In our last post, we asked 350 L&D and HR professionals a simple question: if AI could take something off your plate today, what would you hand over first? Course building topped that list by a wide margin. So we asked the follow-up question that actually matters once you’re building a roadmap: of the agent categories out there right now, which ones are people actually reaching for? 

And this is what they told us:

Agent categoryResonance
Content creation and refreshing68%
Analytics33%
Skill and learner nudges / notifications22%
Knowledge search15%
Compliance monitor13%

The answers track with what we heard on the ground at Inspire 2026, but they also reveal a few gaps between what’s hurting and what’s getting adopted. If you’ve ever sat in a planning meeting trying to figure out where to spend your first pilot budget, this is the data you wish you’d had going in.

The through-line: Content remains king

Anyone who’s spent a Sunday night rebuilding a course outline from scratch will not be surprised by this one. Content creation and refreshing was the top priority practitioners wanted off their plate in our last post, and it’s not close here either: 68% resonance, ahead of every other category combined. When the problem and the solution line up this cleanly, it’s worth paying attention. This is where the appetite for change is strongest, full stop.

It’s also where a lot of the early agent examples live. Think of an agent that takes a messy 20-minute conversation with a subject matter expert and turns it into a clean course outline you can actually build from. 

Or one that takes a policy doc or a URL and spits out a ready-to-publish quiz in minutes, no more staring at a blank assessment template at 4pm on a Friday. 

Or one that builds out a full 30-60-90 day onboarding plan for a new hire the moment you tell it the role and location. 

None of these replace your judgment as an instructional designer. They just get you past the blank page faster, which, if you’ve ever had six of these due at once, is most of the battle.

The surprise: Analytics is really two jobs wearing one label

Analytics comes in second at 33%, but the more we dug into it, the more it looked like two different requests wearing the same name tag. Some people want to talk to their data the way you’d talk to a colleague: ask a question, get an answer, then come back in three months with a completely different question when leadership’s priorities shift. 

Other people want something that watches quietly in the background and taps you on the shoulder, an agent that compares role profiles against current learning completions and flags the gaps before you even think to ask.

What if you could build an agent that does both?

The quiet mover: Nudges are pulling more weight than you’d think

Skill and learner nudges land at 22%, and honestly, it’s needed. We tend to think of agents as content machines or reporting machines. But an agent that quietly finds the learners who started something and never finished, then sends a personal nudge to bring them back, is solving a problem every one of us has lived with. 

Same with an agent that looks at someone’s existing skills, matches them to open roles, and suggests the next learning path without anyone having to build that map by hand. Neither of these is flashy. Both are the kind of ongoing, low-visibility work that never makes it to the top of your to-do list, which is exactly why it’s the kind of work worth handing off first.

The gap: Compliance monitor agents are lagging, and that’s not really about the tech

Compliance monitor agents sit at the bottom, at 13%. But this doesn’t read as a knock on what these agents can do. There are already agents that can review overdue compliance status and fire off a personalized reminder without anyone lifting a finger. 

The hesitation isn’t about capability. It’s about what compliance work actually asks of you: final sign-off on regulated content stays with a person, no exceptions, in most of the rooms I’ve been in. 

An agent can draft it. Someone still has to be the one who says yes, this is right. 

Lower adoption here reads less like reluctance and more like people drawing that same line with their pilot budgets that they’d draw with their own signature, which is exactly why the Docebo agents built for this space lean so heavily on permission controls and full audit trails. 

Trust gets built into the system first. Adoption follows after.

What this means for your roadmap

If you’re deciding where to start, the data hands you a pretty clean order of operations. Content creation and refreshing first, since that’s where both the pain and the appetite are strongest. 

Analytics and nudges next, once you’ve been honest with yourself about which analytics job you’re actually solving for. 

Knowledge search and compliance round things out, not because they matter less, but because they’re the categories people want to bring in once the bigger fires are out and the guardrails have been proven.

Explore AgentHub to see these agents in action, or get a demo to walk through which category fits your team’s priorities first.