Your AI Training Strategy Is TikTok. That's Not a Compliment.

There's a version of the AI skills story that flatters HR: organizations are moving thoughtfully, assessing readiness, building structured curricula. The version that's actually happening is different. Employees are deploying AI tools in live work environments, and their primary source of instruction is social media.

That's not a prediction. HR Executive (2025-07-10) reports that workers are turning to platforms like YouTube, Reddit, and TikTok for AI training because formal employer programs either don't exist or aren't keeping pace with the tools employees are already using. The gap isn't theoretical — it's operational. And it's widening every week that procurement outpaces enablement.

The instinct to call this an L&D problem is wrong. It's a governance problem dressed in a training costume.

The ambiguity is the cost

When organizations don't define how AI should be used — what data can flow through which tools, what outputs require human review, what constitutes an acceptable use case — employees fill the void. They always do. And the informal norms they create in that vacuum are rarely the ones legal, compliance, or risk would have designed.

HR Executive (2025-07-10) identifies four distinct types of ambiguity that HR consistently fails to measure: role ambiguity, process ambiguity, priority ambiguity, and decision ambiguity. The thesis is uncomfortable: "we'll figure it out" is not a placeholder for a future policy. It is the policy, and it is costing organizations their best people. High performers have the most options and the lowest tolerance for unresolved ambiguity. They leave. The employees who stay are often the ones willing to operate in fog indefinitely — which is not the talent profile most organizations would choose to optimize for.

Now layer AI on top of that. When decision rights around AI use are unclear, who absorbs the uncertainty? Typically the employees who are already managing the most ambiguity: middle managers, individual contributors in fast-moving functions, team leads who were never given a framework but are expected to produce results anyway. They improvise. They Google. They watch a three-minute video on how to prompt a model and apply it to a client deliverable the next morning.

The McKinsey data point that should unsettle you

The McKinsey HR Monitor 2026 — a benchmark survey spanning Europe, the United States, and China — identifies the people function itself as being at a turning point, with capability gaps in HR accelerating precisely when organizations need HR to lead AI integration. That's the compounding problem: the function most responsible for building workforce AI capability is itself under-equipped to lead the work.

Blackstone offers a counter-example worth examining. The firm's Legal & Compliance AI transformation, documented by McKinsey, succeeded not because the technology was exceptional but because the organization redesigned decision flows first — clarifying ownership, codifying precedent, and establishing who was responsible for what before the models went live. The technology came second. The governance architecture came first. That sequencing is the lesson, and most organizations have inverted it.

What this means for how you're allocating attention

Senior HR leaders are currently being asked to own AI workforce strategy while simultaneously managing headcount pressures, return-to-office politics, and compliance complexity that grows by the quarter. The temptation is to treat AI training as a program to launch — a series of workshops, a vendor relationship, a completion rate to report to the board.

That framing will produce the appearance of progress without the substance. Completion rates tell you that employees sat through a module. They tell you nothing about whether the organization has resolved the underlying decision ambiguity that sends employees to Reddit at 9 p.m. looking for answers their employer never provided.

The question worth taking into your next leadership conversation is not "what is our AI training completion rate?" It is "what decisions have we actually made about how AI should operate in this organization, and who owns enforcing them?" If the honest answer is that the policy documents exist but the decision rights remain unclear — congratulations, you have documentation of your ambiguity, not a solution to it.

The most expensive sentence in HR isn't "we'll figure it out." It's "we have a training program for that."