Hiring for AI? Look Beyond Certifications

Most companies are screening for AI the wrong way. Stop fixating on badges and tool choices. Understand the three kinds of AI work, then hire for people who can change how work happens-not just shave minutes off a task.

The real prize: new capabilities (not time saved)

Time‑savings are table stakes. The transformational value of AI is that teams can do work that was impossible yesterday. Example: I want finance analysts using AI‑assisted coding to spin up one‑off micro‑apps for analytics-solve a problem this afternoon, throw away the code tomorrow. That’s a different operating model: fewer handoffs, faster cycles, and outcomes that match the moment.

If you hire for badges, you’ll get incrementalism. If you hire for outcomes and workflow redesign, you’ll get compounding advantage.

The three kinds of AI work

Working with AI

This is the new “proficient in Excel.” Everyone should be comfortable using AI tools for research, drafting, summarizing, and analysis-without hand‑holding. It’s baseline digital literacy.

Building with AI

This is workflow engineering-embedding AI inside automations and processes so the way the function works changes. Think: turning a messy, multi‑step workflow into a reliable, human‑in‑the‑loop system that runs in minutes instead of days.

Building AI

Model research, tuning, and training. It’s specialized and science‑heavy. For most companies, the smarter move is to consume this capability via platforms rather than hire it directly.

Where to hire first: Most organizations should prioritize 1) and 2). The third is rarely necessary in‑house.

Why certifications are a weak signal

Certifications show intent, not impact. They tell me someone is motivated. They don’t tell me whether they can deliver value in real contexts.

When I review candidates, I look for:

• Proof of doing - personal automations, internal tools, side projects with a clear before/after.
• Speed with discernment - the ability to move quickly and recognize when a draft is not ready for prime time.
• Workflow thinking - evidence they redesigned how work happens, not just bolted AI onto yesterday’s process.

If a candidate lacked latitude at work, they can still prove it in life: automate home routines, run research with citations, build a tiny tool that saves an hour a week. The best people do this because they can’t help themselves.

Governance (the part that actually matters)

A lot of last year’s fears are now tractable with mainstream enterprise platforms and sane operating practices. Treat AI as one component in automation, with the same change controls and access management you already use.

The persistent, real risk is bias. Use the right mental model: you’re coaching book‑smart interns-immensely knowledgeable and sometimes naïve. Be explicit in instructions, require reputable citations where appropriate, and build transparent feedback loops so systems can be corrected and improved.

A non‑negotiable: label AI‑generated output. If people will rely on it, fact‑check it.

What the talent market actually looks like

Expect a rush of applicants for “AI” roles; only a fraction will meet the bar. The differentiator isn’t a badge-it’s shipped work, clear thinking, and evidence of changed workflows. Savvy candidates know they’re in demand; your job is to detect judgment and integrity, not just tool fluency.

If you’re a small business (quick notes)

• Skip the model wars. Choose an enterprise platform that aggregates models and handles privacy/controls. Then build.
• Hire a doer. Ask for a short project that automates a real task; pay for the work test.
• Measure outcomes, not hours. Did cycle time collapse? Did handoffs disappear? Can non‑experts run it?

Bottom line

Treat AI like electricity for your workflows-ambient and everywhere. Hire people who can re‑wire the building, not just flip a switch.