Adoption LessonsAdoption
When someone runs out of Claude, ask what they built
Jason Lawrence
Every leader with Claude seats gets the message within a month of the rollout: “I’ve run out. Can I have more?” How you answer it decides whether people keep using the tool or start hiding how much they use it.
The number: five in nine weeks
Between the start of July and early September, at least five of our people hit their Claude usage limit. Four said so in the channel where we answer AI questions, and one asked for more through Claude itself.
In July I set the rule: extra credit generally won’t be approved. That was not a “slow down” message. It was a nudge to use what we had well, and to pick a lighter model for lighter work. The post ended with the part that mattered. If you do run out, we want to hear about it, as a win. Tell us what you built.
The first person to run dry was named in that post. Twelve minutes later they had posted a walk-through of the Claude Project they had set up for a client engagement: the sources it was connected to, how it behaved, and what everyone else needed to do to use it. That is a better adoption report than anything a licence screen shows.
What went right
A usage limit is the first adoption data you get that isn’t a licence report. Empty seats tell you who isn’t using Claude. Full ones tell you who is, and the question tells you what for.
The person who runs our AI tooling wiki had hit the limit three times in three months, worked out why each time, and not hit it since. They offered to look at anyone’s chats, so nobody had to guess or ration. The causes were ordinary: an effort setting accidentally left higher than intended, one five-hour session, long chats reused for unrelated work. Each is a habit, and habits can be taught. We wrote a page on usage and limits, a page on picking the model for the job, and pointed everyone at the usage screen in Claude’s settings.
What went wrong
The limit message itself. In July it read “your usage allocation has been disabled by your admin”. Three of the four posts quoted it, and each person read it as a sanction: one asked whether to start a new session, another asked for their access back. The first time it appeared, even Claude, answering in the channel, read it as an admin switching the seat off. Both times we checked, it was an ordinary usage limit.
A limit that reads like a punishment gets treated like one. People who think they have been cut off stop posting, and the data goes with them.
And the rule bent once. In September someone asked how to get back to work faster, and I added credit within two minutes. That was the right call. It also means the rule is a default, not a wall, and it is worth saying so.
What we would do now
Decide the answer to “I’ve run out” before anyone asks. Ours would be: show us what you built, then we’ll look at whether you need more or need a different habit.
Tell people in advance what the limit message looks like and what it means. Put the usage screen in onboarding. And read the limit posts the way you read the success log: as a list of the people getting the most from the tool.
One thing to try this week
Next time someone tells you they’ve run out, reply “what did you build?” before you reply about credits.
Licence utilisation is one of the Adoption Diagnostic’s inputs. The seats that run out tell you as much as the ones nobody opens. About the Adoption Diagnostic
More lessons
- Adoption
Ask people to brag
- Adoption
Take the tool to where people already work
Get new lessons by email
Short lessons from a real Claude rollout, each backed by a number.
