Perspective
AI Agents Are Coming to L&D. Here's What They Should Actually Do.
Useful L&D agents should reduce friction around decisions, evidence, and workflow support instead of becoming another chatbot nobody needs.
The word agent is getting attached to almost every AI interface. A text box that answers questions can be useful, but calling it an agent does not create agency.
01A chatbot is not automatically an agent
The word agent is getting attached to almost every AI interface. A text box that answers questions can be useful, but calling it an agent does not create agency.
An agent should have a bounded job, access to the right context, clear rules, and a way to act or prepare action. It should move work forward without pretending to replace human accountability.
02Give it a real job
In L&D, useful agents might analyze intake requests, surface unanswered diagnostic questions, compare learning assets with current policy, prepare SME interview plans, or flag evidence gaps before launch.
They might help a manager prepare a coaching conversation or help an employee find the exact support for a current task. Those are recognizable jobs with a beginning, an output, and a human decision.
An agent earns its place when it helps someone complete meaningful work, not when it performs an AI demo.
03Keep the boundaries visible
An agent should show what it used, what it inferred, and where confidence is weak. If it changes a recommendation when one source changes, the human needs to see that.
Invisible automation is dangerous when the work affects people, compliance, access, or performance expectations.
04Design the handoff
The important interface is often the handoff between agent and person. What must the human review? What can be accepted quickly? What requires escalation? What evidence should be saved?
If the handoff is vague, the agent creates a new pile of AI output for someone to clean up.
05Measure friction removed
Do not measure success by conversations, prompts, or generated words. Measure time saved on a real process, better questions asked, faster access to support, fewer missed requirements, or stronger decisions.
The goal is not to sprinkle agents across L&D. It is to design a small number of trustworthy coworkers with jobs worth doing.
Take the thinking further
Pressure-test the decision inside The Lab
Perspectives name the problem. The Lab helps you challenge the default, examine the surrounding system, and decide what the work actually needs.
Explore The Lab