A launch event is not adoption. We call the gap AI theatre. Announce, Enable, Embed is a quick diagnostic for which layer your effort is missing, and what leaders can do.
There is a familiar sequence. A leader announces that the organisation is "embracing AI". A tool is licensed. A launch event is held, a channel is created, and a champion posts a few enthusiastic examples. Six weeks later, the channel is quiet and the work is done exactly as before, except that a few people are using the tool privately and a few are avoiding it deliberately.
We call this AI theatre: visible activity that produces the appearance of adoption without changing how anyone works. It is rarely cynical. Most leaders who stage it sincerely believe the announcement is the first step. The mistake is treating the announcement as the intervention, when it is only the trailer.
People do not change how they work because they were told to. They change when three things are true at once: they know what is now allowed, they have time and a reason to try, and they see that people above them do it too. An announcement supplies the first at best, in vague terms. It supplies none of the second and third.
There is also a subtler cost. When a public commitment is not followed by visible change, staff learn that leadership statements on AI are decorative. That teaches them to discount the next one, which makes real change harder than if nothing had been said. The 2026 survey of 346 nonprofits on our website found 81% of AI use was individual, with no shared workflows. That is what a well-announced, poorly embedded programme looks like: enthusiasm at the top, private improvisation below.
Our diagnostic sorts adoption effort into three layers. Theatre is what happens when an organisation invests only in the first.
A useful test is to ask of any initiative which layer it serves. A launch event serves Announce. A policy serves Enable. Only a changed routine serves Embed, and Embed is the only layer that leaves evidence in the work itself.
The behaviours that move an organisation are small and slightly uncomfortable, which is why they are rare.
If the people who announced the programme cannot describe how they used it last week, the programme is a broadcast.
Boards and executives can look for evidence that does not depend on anyone's enthusiasm.
We would add a warning about metrics. Counting logins or prompts rewards activity, not change, and teams will optimise for whatever is counted. A better indicator is one measured before-and-after on a real piece of work: how long it took, how much rework it needed, and who checked it.
It is easy to read this as a criticism of leaders. It should be read as a criticism of the default sequence, in which the cheapest layer goes first and the hardest is left to the enthusiasts. Leaders are usually busy, are given little guidance on what visible practice looks like, and are often more cautious than their staff because they see the consequences of errors. The remedy is not exhortation. It is a short, specific list of behaviours, agreed and repeated.
Announcements grant permission, but only visible practice changes behaviour, so leaders should spend less on the launch and more on using AI on their own work, in view, with the checking shown. The three layers, Announce, Enable and Embed, give a fast way to see which part of the effort is missing.
Prepared by M.A.I. Consulting.
If this is the question on your desk, a thirty-minute call tells you whether the service fits, or that you do not need us yet.