Every deployment has a list. The early adopters, the champions, the people who turn up to every demo with feature requests. And then the other list: the people who are perfectly good at their jobs and see no reason to change how they do them.
This is a story about a man from the second list.
The setup
A lending director at a UK bank, decades in the job, dealing in the kind of lending where the source material arrives as a 70-90 page information memorandum. He had a personal ChatGPT subscription and used it regularly. He just was not using ours.
For months, the platform sat there and he did not log in. By every adoption metric we track, he was a late adopter, and the honest version of the story is that no amount of evangelism from me was moving that number.
What we believed
The early version of the platform asked more of its users, more structure and more steps, and his work did not arrive in that shape. We believed, like most vendors believe, that the gap was awareness and training. So we kept offering both, and his adoption line stayed exactly where it was.
What changed
The platform moved to a conversational experience: a place to put a document and ask for what you actually want.
He started uploading the information memorandums. All 70-90 pages, with a reusable script he keeps and pastes in. Back comes a 2-3 page summary, which he exports and emails on.
His trust came from his own red pen. He checked the output against the source document, line by line, on four or five live deals with real money attached. His verdict, in his words:
This is really good, it's looking at the information I would be looking at, and it's accurate in its analysis.
The bit that surprised me
He also runs customer research through the platform, and he compared it, same prompt, with his personal ChatGPT. He found ours consistently more accurate and more relevant to his day-to-day banking work. His phrase for it was that the platform "feels aware".
He had a general-purpose AI in his pocket the whole time. When the work mattered, he chose the domain-specific one, because it had the banking context his deals actually live in.
The bill
Months of a flat adoption line for one of the most senior potential users on the platform. The early design gave him no way in that fitted how his work arrives. He never raised a ticket about it. He simply stayed away. A quiet flat line on a senior name rarely comes with a complaint attached. Ask why before you assume disengagement.
A second lesson came free in the same session. His wider team's usage had dipped the previous month, and when we asked about it, the answer was that the month had been quieter: lighter deal flow, fewer files, less need. Their usage was tracking deal flow. A re-engagement plan would have been a fix for the wrong problem.
Take this with you
Late adopters convert when the tool meets the shape of their actual work, and they stay converted when they can verify the output against their own expertise. Design for the verification moment, make it easy for the sceptic to mark the AI's homework, and the sceptics become your best evidence.
%201.png)
