Fintech/AI automation
Mr Assistant
Staff ask the organization once instead of pinging five people. Appointment requests go through the assistant.
Confidential commission. We describe the system. We do not name the organization or show their data.
Desktop assistant · Private corpus · Appointments
Result
Someone inside the firm can ask a normal question (who handles this, what we charge, what we offer) and get an answer from the organization’s own knowledge. If they need a meeting, they book it there.
The situation
Fintech teams run on specifics: products, rates, who owns which book, how a client gets on someone’s calendar. That material is written down somewhere and still asked out loud every day.
A public chatbot does not know the firm. A shared drive does not answer. They needed a personal bot for the company, trained on their world, not the internet.
Constraints
Answers had to be about this organization: contacts, rates, services. Generic finance advice was out of scope.
Booking had to live in the same conversation.
The company stays unnamed. No customer lists, no rate cards on this site.
Approach
A familiar chat in front of a corpus the firm already owns. Retrieval first: who-to-contact, rate, service. Then a composed answer.
When the intent is time with a person, the path switches to booking instead of inventing a calendar.
What we shipped
Chat trained on organization knowledge. Routing: whom to contact for what. Rates and services as the firm states them. Appointment booking from the same thread.
A desktop shell so it sits where people already work.
How it was built
A private knowledge layer, a conversational UI, and a booking action. We keep the model and store undescribed here so we do not sketch their security boundary.
What we’d do next
If they allow it: one anonymized question and answer on this page. Otherwise the story stays at the system level.