Inspired by Courtney Vinopal's reporting in HR Brew, "To nab top AI talent, companies are pricing roles against a new set of competitors" (Aug 11, 2025). I'm adding a mid-market lens and what I'd actually do about it.
Here's the uncomfortable question I keep asking clients: is AI talent important enough to your business that you'll pay for it before the data catches up?
Because the data hasn't caught up. Your comp surveys were built for a world where "machine learning engineer" was a rare title. Now you've got Applied ML, ML Platform, GenAI product managers, and AI-heavy roles popping up in customer support. The surveys lag. The headline offers you read about are wild. And you're stuck trying to price a role nobody has good numbers on.
Why your usual benchmarks break
Titles mean nothing. Two "AI engineers" can have completely different jobs, and surveys can't keep up.
A few crazy offers skew everything. Treat them as context, not your pay policy.
There's new data out there. Pay transparency postings and your own offer and decline data belong in the mix.
Build an AI peer group
Think of it like an executive comp peer group, but just for these roles.
Start with where talent actually comes from. Where do your finalists come from? Who keeps poaching your best engineers?
Add a few big names as a ceiling. Even if you'll never match them, you need to know where the top is.
Balance it with companies you can actually beat. The places where you win offers today.
Write down the rules. Which roles count, which data you trust, how often you refresh it. And don't forget the non-engineering roles. AI is showing up fast in customer service and support.
Have the money talk before you post the job
Compression. What happens to the people already on your team when you bring in someone at a new number?
Trade-offs. If cash is tight, can scope, equity, or a chance to make a big impact fast close the gap?
Walk-away lines. Set your floor and ceiling up front, and stick to them.
What to do this week
Separate the roles that need deep AI expertise from the ones that just need to be AI-literate. The second group probably fits your normal pay bands.
Draft a peer list of 8 to 15 companies and write down why each one is on it.
Pull from multiple sources: surveys, public postings, your own offers and declines, and what recruiters are hearing.
Price three scenarios: top of market, competitive, and budget-friendly with non-cash perks.
Walk your leaders through all of it before recruiting starts. No surprises.
We're at the very beginning of how AI reshapes the workforce. The companies that figure out their pay philosophy now won't be scrambling when every role has an AI piece to it.
