The context
An insurance general agency, where most of the value is created face to face with the client, and where everything said during a meeting then has to be honored.
The problem, as it was lived
A client meeting produces two things: a conversation, and a list of commitments made out loud. Send a document, call a policyholder back at a given date, check a contract, come back with a quote. Those commitments are written down nowhere, except on a notepad or in the agent’s memory.
The notepad holds up as long as the day is calm. Past a certain pace, it becomes a bet. And in this business, a callback that doesn’t happen isn’t an administrative slip. It’s a relationship taking damage.
On top of that constraint sits a second, non-negotiable one: the portfolio data is policyholder data. It can’t go wandering.
What we built
The recorded meeting becomes an automatic write-up. Every commitment made during the meeting becomes a proposed task, which waits for the agent’s approval before it exists. Nothing goes out without him: the system is not allowed to create an action the agent hasn’t seen.
On the data side, the rule is set upstream of the processing, not after it. Portfolio extracts are pseudonymized on the machine, before they ever touch an AI. Client names never leave the practice.
What it changed
The notepad is no longer the weak link. The agent walks out of a meeting with the write-up already drafted and the list of commitments already laid out in front of him, to approve or discard. He keeps control over what exists and what doesn’t, and he keeps it without re-typing anything.
The proof
In production.
If you’re in a regulated business and that’s precisely why you haven’t tried anything yet, that constraint is an input to the audit, not an obstacle discovered at the end.