Most engineering leaders I talk to have an AI hiring plan. Almost none can tell me how they'd know a good AI-capable engineer when one walks in the door.
The numbers say I am not imagining it. Karat, the technical interviewing company, has published a preview of its 2026-2027 AI Workforce Transformation Report, based on a survey of engineering leaders in the U.S. and China. Two findings sit next to each other and should make you uncomfortable:
- 70% of engineering executives plan to expand their AI capabilities through strategic hiring.
- Fewer than 30% are investing in systems to identify AI-ready talent.
Read that again. Seven in ten are going to hire. Fewer than three in ten are building the means to tell who is any good. That's not a strategy. That is a wish with a requisition number.
A caveat on the source: this is a preview page, not the full report, and Karat sells interviewing services, so it has a horse in this race. I would not bet a budget on any single percentage. But the direction matches what I see on the ground, and the rest of the preview is worth your attention.
The productivity gain is real, and it is not free
The preview's central theme is that AI boosts productivity, but new workstreams offset some of the gain: reviewing AI output, writing better specifications, and testing. This is the part most AI business cases leave out. Code gets cheaper to produce. Judgment doesn't.
Google Cloud's Lee Boonstra, a software engineer in the Office of the CTO, put the shift plainly:
"The bottleneck didn't disappear. It moved from the code to the people reviewing it."
She wrote that in an April 2026 Google Cloud piece on who reviews AI-written code.
So the engineer you want is not the one who can prompt fastest. It is the one who can look at a confident, plausible, wrong pull request and say so, fast. That skill doesn't show up on a résumé, and a keyword search won't find it.
Top talent is being priced like it matters
Another number from the preview: 29% of leaders say their top engineers are now worth five times their total compensation. Whatever you think of the precise figure, the sentiment is the point: a meaningful share of leaders now see their best engineers as worth multiples of their pay.
If you believe your best engineers are worth multiples of their pay, then your process for finding the next one deserves more than a recycled coding quiz and a gut feeling.
What I would do on Monday
Start by defining "AI-ready" in writing before you open the req. Is it someone who ships with AI tools daily? Someone who builds the systems around models? Someone who can evaluate output? These are three different hires. Most job descriptions blend them into one unicorn and then wonder why the pipeline is empty.
Next, test the review skill, not just the writing skill. Give candidates AI-generated code with real defects and watch what they catch. You learn more in 30 minutes than from a stack of take-homes. (I made the broader case against broken evaluation signals a few days ago.)
Then put a measurable bar on it. Karat's preview suggests fewer than 30% of your peers are investing in a system, so a simple, consistent scorecard is already a competitive advantage. You don't need a platform. You need the same questions, the same rubric, and a debrief that happens the same day.
Last, use a recruiter who can run the technical conversation. Here's where I'm direct about our shop. The cheapest place to catch a bad AI-skills claim is the first screen. A recruiter who has placed engineers and understands what a good answer sounds like saves your team the hours of a failed loop.
The bottom line
The hiring plan is the easy part. Everyone has one. The scarce asset is the ability to evaluate, and the companies that build it first will hire the people the rest of the market is still trying to find.
If you are staffing an AI-heavy roadmap and your interview loop hasn't changed since 2023, that is the first thing to fix. Not the tooling. Not the job title. The loop.
Source: Karat, "2026-2027 AI Workforce Transformation Report" preview, karat.com/resource/ai-workforce-transformation-report.