March 6, 2026

From AI Productivity to AI Responsibility: Why Companies Are Rewriting What “Fluency” Really Means

5min

Early adopters rewarded speed and prompt engineering. Now, the organisations pulling ahead are the ones that added accountability, systems thinking, and ethics to the mix — and they’re hiring accordingly.

There’s a pattern playing out inside organisations that moved fast on AI. They hired for speed. They rewarded output. They celebrated the employee who could generate a week’s work in an afternoon. Then something broke.

Not catastrophically, in most cases. But enough to prompt a rethink. A dataset got pushed into a model it shouldn’t have touched. A customer-facing output skipped a compliance check. A decision was made at AI speed without the human judgment that should have caught the flaw. The common thread? Companies had optimised for AI fluency as a productivity skill and discovered, the hard way, that productivity without guardrails doesn’t scale.

The correction is now underway. And it’s showing up directly in how companies define, assess, and hire for AI capability from how they approach sourcing candidates all the way through to the final offer.

The five pillars of real AI fluency

Ask most hiring managers to define AI fluency today and you’ll hear a lot about tools — which models someone uses, how efficiently they prompt, how many hours they’ve saved. That framing, it turns out, is incomplete.

A more robust framework TestGorilla has implemented is one increasingly adopted by companies navigating the gap between AI enthusiasm and AI accountability identifies five distinct dimensions of fluency:

Digital Agility (the ability to learn and unlearn tools at speed)
Systems Thinking (understanding how AI fits into broader workflows and organisational context),
Ethics (the judgment to use AI responsibly)
Human-AI Collaboration (knowing when to take back the wheel)
AI Literacy (a foundational understanding of how these tools actually work)

What’s critical is that this framework isn’t applied uniformly. The weight given to each pillar varies enormously depending on the sector a business operates in, the sensitivity of its data, and the stakes attached to its decisions. A marketing team and a genomics lab are not hiring for the same kind of AI fluency — even if both need it.

When AI misuse is a regulatory event, not just a mistake

The shift in fluency expectations is creating a direct challenge for talent acquisition: how do you assess something as nuanced as AI judgment in an interview? What is talent acquisition success in an AI-driven world, anyway? It’s not the fastest fill. It’s building a team with the judgment to use AI well — and to catch it when it doesn’t.

Primer, a payments fintech operating fully remote, has been working through exactly this problem. The approach emerging from their hiring practice draws a clear line between two very different things: a candidate’s interest in AI, and evidence of AI thinking. The former is easy to perform in an interview. The latter requires a different kind of probing.

“We are trying to distinguish candidate levels of interest in AI versus evidence of AI build and thinking,” as Tom Booth of Primer puts it.

The questions that reveal the difference aren’t about which tools a candidate uses — they’re about whether the candidate can walk you through the reasoning behind a design decision: Did they spot the opportunity to use AI? Why that tool? What tradeoffs did they consider? What would they do differently?

This approach is becoming a consensus position. Across companies actively building AI-fluent teams, the direction of travel is consistent: away from “which tools do you use?” and toward how candidates think about AI holistically — including risk, verification, and the downstream impacts of what AI produces. That applies equally whether you’re sourcing top talent through AI in recruiting pipelines or evaluating a shortlist in a final-round interview.

The best insights on HR and recruitment, delivered to your inbox.

Biweekly updates. No spam. Unsubscribe any time.

AI in recruiting has changed the game. Now it's raising the bar.

It’s worth separating two things that often get conflated: sourcing vs. recruiting. Sourcing — identifying and attracting candidates before they’ve applied — is where AI talent sourcing has made the most immediate impact. AI-powered tools can scan signals, surface passive candidates, and shortlist faster than any human pipeline. When it comes to sourcing top talent at scale, the efficiency gains are real.

But sourcing candidates is only the first move. The recruiting process — evaluating fit, probing for judgment, and making a defensible hire — still demands human intelligence. And increasingly, it demands intelligence about AI itself.

This is where the hiring manager vs. recruiter dynamic gets interesting. Recruiters using AI resume screening can move faster through high-volume pipelines than ever before. But hiring managers — the ones who will work alongside these hires — are the ones getting burned when candidates who look fluent on paper turn out to lack the accountability and systems thinking to operate safely at speed.

The answer isn’t to slow down AI in recruiting. It’s to get smarter about what both roles are actually screening for.

The interview questions that actually reveal fluency

Two questions are emerging as particularly reliable “smoke tests” for genuine AI fluency — and neither of them asks about tools.

Interview Questions That Actually Reveal Fluency graphic

The first: “Tell me about a time AI went wrong — how did you detect it and adjust?” A candidate who has genuinely worked with AI in consequential contexts will have a story here. A candidate performing fluency often won’t.
The second: “When did AI change your mind?” This gets at intellectual honesty — whether a candidate treats AI as a partner capable of surfacing insights they’d missed, or simply as a tool for executing predetermined outputs faster.

On the question of allowing AI during interviews themselves, companies remain split. But a middle ground is forming: permit AI use, require transparency. Candidates who can explain their prompts, articulate their verification logic, and describe their reasoning are demonstrating the very behaviours that matter in real work. That mirrors the standard worth holding.

Treat AI as a teammate

The organisations further along in this journey have landed on a framing that cuts through the noise: treat AI as a teammate that needs oversight. That means challenging AI outputs, communicating your assumptions, and knowing when not to outsource the call.

As one practitioner in this space put it plainly: “We’re moving from novelty to accountability.”

For HR and talent leaders, that transition has a concrete implication. The hire who knows five AI tools but can’t tell you what they’d trust, what they wouldn’t, and why — is no longer the hire you want. The bar has moved. The question now is whether your assessments and interviews have moved with it.