How to Write an AI Job Description That Attracts Real Talent

Companies write bad AI job descriptions because they haven't decided what they're actually hiring for.

That decision, job design, is the real work. The job description is just the artifact that comes after it. Most companies skip the design and go straight to the artifact, then wonder why no one fits.

The two buckets

AI JDs cluster into two failure modes.

The wish list: "5+ years AI experience, LLMs, RAG, vector databases, fine-tuning, MLOps, Python." The kind of description assembled by someone who typed "AI skills" into ChatGPT and added "required" to everything. The company doesn't understand the space well enough to know which of these skills actually matter for the problem they're solving.

The other extreme: "Looking for an AI leader to help us leverage AI across our organization." This is the non-description. What does this person do on Monday? Who do they report to? What does success look like in six months?

Both fail for the same reason: the question wasn't answered before the posting went up.

Start with the problem

Before writing a single bullet: what decision or capability is this hire supposed to unlock?

The hire follows the problem. Someone to figure out where AI adds value? Fractional AI strategist, not a full-time hire. Engineers with no technical direction? VP of AI. Support team drowning with no idea how AI fits? Scope the project first, you're not ready to post anything.

The role follows the problem. The description follows the role, and neither step can be skipped.

Judgment or execution — pick one

AI roles split into two fundamentally different functions.

Judgment: deciding what to build, where to apply AI, how to sequence work, how to manage stakeholders, when to push back on a vendor. That's what a head of AI or AI consultant actually does.

Execution: building, fine-tuning, evaluating, and shipping models. That requires hands-on ML experience and direct exposure to production systems.

Most job descriptions want both. That person exists, is rare, and costs accordingly. More often, companies actually need judgment (better decisions) and they write for execution because execution skills are easier to describe.

Decide which one you need and write for that.

What to say instead

The descriptions that attract real talent have something in common: they're specific.

Specific about the problem: not "help us leverage AI" but "own our AI roadmap for sales and support, starting with the triage system we've already scoped."

Honest about where the company is: "We don't have a data team. You'll be working with what's in Salesforce and Google Sheets." Candidates who understand what they're walking into perform better and stay longer.

Clear about the first 90 days. What does this person ship, decide, or define before month four? If you can't answer that, you're not ready to hire.

The signal you're sending

Strong AI practitioners have options, and they're evaluating you as much as you're evaluating them.

A description full of buzzwords says: we don't understand this space yet. A description with specific problems, a real tech stack, and a concrete first 90 days says: we've thought about this seriously.

That gap, between the company that did the thinking and the one that didn't, is what separates good hires from expensive misses.

Write for the person who will do the work, not the résumé you hope to see.