Prexo

Soon after I arrived in Canada, a WFG agent got in touch, the way they do when you are new and meeting people. WFG is a financial services company whose agents sell insurance and investment products, and for those agents, prospects are everything: the people they have met and need to follow up with. When I saw how they kept track of them, I was shocked: they used spreadsheets, and the agents with spreadsheets were the best in their industry. A friend and I offered to build them something better.

An agent walks out of a meeting, opens Telegram and says what happened: a mortgage renewal coming up, a promise to send numbers, call back in two weeks. Prexo transcribes it, works out which client it is about and proposes the changes: a note on the file, a task, a date to follow up. Then it waits for the agent's answer. Before the next meeting, the brief arrives on its own: who the person is, their family, where things stand, what was promised last time and what to bring up.

All of the code was mine, from May to August 2026. The rule I held to was that the AI never changes a client's file by itself: it proposes, and the agent says yes, partly or no. Every one of those answers is counted, so we could see how often it was right.

It works and it is live, but agents never used it much. We were not a big company people trust easily. We were people like them, and prospects are the one thing a WFG agent guards, so they worried their client lists might be worth something to us.