For anyone whose pilot worked and then went quiet
Why AI adoption stalls, and what does work
The technology is rarely the problem. It stalls between the announcement and the daily work: people don't know what's allowed, have no time to practise, or can't see what it gains them. Four patterns we see in almost every organisation.
Plan an introductionThe pattern
The pilot worked. Then it went quiet.
The sequence looks the same in many organisations. A keynote or a pilot lands, people are enthusiastic, a small group gets going. Three months later that same group is still using it and nobody else is. That is rarely down to the model or the licence. It comes down to the four things below, all of which are about people and none about technology.
Four patterns
Where it stalls in practice
We see these four in nearly every organisation where we run discovery, regardless of sector or size.
Nobody knows exactly what is allowed
Without clear boundaries, everyone decides for themselves what is sensible. The cautious do nothing, afraid of mishandling client data. The enterprising do everything, through a personal account. Neither group produces adoption you can build on, and the second adds risk on top.
There is no time to practise
A half-day training makes people enthusiastic, but the following week the calendar is full again. New habits do not form in a session, they form in the weeks after it. If no moment is scheduled in those weeks where it is allowed to be awkward, everyone falls back on what already worked.
It is nobody’s job
AI adoption often lands with someone doing it on the side, usually from IT or an enthusiastic team lead. With no mandate, no hours and no place on an agenda where it gets discussed, it slides the moment the first real project comes along.
The gain stays invisible
People do not believe figures from a research report the way they believe an example from the colleague two desks away. As long as nobody can show what it concretely saves them in this work, with these systems, it stays a story about the future instead of a reason to do something differently today.
What does work
Three things that make the difference
None of these three is about picking a better tool. All three are about setting up the environment in which people will use that tool.
Start with the work, not the tool
Look first at where daily work leaks time into things nobody enjoys. That is where adoption happens by itself, because the gain is immediately felt. We do that with interviews and shadow days, because in a meeting people describe something different from what you watch them do.
Make the boundaries as boring as possible
A list of approved tools, a clear rule on what data may go into them, and who to call when in doubt. That is all. The shorter and more concrete that document, the greater the chance people follow it and dare to start.
Put a rhythm on it
A recurring moment where people show what they tried, including what did not work. That produces two things: colleagues see real examples from their own work, and the topic does not vanish from the agenda when things get busy.
Where this lands
How we work on this
Discovery on the shop floor
Interviews, shadow days and workflow analysis to find where time actually leaks. Six weeks from signal to a validated pilot plan. At Veris this freed up to 20% of the working week.
Training in your own tools
Hands-on work in the environment you already have, with your own examples. Four modules, 8 to 15 people per session. Over 700 people trained so far.
The boundaries on paper
Which tools are allowed, what data may go into them, and who decides. Without that it stays enthusiasm nobody dares turn into use.
When adoption isn't your problem
- Is a process stuck on unclear agreements or a system that stopped fitting years ago? That is a process problem, and AI will not solve it. Discovery finds that regularly, and when it does we say so.
- Is part of the organisation already using AI daily and independently? Then an adoption programme is redundant. The question then is how to make sure it happens safely and repeatably.
Over fifteen agents run across seven departments of our own company. What we propose to clients, we tried on ourselves first.
What that looks like, including what went wrong, is written up. Read the piece
Frequently asked questions
01 How long before AI adoption lands somewhere? +
The first visible time saving usually comes within a few weeks, provided you start with work people already find painful. Turning it into a rhythm that no longer fades takes more like six months. That gap is exactly why a one-off training day leaves so little behind.
02 Do we need a strategy first? +
Not necessarily. A strategy helps you choose what to focus on and in what order, but it is not a precondition for starting. In practice it often works the other way round: a few concrete results make the strategy conversation a lot sharper.
03 What about people who really don't want this? +
Resistance is usually a good question nobody answered: what does this mean for my job, and will I be judged on something outside my control. Take that question seriously and a lasting problem rarely remains. Wave it away and it comes back later anyway.
04 Is a keynote enough to get this going? +
A keynote is good at creating a shared starting point: everyone has seen the same thing and can talk about it afterwards. If nothing follows, that effect is gone within a few weeks. We recommend it as the opening of a programme, not as the programme itself.
Curious where your time is leaking?
Six weeks from signal to a validated plan to build on.