• VC firm Flybridge is using AI for writing investment memos, a critical part of any diligence.
• It's making the generator available for public use via its website and GitHub.
• Flybridge aims to help founders refine pitches and save investors time on routine tasks.

When Chip Hazard, a longtime startup investor at Flybridge, finds a startup he wants to fund, he writes a multi-page investment memo for his partners, outlining in detail all the risk factors and opportunities of a business. Now, he's using AI to help write these documents and saving hours of work.

It's the latest example of how venture firms are eating their own dog food. Investors aren't only funding startups exploring the application of large language models and "agents." They're also experimenting with these tools internally to improve how they source deals, research companies before investing, and track performance.

"If we could free up capacity on more routine tasks and therefore give us more time for judgment," said Hazard, "that's a good trade in our business."

This week, Flybridge is unveiling the AI-powered memo generator to the public — free for anyone to use.

How it works

Following the release of ChatGPT, Hazard made a cheeky bet with an associate at his firm: he promised a bottle of wine from his private collection if the associate could build an artificial intelligence that writes investment memos.

Flybridge's memo generator looks like a simple web form. The user uploads a pitch deck and a transcript of the pitch, fills in the round size and valuation, and adds links to the founder's LinkedIn page and the company website.

Under the hood, the memo generator was built on top of OpenAI's o1 model, according to Daniel Porras Reyes, a self-taught developer and Flybridge associate. This model is considered superior to its successors because o1 was designed to spend more time thinking before providing an answer, improving its output quality.

Flybridge used CrewAI, a portfolio company, to build "agents" — a new set of artificial intelligence tools that can work autonomously without much human supervision. Those agents can search the web through Exa, a search engine designed for use by agents, and create content about a company's competitive landscape or market size.

In as little as three minutes, the memo generator spits out a Word document with sections on the opportunity, risks, business model, go-to-market strategy, and team. It also proposes a list of follow-up questions for the founders.

Hazard said that by releasing the tool to the general public, Flybridge hopes that founders will run their pitch decks or investment memos through the generator before they meet with the firm. This could give founders an idea of how their pitch comes across so they can smooth out the kinks ahead of time.

Like most generative AI tools, the memo generator sometimes gets it wrong. It might leave a competitor off the list, said Porras Reyes, or reach a wrong number in the financial projections, Hazard added. Still, Hazard said the product shaves hours off of production; he spends less time writing a founder's bio, for instance, and more time thinking about whether they have the right characteristics and the best idea.

Hazard said if the memo generator was more capable, users might be tempted to "check their judgment at the door." He continued, "The point is to have it 'good enough' that you can then start to really apply your judgment."

Founders and investors can access Flybridge's open-source investment memo generator on its website or clone the project on GitHub.