At Social Leverage we're a small team running a lot of surface area: team meetings, founder calls, portfolio check-ins, LP communications. The raw material of venture is information, and most of it was living in someone's memory or half-captured in notes. We started building things to fix that. One problem at a time.

What we ended up with isn't a platform or a grand system. It's a set of agents and workflows that help a small team operate with the output of a much larger one. Here's what we built, and what we learned.

Conversations as a System of Record

The first problem we tackled: turning meetings into something retrievable.

Most of the work in venture happens in conversation, and historically, that information evaporates. A great call with a founder becomes a vague memory. A portfolio check-in with a dozen updates goes mostly uncaptured.

We built a system that pulls transcripts from every meeting (team, pitch, portfolio, investor) and runs them through an AI pipeline that identifies companies, people, and key themes, then logs everything automatically into our CRM. No manual note entry. No relying on someone to remember to update a field.

The result sounds simple, but the compounding effect is real: every conversation becomes searchable context. When we're prepping for a call, we can pull up a company and see every time we've discussed it, what we said, who was in the room. Organizational memory goes from fragile to durable.

A Portfolio Monitor That Actually Runs

Staying current across 100+ portfolio companies is one of those things that sounds manageable and isn't. Things fall through the cracks: a major hire, a press mention, a market signal, unless someone is actively looking for them.

We built an internal agent that continuously monitors news and updates across our portfolio and surfaces what matters. No one is manually checking. The agent delivers relevant updates, and we package the best of it into our weekly newsletter: portfolio achievements, media shoutouts, milestones worth sharing.

The shift is from reactive to ambient. Information finds us instead of waiting for us to go looking.

From Blank Page to First Draft

Investment research is another area where AI has changed our workflow more than we expected.

We use AI heavily in the research and memo-writing process. Given a deck, call notes, and some context, we can generate a first-pass investment memo, map out the competitive landscape, and surface the key risks and questions worth pressing on, in a fraction of the time it would take starting from scratch.

This isn't about replacing judgment. The judgment is still ours; that's the whole job. What AI eliminates is the blank page. We're editing and refining instead of bootstrapping. That compression, from inputs to a structured first draft, meaningfully changes how fast we can move and how much we can cover.

LP Updates That Don't Take All Week

One of the more underrated use cases has been LP communications.

We get a constant stream of unstructured updates from portfolio companies: emails, metrics updates, check-ins, notes. Synthesizing all of that into coherent, consistent LP updates used to be a fragmented, time-consuming process. Something you'd put off until you couldn't anymore.

We built an agent that reads across these inputs, structures them into a consistent format by company, and drafts monthly LP updates. What was once a multi-day scramble is now a repeatable workflow. We spend our time reviewing and sharpening, not assembling.

Shipping a New Website

Gary on our team recently redesigned and relaunched the Social Leverage website, and the process looked nothing like any web project we'd done before.

Instead of Figma mockups, agency handoffs, and long iteration cycles, we worked directly in a tight loop. The same person thinking about positioning could immediately translate that into a live product. Copy and structure iterated in real time. Code generated, refined, and shipped without a full development cycle standing between idea and execution.

The biggest shift wasn't speed, it was ownership. AI collapses the gap between concept and execution in a way that hands creative control back to the people closest to the thinking. What used to take weeks to months across multiple people happened in a single continuous workflow.

What We've Actually Learned

A few things that held up across all of this:

Start smaller than you think. The best workflows came from solving one specific, annoying problem, not from designing a comprehensive system. The comprehensive system emerged later, after the small things worked.

Embed into existing behavior. If an AI workflow requires people to change how they work, it won't get used. The best ones live inside tools and habits that already exist. Our agents plug into the communication flows already in motion.

Optimize for outputs, not analysis. The biggest gains came from turning inputs into something usable: a memo, an update, a structured brief. AI that surfaces information is helpful. AI that produces a first draft you can react to is transformative.

The thing we keep coming back to: AI doesn't replace judgment in venture. What AI does is free up more of our time and attention for exactly that. Less organizing, more deciding.

Social Leverage is an early-stage venture fund. If you're building something, we'd like to hear from you.