From 265 to 5,000: What We’ve Learned Building AI Into Our Investment Process

HearstLab's Lisa Burton took the stage at this years AI Pitch & Showcase during NY Tech Week to share how HearstLab is embedded AI into its investment workflow, widening the top of the funnel while keeping human judgment at the core.

A year ago, HearstLab made a conscious decision: we would not just talk about AI, we would build with it. At this point last year, we had evaluated 265 startups. This year, we have evaluated more than 5,000.

Same team. Same hours in the day. A very different way of working. The shift came from building AI into the workflows our team already uses every day, especially at the top of our investment pipeline.

Meet Ivy, HearstLab's AI Investment Agent

Building with AI meant identifying our bottlenecks, creating what would save us time, and giving our AI tools a job description. Now, we have three AI teammates on our team: Margo, Theo, and Ivy. Margo handles CRM and partnerships. Theo runs our reporting and automations. Ivy is our sourcing and diligence agent, and that's where much of our investment work now begins.

Before Ivy, our AI investment agent, a new Y Combinator batch meant hours of manual work. Someone on the venture team would research every company, look up every founder, assess whether the startup could be a fit for HearstLab, and log the information into our systems.

Now, Ivy handles much of that first pass.

She screens companies against our criteria, enriches the data, flags potential strategic alignment with Hearst, suggests real use cases, and recommends whether we should reach out or pass. When the team agrees there may be a fit, Ivy can also draft the first outreach email.

To date, Ivy has sourced and screened about 4,600 companies and saved our team roughly 2,000 hours, the equivalent of about three full-time employees' worth of capacity.

What AI Changed About Our Workflow

The point of Ivy is not to replace investors.

It is to give experienced investors more time for the work that actually requires them: founder meetings, investor relationships, diligence conversations, portfolio support, and judgment calls.

That distinction matters. In venture, more data is only valuable if it helps you ask better questions. AI can organize information, surface patterns, and speed up research. But it cannot replace context, trust, or lived experience.

For us, AI has become a way to widen the top of the funnel without diluting the human layer of the investment process.

What Our Data Is Teaching Us

Once our data lived in one place, in a consistent format, we could start asking better questions of it. One question we asked: what signals in our portfolio seem connected to startup success?

Founders who hit, or came close to, their financial projections succeeded about 70% of the time. Founders who severely missed their projections have not yet produced a success case in our data. The sample size is still small, so we treat this as a trend rather than a rule. But it has already changed how we think about diligence.

The Human Layer Matters More Than Ever

The more we build with AI, the more convinced we are that human judgment matters.
AI helps us see more companies, move faster, and ask sharper questions. But venture still comes down to people: understanding founders, building trust, spotting nuance, and knowing when the data is pointing to something worth a deeper conversation.

For HearstLab, the promise of AI is not that it removes people from the process. It creates more room for the work only people can do.

For founders and investors building with AI: what is the biggest operational shift you are seeing right now?