The AI Productivity Gap Is Real — and Your Hiring Pipeline Is Making It Worse

The most dangerous assumption in AI strategy right now isn't that automation will eliminate jobs. It's that the people you're hiring already know how to work with AI. They don't — but they're getting very good at pretending they do.

Recent research found a spike in the number of job candidates using AI to cheat in the hiring process, with most getting through undetected, according to HR Executive (2025). That single finding should unsettle every executive who believes their talent pipeline is a reliable proxy for organizational AI capability. You are not hiring for AI fluency. You are hiring for AI-assisted performance theater.

Why are most AI productivity gains still theoretical?

The vast majority of executives expect to realize AI performance benefits later rather than now, according to HR Dive (2025). "Later" is doing a lot of work in that sentence. It functions less as a forecast and more as a permission structure — a way for leadership teams to defer accountability while continuing to invest. The board conversation you need to have isn't about which AI tools you've deployed; it's about why the productivity numbers haven't moved yet.

Here is why: AI transformation is not primarily a technology problem. It is a human capability problem. McKinsey (2025) argues that reaching the next productivity frontier depends on workers across the U.S. economy acquiring new AI habits and skills — what the report terms "AI fluency." AI fluency, defined plainly, is the capacity to direct, interrogate, and critically evaluate AI outputs rather than simply accepting or rejecting them; it is a cognitive and professional practice, not a software certification. Organizations that treat AI adoption as an IT rollout rather than a workforce development challenge will keep having board conversations about future gains.

What skills does AI actually leave behind?

Employers are increasingly in need of what researchers call "durable skills" — judgment, communication, ethical reasoning, and adaptability — precisely because AI is systematically eroding the value of the narrow technical tasks it can perform reliably, according to HR Dive (2025). The irony for hiring teams is severe: the skills that AI cannot replicate are exactly the skills that AI-assisted interview performance can most convincingly fake. A candidate who uses AI to generate polished answers to situational judgment questions has demonstrated the opposite of durable judgment — and your interviewers probably can't tell the difference.

This is a structural design flaw, not a character flaw in candidates. When organizations don't test for durable skills rigorously and in conditions that make AI assistance impossible or irrelevant, they get what they select for. The question isn't whether candidates are being dishonest. The question is whether your hiring process is built for the world it now operates in.

McKinsey (2025) frames AI transformations as a reinvention of how work gets done — one that requires change leadership rather than change management. That distinction matters. Change management assumes a stable destination; change leadership assumes the destination is itself being constructed. Organizations that are still running 2019-era competency frameworks through 2025 hiring pipelines are not in change leadership mode. They are managing toward a destination that no longer exists.

Meanwhile, workers are moving faster than employers. Employees are already using AI to identify development opportunities inside and outside their current organizations, according to HR Dive (2025). The workforce is self-organizing around AI capability in ways that have nothing to do with corporate learning and development roadmaps. If your people are using AI to plot their exit to more AI-literate environments, your internal mobility strategy has a problem your engagement survey will not catch in time.

There's a useful frame from Stanford University economist Erik Brynjolfsson, who argues that technology isn't the biggest barrier to AI-era progress — people, organizations, and institutions are, according to MIT Sloan Management Review (2025). That is not a comforting observation for senior HR leaders. It means the constraint is squarely in your domain.

The organizations that will close the AI productivity gap are not the ones with the best tools. They are the ones that stop measuring AI capability through proxies — certifications, interview performance, headcount of AI-adjacent roles — and start measuring it through actual work outputs, under real conditions, over time. The gap between AI investment and AI results is, at its core, a measurement problem dressed up as a technology problem. How long can you afford to keep treating it as the latter?