At the end of each quarter, the team at Toronto-based Ripple Ventures would spend weeks  gathering updates from dozens of portfolio firms scattered across emails, PDFs, WhatsApp messages and meeting notes, before turning that information into reports for investors. 

Today, managing partner Matt Cohen says much of that work is handled by AI in just a few hours. 

The reporting process is one piece of what the early-stage venture capital firm calls “Ripple OS,” an internal AI system it started building in earnest about a year ago. The platform runs on about 15 AI agents that help carry out market research, source investment opportunities, recruit talent for portfolio companies and connect them with potential customers and investors. It also helps with due diligence, writes term sheets and tracks company performance. The AI-powered operating system lets Ripple’s four-person team do the work of up to about 100, according to Cohen. 

“It’s like having the best COO, CFO and CEO all in one application for us to do a lot of very sophisticated tasks,” Cohen told The Logic. 

Ripple’s platform offers a glimpse at what the AI-powered future of venture capital could look like—skeleton teams of investors overseeing dozens of agents sourcing deals and managing investments. It also shows how a small team armed with the right AI assistants may be able to compete with far larger organizations.

Many Canadian tech investors have been early adopters of AI, a technology that is both creating a new generation of startups to invest in while threatening to render others obsolete overnight. While many VCs may use Claude Code to build a financial model, or ask ChatGPT to draft a LinkedIn post, Ripple has taken it several steps further by building an agent-run system that underpins almost everything it does.

Ripple OS has attracted attention from outside the firm. Cohen said other investors have asked to buy the technology, which he describes as a packageable “cheat code” that VC firms could use to boost their productivity and, ideally, returns. So far, he’s declined their offers, preferring to keep the AI system as a competitive advantage for the firm. 

Ripple began experimenting with AI about two years ago, when it became clear how disruptive and advantageous the technology could be. It only started getting serious about using AI last summer after hiring a head of AI, Jay Lee, a former Ripple intern who had spent three years at McKinsey, working on the consulting firm’s AI initiatives. “He came in with this idea that we should start building our own systems to run a lot of the tasks and work that we do at the firm,” said Cohen. 

Lee set about building what’s effectively a software system that houses all of the firm’s institutional knowledge, scraped from years of emails, deal summaries and other internal records. 

The new AI system was first put to work sourcing deals and carrying out market research. Ripples used AI to scan founder profiles and online activity, identify potential deals and analyze industries it was interested in. From there, the system expanded into nearly every aspect of the firm’s work. Today, employees can query Ripple OS through the firm’s Slack and Notion workspaces, asking for specific information on a company, founder or details about an investment, for example. They can also use the technology to model investment returns, estimate when the firm will need to pull more capital from its limited partners, and test different investment strategies before writing another cheque. 

Mike Kim, co-founder and CEO of cybersecurity startup Mycroft, said his own experience as a portfolio company has changed since the arrival of Ripple OS. Kim said the VC firm—which led Mycroft’s pre-seed round in September 2024—has always had a wide network and has been good at connecting his team to customers, talent and investors. In the latter half of last year, he said, “that got turned up to 11.” Now, he said, “they’re able to introduce me to people that they don’t even know they have access to.” 

For example, if Kim is looking to hire an engineer, he can send out a request through Ripple’s Slack workspace. Within minutes, Ripple OS returns a report identifying promising candidates and whether they’re connected to a Ripple partner or founder who can make an introduction. Kim said other investors may take a week or two to produce a similar report. 

It’s comforting, Kim said, to have an investor that’s as steeped in AI as he is. Mycroft is built on AI and he said Ripple can understand their work on a deeper level, based on the firm’s experience. “That helps really reflect on how they can empathize with founders a little bit more,” Kim said. 

Ripple may be ahead of many of its peers, but plenty of VC firms are using AI in increasingly sophisticated ways. Emily Walsh, a lead investor at Georgian, said her firm, for instance, uses the technology to create briefs on startups, summarize notes, review due diligence material and to do a first pass on financial analyses. 

Walsh said there are barriers to getting more out of AI. The systems can get things wrong, she said, which can make them hard to trust. “For anything that touches numbers or other critical pieces of data, we rely on built-in sanity checks, a human confirming before anything is finalized and the legacy process kept as a fallback,” she said. 

Cost is another factor. Georgian has a defined compute budget, said Walsh, and the proprietary data and infrastructure that gives AI systems an edge is expensive to develop and scale. 

For Ripple, being a relatively small and young firm—it was founded in 2018—has made it easier to go all in on AI, rather than tack the technology onto old systems. “We have the flexibility, that wasn’t built on legacy infrastructure, to do a lot of the things that we do,” Cohen said.  

Walsh said there’s a temptation to “AI-ify everything.” Giving into that urge won’t make sense for every firm, she said, adding that it takes discipline to focus on where the technology can actually improve outcomes. 

Still, AI has changed her expectations of what a modern venture capital firm will look like. “The firm of the next couple of years isn’t one where agents make the decisions. It’s one where agents do most of the front-end production and humans retain the judgment,” she said. “The firms that get this right will feel less like they’ve automated investing and more like they’ve given a small team the leverage of a much larger one.”

Cohen said he’s living that prediction right now. For all the time Ripple OS has saved the firm, it hasn’t created any downtime for its partners. “We’re actually busier than we’ve ever been,” he said. “We’re running on fumes, but we’re getting way more work done in a way shorter time period.”