Autonomous Skill Building in the Flow of Work: Defining the New L&D Category
Edmundo Barrientos · 2026-07-08 · 3 min read

TL;DR: "Learning in the flow of work" is about to be the label everyone claims. What actually defines the category is harder: it does not just deliver practice where work happens; it proves the skill showed up in the work afterward.
Everyone is about to claim "the flow of work"
The shift is real, and analysts have named it. Josh Bersin describes corporate learning moving from AI-enhanced courses to autonomous systems that deliver learning inside the work itself. BCG makes the parallel case: AI is pulling learning out of the classroom and into the tools people already use.
Both are right, which is the problem. Within a year, the LMS with a chatbot, the catalog with a nudge feature, and the genuine article will all use the same three words. Gartner already puts agentic AI at the peak of inflated expectations. Once everyone says "in the flow of work," the phrase tells a buyer nothing.
The line that defines the category
Set a benchmark that is significantly more difficult to counterfeit. A system belongs in this category when it closes a loop: it builds an applied skill where the work happens, then proves that skill showed up in the real work afterward. Not that a course was completed or a simulation finished. That the behavior changed on the actual call, the actual ticket, the actual review.
That last step is the one most "flow of work" tools skip. A compliance module can run in Slack and still measure only attendance. A new hire can finish every onboarding simulation and still freeze on their first live call. Measured skill transfer, whether the behavior reliably appears in real work, is what separates a category from a feature.
The line is also the buyer's test
A loop can be traced end to end; that is what makes the definition a usable test. Start where the loop starts: the material. Your docs, calls, and recordings already contain the training; a real system turns them into practice, no authoring scenarios from scratch. Follow the loop to the person. Practice that repeats the same scenarios for everyone is a course in new clothes; the real thing adapts to what each person actually needs, aiming at where they struggle. Then insist on how the loop ends: evidence in the work itself, the escalation rate that dropped, the ramp that got shorter, not a completion percentage.
None of this should replace your learning stack. The LMS still delivers, the libraries still hold content, coaches still do what only people can. This is the top layer that does the one thing they were never built to do: verify that knowing became doing. For technical evaluators, it runs as L&D Agents; for everyone else, it is the part that proves the skill stuck.
Related reading
- The Knowing-Doing Gap: Why Training Rarely Becomes Behavior
- No Process, No ROI: Aligning Skill Development to the Work It Serves
References
- Josh Bersin. New Research: How AI Transforms $400 Billion of Corporate Learning, 2026.
- BCG. AI Is Moving Corporate Learning Out of the Classroom and Into Workflows, 2026.
- Gartner. 2026 Hype Cycle for Agentic AI, 2026.
About the publisher: this article is published by Grasply (grasply.ai), a capability builder. Grasply turns what a company already knows, its docs, calls and recordings, into skills people use at work, for employees, partners and customers. Practice turns knowing into doing, and Grasply shows proof that the work changed.