For accelerators, incubators and university programs
Practice a difficult investor meeting.
FAUVION gives each founder in your cohort a private, AI-led mock meeting built from their own company brief and pitch deck. It is a live spoken conversation: they talk, and a sceptical practice investor answers aloud in an AI voice. It asks the questions they have not thought about yet, follows up, and at the end hands back a written account of where their story does not hold together — with coaching on pace, pauses and clarity.
The demo uses invented data and needs no account.
Founders rehearse in private. You get the part you actually need.
What a program manager sees
- Who has joined the cohort, and who has not.
- Who has started a session, and who has completed one.
- Nothing from any founder’s report. Counts of recurring themes across a cohort, in groups of at least 3 founders, are built but switched off until they can be made safe enough: the privacy page says why.
What a program manager never sees
- The founder’s pitch deck, or any text taken from it.
- The meeting itself, or anything said in it.
- The figures the founder’s browser measures while they speak — timing, pauses, pitch movement and loudness variation included.
- The founder’s own feedback report.
There is no recording and no transcript to see. The founder’s microphone audio goes from their browser straight to the live-voice provider (OpenAI) for the length of the meeting; this service never receives it, so it is not recorded, uploaded or stored here. Only the written report and the measured numbers inside it are kept.
This is enforced in the database, not by hiding buttons. The tables holding a founder’s brief and report have no manager access policy at all.
You set up the cohort
Create your program and a cohort
Name, dates, and a demo day if you have one.
Generate invitation links
Set how long each link lasts and how many people may use it. Copy it and send it however you already talk to your founders.
Watch completion, not content
A roster showing who joined, who started and who finished. Nothing from anyone’s practice material.
Each founder rehearses
Read the notice, then decide
Before anything is uploaded, they see what is collected, who can see it, and when it is deleted.
Write a brief, optionally add a deck
A PDF deck is read server-side for its text. Either half works on its own.
Sit the meeting, out loud
A live conversation with an AI voice that listens, follows up on what they actually said, and can be interrupted. It needs a microphone; headphones help. Up to 55 minutes of conversation (a session’s usage limit can end an unusually costly one sooner, with notice); they can mute their microphone or end the meeting whenever they want, but a meeting they leave cannot be picked up later, because nothing said is kept.
Read the report
Written once, at the end, by the same AI model that held the conversation: what held up, weak or unsupported answers, contradictions with the deck and the brief, and whether each question was addressed. Coaching on pace, pauses, clarity and filler words is the model’s impression from listening; the browser’s approximate timing figures sit beside it.
A 7-day deadline on everything a founder creates.
The clock starts when a founder first saves their brief, and it does not restart when they come back. A deletion job is queued for the deadline straight away. The deck, the extracted text, the brief, the session record and the report — measured figures included — are then removed by a cleanup job that the operator schedules — and until that schedule is set up, the founder’s data page says so. A founder can trigger the same cleanup immediately from their own dashboard. No audio and no transcript is ever stored here, so there is none to keep.
Two honest caveats, which we put on the privacy page too. Deleting from the live product is not the same as deleting from a hosting provider’s database backups, which roll off on their own schedule. And the live-voice provider (OpenAI) handles what it receives under its own terms: it may keep abuse-monitoring logs, which can include content, for up to 30 days, and we cannot delete those. We describe what actually happens rather than promising more.
What this is not
Not an investor, and not anyone’s stand-in
It does not impersonate a real investor, firm or person, and it has no affiliation with any of them. It is a practice interviewer. A founder may pick a firm as a practice focus; that only steers the questions, and no firm is involved.
Not a prediction about fundraising
There is no score, no rating and no probability of raising. Numbers like that imply a rubric nobody has, and founders read them as verdicts.
Not advice
Nothing it produces is investment, legal or financial advice. It is feedback on how a story holds up under questioning.
Not a grading tool for your program
You cannot read a founder’s report, so it cannot be used to rank them. That is deliberate: founders are only candid in rehearsal if rehearsal is private.
Not a judgement of the person speaking
The AI model hears the founder directly, and its notes on pace, pauses and clarity are impressions from listening, not measurements; it can mishear. Separately, the founder’s browser measures timing and pauses, how far pitch moved and how far loudness varied. Those are approximate numbers, and the pitch and loudness figures are never sent to the AI. Nothing infers emotion, confidence or personality, comments on accent or voice, or identifies anyone by voice, and it is never a verdict on whether a company will raise.
Look at it before you decide anything
The demo runs the whole thing on invented data — the manager dashboard, a founder’s session, and the report at the end. No account, and nothing is sent to us or to an AI provider. The demo meeting is a labelled simulation: the investor’s lines are scripted and the report is a pre-written example, not a response to what you say. If you switch on your microphone, your browser measures your turns on your device and the audio never leaves the page.