Why most AI enablement programs stall at "documented"
A few months into most AI rollouts, the person running enablement has done everything the playbook asks for. Licenses are out. Training happened. There is a channel, a wiki page, maybe a monthly office hours. Good prompts exist, written down, in a place people can find them if they go looking.
And it still is not landing. You can feel it without being able to prove it. Usage is uneven. The same task gets done a dozen ways. Leadership asks whether the program is working and the honest answer is a shrug wrapped in a survey result.
If that is you, the useful thing to know is that you are not stuck because you did the work badly. You are stuck at the most common place to be stuck, and getting unstuck is a different kind of problem from the ones that came before it.
The five stages nobody tells you about
Every organization using AI is somewhere on a short ladder, and the rungs are not about how much AI you use.
The first two are about access. At stage one, people use AI and nobody knows who or for what. At stage two, there is enthusiasm and a license rollout, and everyone invents their own approach. Most enablement effort goes here, and it works: getting AI into people's hands is a solved problem.
Then comes stage three, documented, and this is where programs stop. Good practice has been captured. Someone worked out the best way to draft an incident statement or summarize a contract or reply to a regulator, and it got written down. You have done the hard part. It just is not spreading.
The two stages beyond it, circulating and evidenced, are the ones almost nobody reaches. And the jump from documented to circulating is the whole game.
Documented is not the same as circulating
Here is the quiet mistake buried in most programs. Writing something down feels like the finish line. It is not. It is the moment the constraint changes, and almost nobody notices it change.
Up to stage two, the question is whether people can use AI. Once good practice exists and is written down, the question becomes something else entirely: does the person who just worked out the best way to do a task ever reach the forty other people who do that same task?
Usually they do not. The forty sit in four different departments. They will never open the wiki page it was filed under, because it lives in a space that belongs to the team that wrote it, and they are not on that team. The good work is real, and it is stuck exactly where it was created.
The person who figured it out and the people who need it never meet. Not because anyone failed. Because nothing routed one to the other.
That is the circulation gap. It is not a training problem, because the people already know how to use AI. It is not a tooling problem, because you have plenty of tools. It is a routing problem, and routing problems are invisible until you name them.
Why the org chart is the wrong map
Most attempts to fix this reach for the org chart. Share it with the team. Post it in the department channel. Roll it up to the business unit.
But the people who do the same kind of work do not sit together. External Affairs, Government Relations, and site-level community relations may all draft the same kind of stakeholder correspondence while reporting to four different leaders in three different buildings. A standard that travels along reporting lines reaches the wrong rooms.
What you actually want is to route proven practice by the kind of work someone does, not by where they sit. The unit that matters is not the team. It is the recurring job: "when we draft an incident media statement, we do it this way." Everyone who does that job, wherever they are, is the audience. That is the difference between a standard that spreads and a document that gets bookmarked and forgotten.
This is also why generic "prompt library" tooling tends not to move the needle. A library scoped to a team is a better-organized version of the wiki page nobody outside the team opens. The scope is the problem, not the storage.
The part you cannot skip: seeing it
Say you fix the routing. Practice starts reaching the right people. How do you know?
This is where stage four turns into stage five, and where most programs that get moving still cannot prove they are working. Circulating is something you can feel. Evidenced is something you can show.
The measures that matter are not license counts or a satisfaction score. They are things like: is a given prompt being reused by people other than the person who wrote it? Did a practice actually escape the group that invented it, or is it just heavily used inside its own corner? Is the program self-sustaining, or does it only move when you push? Which kinds of work have a standard, and which are still ad hoc?
None of those come out of a survey. They come out of watching what people actually reach for, over time. And the reason they matter is not vanity. It is that the enablement lead who can put a map in front of leadership, showing where a standard way of working landed and where it has not, is in a completely different conversation than the one holding a deck of anecdotes.
What to do on Monday
You do not need a tool to start closing the circulation gap. You need to change what you are optimizing for.
- Stop counting access. Start watching reuse. The signal you want is not how many people have a license. It is whether one person's good work gets picked up by someone who did not write it.
- Pick one kind of work, not one team. Choose a recurring task that people across the organization do inconsistently, and make the audience everyone who does that task, wherever they report.
- Make contributing a byproduct, not a chore. If capturing what works takes effort, only the program team will ever do it. The good stuff surfaces when saving it is part of the normal flow of the work.
- Decide what "working" would look like before you need to report it. If you cannot name the evidence now, you will be assembling anecdotes under pressure later.
The circulation gap is the most common reason AI enablement stalls, and it is also the most fixable, because once you see it you stop mistaking it for a training or tooling problem and start solving the thing that is actually in the way: proven practice has to travel to the people who do that kind of work, and you have to be able to show that it did.
That is the whole jump from documented to circulating. Most programs never make it. The ones that do are not the ones with the most training or the most licenses. They are the ones that got the good work to move.
Where does your program stand?
Most AI enablement stalls in the same place. The maturity assessment tells you where yours is in twelve questions, with no sign-up and no email to see your result.