What Blind L&D Costs HR

The result is a widening gap between what training programs schedule and what employees actually need. Josh Bersin puts a number on it: 74% of organizations say they can't keep pace with the demand for skills, across a corporate learning market worth more than $400 billion globally. Companies are pouring resources into catalog management while the real learning need surfaces elsewhere, quietly, in a chat window. The problem isn't the budget, it's the model, and that's the core challenge of shifting L&D from training plan management to learning ecosystem architecture.

Why does this signal disappear instead of surfacing? The reasons are rational, not malicious. Fear of admitting a knowledge gap and being seen as replaceable. Reluctance to ask a question that feels too basic in front of a colleague or direct manager. The friction of a heavy internal tool, when an external AI responds in three seconds. And the straightforward reality that only 9% of employees have a company-issued AI tool to begin with. When the organization doesn't provide an alternative, the need doesn't disappear, it migrates out of view.

On the fear of disclosing AI use: "The more I use it, the more I encourage the organization to go AI-first, and the more I risk making myself redundant. I don't know anyone who saws off the branch they're sitting on. If I stay under the radar, I boost my productivity, but I don't share those numbers with the company."

(Webinar: "Your employees are already learning with AI, how do you take back control?")

Capturing the Signal Without Surveilling Anyone

There is, however, a viable middle path. Capturing this signal doesn't mean monitoring every query. It means working with anonymized, aggregated data, and targeting collective intent rather than individual behavior. What matters isn't "who asked what," but "which skill need comes up often enough to warrant a team-level response." Shadow AI then becomes a virtuous cycle: one employee's individual discovery becomes a micro-learning moment that benefits the whole team.

"We can go further: capturing intent through anonymized platforms is a key lever for identifying the questions employees are asking day to day, analyzing skill needs by team. This data can be collected, integrated into the training plan and reporting, and shared with employee representative bodies."

Clément Lhommeau, Co-founder of Blify (Webinar: "Your employees are already learning with AI, how do you take back control?")

To keep this signal inside the organization, the deeper fix is to stop sending employees to learn elsewhere in the first place. That means embedding learning in the flow of work, where the question arises, inside Slack, Teams, or the tools employees already use, rather than a catalog they have to log into separately. The model shifts from inventory to flow, from declared intent to actual behavior. This is precisely the transition described by the move from LMS to Learning Operating System. A platform like Blify deploys AI learning agents directly inside everyday work tools: the need is met in place, and the signal it generates stays within the organization instead of evaporating into a third-party AI provider.

Conclusion

Shadow AI isn't a leak to plug. It's a compass that most organizations choose to ignore, out of a control reflex, at the exact moment it's most reliably pointing the way forward. Every question sent to an external AI is a skill need made explicit: timestamped, contextualized, phrased in the employee's own language, and free. Captured and aggregated, this stream tells you in real time what no annual survey will ever surface: where your teams are stuck, on what, and when. Left unaddressed, it continues fueling a body of knowledge that builds outside the organization instead of circulating within it, widening the gap between what training programs schedule and what the work actually requires. The divide taking shape in 2026 doesn't separate companies that allow AI from those that ban it. It separates those that keep guessing what their workforce needs from those that have finally decided to listen.