The gap that opens after the programme ends

Most AI learning programmes close with a satisfaction score and a list of tools people have been shown, and neither tells you whether a single manager now runs their team's work any differently. Judging a programme that way is a bit like judging a cookery course by how much everyone enjoyed the tasting at the end: pleasant to know, and close to useless if what you wanted was better dinners at home. If you've ever had to defend an L&D budget with a satisfaction score, you'll know how thin that argument feels once someone asks what it bought. The programme may well have been well received. What matters is what people did differently once they were back in the job.

Finding that out takes a specific conversation between HR or L&D and each line manager, roughly four to eight weeks after a cohort finishes. It has a structure, and it asks for evidence, because impressions are exactly what the satisfaction survey already gave you.

I've written separately, in The AI adoption that already happened, about the broader reconnaissance of where AI use is already showing up across an organisation. This piece zooms in on one manager and one review, and on what that conversation should contain.

Why the manager is where you look

Managers decide whether experimentation has the time and permission to become part of the work. A programme can lift individual capability while team practice stays exactly where it was, and the usual reason is that the manager hasn't changed how they hand out work, review what comes back or signal what good looks like. Put simply, the programme teaches people, and the manager decides whether the team gets to use what it learned.

Whether the manager attended the programme changes the conversation, although both kinds of manager need one. If they were in the room, you can ask what they've changed in their own practice. If they weren't, the better questions are what they've noticed in their team's work and whether they've made any space for it.

Neither version should feel like an audit. Run it as a structured conversation that surfaces what changed, what didn't and what the manager needs next, and you'll get franker answers than any evaluation form is ever going to collect.

A hypothetical review: one manager, four weeks out

What follows is invented to show the mechanics; it doesn't describe a real programme or client.

Picture a senior manager in a financial services operations team. Her team of nine has completed a three-day AI practitioner programme, of which she attended the first day. Four weeks later an L&D business partner sits down with her for thirty minutes, and the conversation moves through four areas.

It opens with the work itself. The business partner asks her to name one task or process her team now handles differently, and she describes a weekly reconciliation report that used to take a team member around three hours and now, with an AI-assisted drafting step, takes considerably less. Two follow-ups matter here: where does the recovered time go, and has anyone checked the quality of the output since the change? She can answer the first but not the second, and that gap becomes an action item.

Next comes what she's seen but not yet acted on. Has anyone on the team tried something she noticed but hasn't discussed with them? She mentions an analyst who uses an AI tool to summarise regulatory updates, and admits she isn't sure the tool is on the approved list. That surfaces a governance question the programme never resolved for this team, and it tells the L&D function that the line between sanctioned and unsanctioned use is still blurry where the work actually happens.

The third area, how she has changed the way she manages, is the one most review conversations skip, which is a pity, because it tells you whether anything is likely to last, and if I only had five minutes with a manager it's the one I'd ask first. Asked whether she's adjusted her own routines (how she briefs work, how she reviews outputs, how she runs the weekly team meeting), she says she now asks people to flag when they've used AI on a piece of work, so she can see where it's happening. It's a small change in management practice and a meaningful one, and it quietly creates a data trail that nobody had to design.

Finally, the business partner asks what would make it easier for her to support the team's AI practice over the next quarter. She names two things: a clearer list of approved tools, and a short session on reviewing AI-assisted work for quality as well as speed. Both are requests the L&D function can take back and act on.

In under thirty minutes the review has produced one confirmed practice change, one quality-assurance gap, one governance question and two development needs, which is more usable evidence than a satisfaction score will ever give you.

The question set

Those four areas translate into six questions that an L&D business partner or HR manager can use as they stand.

  • Name one task or process your team now handles differently than before the programme.
  • Where did any recovered time or capacity go?
  • Has anyone reviewed the quality of AI-assisted outputs, and how?
  • Is there anything a team member has tried that you've noticed but not yet discussed with them?
  • Have you changed anything about how you brief, review or discuss work since the programme?
  • What would make it easier for you to support your team's AI practice over the next quarter?

Push for specifics. A manager who can't name a concrete change in answer to the first question has still told you something useful, because it suggests the learning stayed in the room and never made it back to anyone's desk.

What to do with the outputs

The review earns its thirty minutes only if the outputs go somewhere. Aggregate them across the managers in a cohort and the patterns appear quickly: which practice changes are spreading, which governance questions keep coming back and which development needs are shared across teams rather than individual.

Anyone who has sat through an evaluation meeting where the completion rate was the headline will recognise the problem: no dashboard of completion rates and satisfaction scores can show you any of that. The manager review gathers evidence that sits close to the work, which is precisely what makes it harder to collect and far more informative once you have it.

It also helps you tell two quite different problems apart. When several managers in the same cohort can't name a concrete practice change four weeks out, the programme design deserves a second look. When the governance questions cluster around the same ambiguity, more training won't fix it, because that's a policy gap and it belongs with whoever owns the policy.

Timing and frequency

Four to eight weeks after the programme is a sensible window for the first review: recent enough that people remember what they learned, and late enough for new habits to have either taken hold or quietly stalled. A lighter check-in at around three months then shows whether those early changes held or faded.

The review itself doesn't need to be long, and thirty minutes with a structured question set is plenty. What it does need is a business partner who's willing to push when an answer is vague, and to record what they hear in a form that can be compared across the cohort.

Most organisations stop measuring at the moment the programme ends, which is roughly the moment the interesting part begins. Thirty minutes with each manager, a few weeks later, is a small price for finding out what your people actually did next.

Written for this site.