One agent per client, not a chatbot
Lisa is Redexe's AI agent, in a dedicated instance for each client: she lives in the working group, knows only the data of the client she serves, and actually does things — delivers reports, answers on the numbers, launches computations, simulates scenarios. Since the summer of 2026 she serves every Redexe client.
What she is
Every Redexe client has their own working group, in the chat they already use every day: in it are the client's contacts, Riccardo Donati and Lisa. No app to install, no portal, no new password.
Lisa reads everything but only steps in when needed: she participates, she does not dominate. She answers with the numbers highlighted and the conclusion before the details, always quoting the analysis date. If a job failed she says so plainly; if she does not know something, she admits it without fuss.
She is not a generic assistant trained to chat: she is an AI framework developed by Redexe, with a precise perimeter — her client's data, documents and computations. Everything the client needs to know about Lisa, Lisa teaches them herself.
What she does
Capabilities vary from client to client — the perimeter is bespoke — but the substance is this:
A typical meeting
The scene is real, from the summer of 2026. The investment committee is discussing the model portfolios; in the group someone writes: "Lisa, try +2 on gold and offset on governments". A couple of minutes later the numbers arrive: the effect of the move on diversification, risk and expected return, before and after.
Compound moves are all-or-nothing: either they all pass or none does — the scenario never stays half-applied. You explore with "show" and "list", you go back with "reset". And when the committee is convinced, "save" publishes the new weights — but only after Lisa has asked for explicit confirmation and received it. A "sounds fine" is not enough.
It is the difference between a simulation tool and a colleague who speaks the language of the meeting: no interface to learn, no forms to fill. The engine underneath is the same as ever: the Redexe quantitative platform.
Security and governance
Lisa's limits are not etiquette recommendations: they are platform constraints, the same for every agent.
The people
In every working group, next to Lisa, there is Riccardo Donati. When a request is ambiguous, sensitive or out of perimeter, Lisa does not improvise: she logs it and brings it to him. When content is delicate — the commentary on a projected balance sheet, the analysis of a bond — it does not go out without passing through a person.
It is the operating model we chose: the agent does the routine, always, immediately and without slips of attention; the person does the judgement, the research and the supervision. It is not a fallback: it is the design. And it freed the founder for the one thing a model cannot do alone: building the new models.
FAQ
No. A chatbot converses; Lisa operates: she delivers documents, launches computations on the Redexe quantitative platform, validates incoming data, simulates scenarios for the committee. And she is dedicated: every client has their own instance, with its own perimeter.
Nothing: no app, no portal, no password. Lisa joins the working group in the chat the client already uses every day, together with the authorised contacts.
She can, like any system — which is why the service is built not to pretend: she only answers on data she has actually read, quotes the analysis date, declares failed jobs, and on delicate matters hands over to a person. Irreversible actions always require human confirmation.
No, never. She reports the model's analyses — "according to the latest available Redexe analysis" — as decision support. Choices remain with the client and their governing bodies.
Lisa is an AI framework developed by Redexe, combining state-of-the-art language models with the proprietary quantitative platform built over twenty years. Implementation details are confidential.
Yes: every Redexe service is delivered this way today. Activating a new agent is a matter of days, because the platform is designed to be configured per client, not reprogrammed.
The services
Risk reporting Portfolio optimisation ALM for foundations AI securities analysis Production ICT Real estate
Contact
Message Riccardo Donati on LinkedIn: we start from the perimeter — your data, your analyses, your contacts — and Lisa comes after.
Redexe works strictly B2B: banks, portfolio managers, advisory firms, foundations and institutions.