AI May Free L&D Capacity, Unless More Content Becomes the Outcome

The real decision is not whether agentic AI belongs in L&D. It is whether automation will free time for diagnosis and work design, or simply create more content and activity. Early use is more likely to compress the function's operational layer than to change capability strategy. That creates value only if the function uses the capacity to redesign work, decide what should be automated, and measure whether the change improves performance.

Docebo reports survey data from 350 L&D and HR professionals on the agentic AI categories they say they are adopting, including content creation and compliance monitoring. That is evidence that people are paying attention to creation, administration, workflow support and task automation. It is not evidence that those uses have improved productivity, performance or workforce capability.

The useful question is what work should follow the automation. A content workflow may reduce production effort, while leaving the harder work untouched: identifying the performance requirement, changing tasks and decision rights, putting support into the workflow, and deciding what evidence would show improvement. If those steps do not change, faster content creation can increase supply without improving readiness. The wrong diagnosis is to treat automation as a capability program simply because the function can produce more material.

I would ask the CHRO, CPO or senior CLO to trace one automated workflow from request to business result. The review should identify the task removed, the owner of the resulting decision, the manager or performer whose work should change, and the measure that distinguishes time saved from capability gained. This sequence separates a capacity release from a new form of activity reporting.

The source is a vendor survey and usage snapshot, so it may reflect experimentation rather than mature operating model adoption. The next useful evidence is independent reporting on cycle time, quality, manager load and performance after specific workflows change. Until that is visible, the sound decision is to fund targeted automation only alongside a plan for diagnosis, work design and evidence. AI can narrow training to the cases where training is actually the right instrument, but adoption alone does not prove that shift has occurred.