When to require AI skills in a hire

Hi Everyone,

In April 2025, Duolingo CEO Luis von Ahn told employees their AI use would count in performance reviews. Shopify's Tobi Lütke had sent a similar memo a few weeks earlier, calling AI use "a baseline expectation" and telling teams to prove AI couldn't do a job before asking for new hires. A year later, von Ahn dropped the performance review rule. Employees kept asking whether they were meant to use AI for AI's sake, and he has since said the rule wasn't right.

Today we're sharing a quick way to decide, role by role, whether AI skills belong in the job post or role expectations.

Start with these two questions

Whether a role should require AI capability comes down to two questions:

Does AI already sit inside the role's daily work? A content marketer spends the day drafting and reworking copy, and AI is now part of how that work gets done. A field sales rep spends the day in customer conversations, where AI helps with preparation but not with the conversation itself.

And who does AI help in this role? A large study of customer support teams found that AI assistance made new reps about a third more productive, while the most experienced reps barely improved. AI capability pays most where the work is high-volume and repeatable, and least where the value comes from judgment and relationships built over years.

Where each role ends up

Once you've checked whether AI sits in the role's daily work and who it helps, you're ready for one of four decisions — require AI skills in the hire, treat them as a plus, train them after hiring, or ignore them.

Require: AI sits inside the daily work, and the evidence says it pays. Marketing production, sales operations, junior analysis, and customer support belong here. Put the requirement in the post and test it in the interview.

Treat as a plus: Most other knowledge roles. AI helps, but a strong candidate who hasn't used it much can learn within weeks. Treat it as a plus, never a filter.

Train after hiring: Roles where your strongest candidates may not have touched AI yet, like experienced managers and senior client-facing hires. Hire for the core skill and teach the tools in onboarding. We covered how to spread AI skills across an existing team in this issue.

Ignore: Roles where the daily work happens away from a screen. An AI requirement here removes good candidates and adds nothing.

The fastest check is the job ads currently on your careers page — if one says "AI skills required" for a role where you'd train after hiring or ignore the skills, delete the line before the next candidate reads it.

Won't we fall behind?

A common worry with this approach is that companies requiring AI everywhere will pull ahead. Zapier requires AI fluency for every single hire, which sounds like one rule for the whole company. In practice, what a candidate has to show depends on the job — a marketer proves they can use AI in marketing work, an engineer in engineering work.

The company pushing hardest on AI hiring still makes these decisions role by role, so making them yourself doesn't put you behind. Duolingo tried the version without them, and dropped it within a year.

Go deeper

👉 Zapier: Raising the AI fluency bar for every Zapier hire — steal their role-by-role fluency levels before writing your own hiring bar.

👉 Fast Company: Duolingo's CEO admits where he got AI wrong — read this before announcing any AI rule to your team; von Ahn explains what he'd do differently.

👉 NBER: Generative AI at Work — the study behind the 14% and 34% numbers, worth skimming if you want to see which of your teams would benefit most.

👉 Tobi Lütke: The original Shopify memo — the full text everyone copied, so you can judge for yourself which parts fit your company.

Coming up on Monday

On Monday, we're covering why clients and agencies disagree about why their relationships end, and the review that puts both on the same standard.

Have a great weekend. See you on Monday.