Research that updates itself. Judgment that does not.
Investment research automation keeps a living memo current as a deal moves forward, while human conclusions remain clearly human.
The situation
The VC fund receives applications from startups seeking investment. Collecting the application was never the problem. Everything after it was.
Analysts manually pulled information from pitch decks, company websites, founder profiles, meetings, market research, and internal discussions, then shaped it into the fund’s existing memo format. As a deal moved through the pipeline, that picture kept changing: new meetings, new research, and new observations from the investment team. Without a connected system, someone had to gather it again, update the CRM, update the memo, and confirm that the current version reflected what the team knew.
The constraint
The system had to automate research and documentation without automating investment judgement. Letting a model assemble the whole memo and asking people to edit it afterwards would mix generated and human text. Once that happens, nobody can tell which is which, and a memo nobody trusts is worse than no memo.
Human-written sections are therefore stored separately from generated ones. AI agents update the research around them but never overwrite an analyst’s conclusion or an investment-committee comment. A reader can always tell which parts are researched, which are synthesised, and which are someone’s judgement.
Every meaningful change is traceable: when the memo changed, what new information caused it, and what it said before.
The second constraint was cost. Running AI-assisted deep research from scratch for every company at every stage is expensive and mostly redundant when much of the information has not changed since the last run. The workflow had to be deliberate about which events trigger research and which facts are stored and reused.
What we built
We built a research and memo system around the fund’s existing deal flow. A startup applies with its website, pitch deck, and basic details. When the deal reaches the right stage, the system expands that into a company profile: the deck is read and structured, the website is researched, founder and team profiles are added where available, and market research covers sector dynamics, competitors, estimates, and regulatory considerations. Sanctions screening runs automatically at a later stage.
Everything lands in the fund’s own memo format, not a new one.
From there, the memo stays live. Meeting transcripts between the fund and the startup are processed and added as context. New research refreshes only the sections it affects. Analysts and investment-committee members write their views in sections the system will not touch, and the next update works around them.
Process diagram
Where it stands
The system is live across the fund’s deal flow.
Analysts spend their time reviewing, challenging, and recording judgement rather than repeatedly collecting the same information at every stage. The record compounds: the team can see not only the current view of a company, but how that view formed, what changed it, and what the committee concluded.
Next
Personalised video outreach for qualified inbound leads
Ready to boost business results with AI automation?
Book a 30-minute operations review. We will look at where manual work is limiting your operation and whether there is a practical opportunity to create capacity.
Book a 30-minute operations reviewSwiss precision. Valley speed.