Secure Deal Room + AI Insights
A secure deal room, with AI that reads it and describes what happened
Rules check the room's language before anyone sees it, and plain-language summaries describe recorded activity afterwards. AI is a feature of the workspace — never a legal, investment or fiduciary decision-maker.
Rules first, then the reading
Deterministic rules run before any model does. Guaranteed or projected returns, risk-free language, a claim that a regulator approved or recommended the offering, pressure language, profit testimonials, a comparison to a famous company, a missing legend — those are decided by rule, and a rule that blocks cannot be talked out of it.
The model then reads the same material and may only add warnings. It can never clear something a rule has blocked. That order is the whole design: an issuer should not be able to argue its way past a rule by rephrasing the prompt.
What it reads
- Every deal-room section and every uploaded document
- The pitch deck, scored twice — findings that block, and subjects that merely score
- The public notice and its legend, against the room's exemption
- Draft investor emails and social copy, before they go out
What it refuses to do
- It does not read text inside a picture, and it does not listen to a video. The portal says so where the scan is shown.
- It will not judge a deck it could not read, rather than guess from a filename
- It does not recommend an investment to anybody, and it does not rank offerings
- It does not decide that an exemption is available
From behaviour to an observation
Every view, document open, video position and button press in the room is recorded against the offering. The report describes what was observed — for example, that most visits which opened the summary ended there — and links each line to the activity behind it. It does not say why, and it does not tell the issuer what to change.
With a handful of visits, a pattern can be coincidence, and the report says so rather than presenting it as a trend.