing that is now automated. Hiring shifts toward what agents do not supply: judgment, systems thinking, debugging under uncertainty, clear communication, and the taste to tell genuinely good output from output that merely looks right. The profile you are actually hiring for is an integrator and systems thinker, not an ML researcher -- so name the role precisely (agentic coder, AI product engineer, forward-deployed engineer), or you will draw machine-learning and traditional-AI applicants for a job that is really about wiring models, tools, and systems together.

Large orgs are redesigning the screen rather than policing AI out of it. In late 2025 Meta rolled out an "AI-enabled coding" interview where candidates work in a multi-file codebase with an AI assistant, still graded on problem solving, code quality, verification, and communication -- "more representative of the developer environment our future employees will work in," and harder to cheat (Hello Interview). Canva goes further, expecting candidates to use Copilot, Cursor, or Claude rather than "fighting this reality and trying to police AI usage" (Canva).

Why It Matters

Screening for syntax recall now selects for exactly the skill agents have commoditized, while missing the ones that decide whether someone can direct and check an agent. Counter-intuitively, making AI available does not make the interview easier -- weak fundamentals surface faster, the moment a constraint changes or an edge case is probed (ul Haq).

A pipeline, not one screen

Teams are converging on a few complementary rounds:

• An AI-allowed build round. Let the candidate use AI to produce the best solution they can, then judge how they directed and verified it -- "should use AI, but need to show you understand the code; explain the output; test before using; don't prompt your way out of it" (Hello Interview).
• A verbal round that probes understanding -- AI-proofed. Live reasoning where you guard against silent assistance: don't put the question somewhere a model can read it, and watch for answers that arrive suspiciously clean. A reliable screen is simply "tell me about a past project -- what was it, how did you know it was worth building, how did it go?" (Swizec Teller).
• On-site or pairing, to watch the workflow. See which tools they actually reach for and how they work with them -- the part no take-home reveals.

Signals when nobody has twenty years' experience

No one has a decade of agentic-coding experience; everyone is still learning. So read proxies for how someone works: their open-source projects and how openly they share what they learn, and -- most telling -- their tool maturity. Are they still at AI completion and chat assistance, or are they running autonomous agents and doing their own harness engineering? As one hiring manager put it, "how are they using AI today? Because if the answer is 'I don't,' that's a no" -- and the strongest signal is a candidate who pushes back on AI output rather than pasting it (Kore1).

The honest tension: these formats are early and unproven, and the deepest signals resist any single screen. As titles rise, code volume stops being the job -- when a project is off track "it's rarely for a lack of coding velocity; it's a lack of consensus, direction, customer buy-in, system architecture," and "none of the artifacts for those issues get committed to GitHub" (Gengelbach). The durable signals -- can this person think in systems, find the bug nobody else can, and judge what "correct" means -- are also what performance gets measured on later, and no interview format yet captures them reliably.

Last reviewed: 2026-06-26