Russ, CEO of KeyDelta, an operator-led strategic advisory firm helping CEOs and leadership teams scale faster and increase valuations.
“Proficient in ChatGPT” is now on roughly every junior resume crossing your desk. It is the new Microsoft Office. It tells you nothing about whether the candidate can actually direct AI to produce a business outcome. Most of them cannot.
I keep getting the same call from CEOs, CHROs and engineering vice presidents. They are sitting on stacks of “AI-fluent” resumes, and a vanishingly small share of those candidates can do the work. The talent shortage is not generic AI skills. It is the rare combination of judgment, rigor and domain depth that makes AI usable in production. The hiring playbook you have used for a decade is broken, and the gap is widening monthly.
The Talent Market, Just Repriced
Recent grad unemployment hit 5.7% in Q4 2025, with underemployment at 42.5%, according to the Federal Reserve Bank of New York. Stanford Digital Economy Lab researchers Brynjolfsson, Chandar and Chen found that employment for early-career workers in the most AI-exposed occupations dropped 13% from late 2022 through mid-2025, while older workers in the same roles held steady.
New-grad hiring at the 15 largest tech firms is down more than 50% since 2019. The bottom rung of the corporate ladder is being sawed off. The work that used to train juniors, summarize meetings and first-pass memos, and clean data and routine code is what AI does on its first try.
Meanwhile, AI-related skills now command a 56% wage premium, more than double last year’s 25%. You are paying that premium for a competence most candidates do not actually have.
Why ‘AI Fluency’ Signals Nothing
The resume claims have outrun reality. “Used ChatGPT to summarize meetings.” “Built a GPT for customer FAQs.” “Familiar with prompt engineering.” These describe consumption, not competence. Two candidates with identical claims will produce work that differs by an order of magnitude, and the resume gives you no way to tell them apart.
The competence you need is context engineering: instructing a literal-minded execution engine with the rigor a contracts attorney brings to indemnification language. Ambiguity in your instruction becomes hallucination in your output. The skill is translating fuzzy human intent into instructions a machine cannot misinterpret, then designing the checks that catch it when it fails anyway. It is learnable, it is rare, and it is almost never visible on a resume.
So what does count?
What To Hire For Instead
Here are five traits to look for instead:
1. They write the spec before they prompt. Ask how they would have AI draft a routine customer email. The weak answer starts with the prompt. The strong one starts with scope, in-bounds and out-of-bounds inputs and success criteria so explicit that two strangers would grade the output identically.
2. They show outcomes, not tools. Strong candidates do not show what AI they used. They show what changed because they used it: hours saved, errors caught, revenue moved. Three documented projects beat a stack of certificates.
3. They own a domain. AI plus a domain compounds. AI alone is a commodity. The candidate worth hiring can say, “I am the person at the intersection of AI and clinical trial operations,” or supply chain, underwriting or claims.
4. They verify everything. AI fails fluently, with wrong answers in flawless prose. The candidate worth hiring catches the hallucination, the citation that does not exist and the calculation that is off by an order of magnitude. In healthcare, finance and legal, this is the difference between a deployment and a lawsuit.
5. They reach out like operators. The candidates worth interviewing send a two-line thesis on a problem you actually have, plus a short 90-day plan. Treat that note as the first signal of the work product you will get.
How To Evaluate Candidates
Resume claims are often noise. Three tests work better:
• The Spec Test: Give the candidate a real, ambiguous problem from your business. Ask them to write the system instruction, the evaluation rubric and the failure modes they would test for. You will know in 20 minutes.
• The Verification Test: Hand them a plausibly wrong AI-generated output. The strong candidate spots what is broken. The weak candidate accepts it.
• The Outcome Interview: For every project they cite, ask, “How did you measure whether it was working?” and “What did it change in the business?” Specifics mean you are talking to an operator. Generalities mean you are talking to a user.
How To Build Fluency Internally
You will not hire your way out of this. Heroes don’t scale; systems do. Most of the capability has to be built in the team you already have. Three things move the needle:
1. Set the standard at the top. If leadership treats AI as a science project, middle managers will too. Every CEO I advise spends at least two hours a week working alongside AI on something that actually ships.
2. Run internal contracts, not training. Generic prompt courses do not move outcomes. Pick three high-value workflows, assemble a small cross-functional pod, and give them eight weeks to ship a measured improvement. Define it. Measure it. Own it. Close it. Scale it. Those people become your internal teachers for the next wave.
3. Make verification a competency, not a checklist. The most expensive AI failures right now are human, such as an employee accepting a polished, wrong output and shipping it. Reward people who catch errors. Promote them. Make it visible.
The Bottom Line
The companies that win will learn to recognize, evaluate and develop the real competence: directing AI with the precision of a contracts lawyer, the rigor of a senior engineer and the depth of a specialist. The companies still treating “AI fluency” as a hiring signal will spend the next decade wondering why their hires never produced the lift the wage premium implied.
Stop hiring AI fluency. Hire the operator. Build the rest. That is the job now.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
