Business Insider surveyed investors from its 2026 Seed 100 and Seed 40 lists about how they use AI in daily work. The responses, published May 20, describe agents embedded in nearly every stage of the venture workflow: sourcing, diligence, portfolio monitoring, scheduling, and internal tooling.

Back Office Agents and Portfolio Intelligence

Salil Deshpande, general partner at Uncorrelated Ventures, told Business Insider he “taught my AI agent to run my entire back office.” The agent handles scheduling, runs email campaigns, and triages incoming LinkedIn invitations, auto-accepting portfolio founders and relevant VCs while surfacing others for review. Anyone on Deshpande’s team or investor base can email with “portfolio doctor” in the subject line to query board decks and investor updates stored in Dropbox. The agent synthesizes an answer with specific citations and replies to the thread.

Ann Miura-Ko, partner at Floodgate, described a different approach: visiting a dozen AI-native companies in person over several months and using AI to ingest all field notes, cross-reference them across companies, and surface patterns. “This has become the empirical backbone of my analysis of the AI-pilled startup,” she told Business Insider. “The patterns are converging in ways that are reshaping how I evaluate investments and founders.”

Sourcing Before the Market Sees It

Alex Bard, managing director at Redpoint Ventures, described internal AI systems that identify “high-potential founders leaving top companies and detect early momentum signals in inception-stage startups,” according to Business Insider. The signals include hiring velocity, product launches, and network activity.

Anne Dwane, cofounder of Village Global, said agentic workflows help the firm “pinpoint the ‘first-call’ angels in today’s most promising founder ecosystems,” extending the firm’s network into emerging talent pools continuously.

Sarah Smith, general partner at Sarah Smith Fund, built a custom 100-point AI scoring framework that evaluates every inbound opportunity and decides whether to take a first meeting. The system weights the Stanford ecosystem, founder-market fit, and early signs of exceptional achievement.

From Evaluating Startups to Building Them

Jeff Fluhr, venture partner at Craft Ventures, said he uses AI coding tools daily to prototype startup ideas and evaluate them. “In the last four months, I’ve built four such products,” he told Business Insider. “That has changed how I evaluate new ideas: instead of debating them in the abstract, I can get to a working prototype quickly and see what actually holds up.”

Shan-Lyn Ma, cofounder and co-CEO of Zola, built a custom Gemini “Gem” that tracks fundraising and exit news from roughly 50 portfolio companies daily, delivering automated summaries.

The Broader Pattern

The survey results corroborate a trend VC Lab documented in April: autonomous AI agents are collapsing the operational advantage that large VC teams once held. Decile Group, VC Lab’s parent organization, built an “agentic VC” platform with OpenClaw integration where agents ingest pitch decks, score deals against fund theses, run competitive landscape analysis, and produce diligence checklists. A counterfactual agent then argues against each deal, giving GPs both sides.

The convergence is straightforward. Seed rounds now routinely exceed $10 million, with outliers like Thinking Machines Lab ($2 billion) and Advanced Machine Intelligence ($1.03 billion) redefining the category, according to Business Insider’s Seed 100 report. As capital concentrates around fewer, larger bets, the investors who can evaluate more companies faster with less headcount gain a structural edge. Agents provide that edge.

Julie Lein, managing partner at Urban Innovation Fund, offered the counterpoint: AI adoption has made her firm “deeply aware of the importance of humanity” in the startup world, pushing the team toward more in-person events and founder meetings. “In the age of AI,” she told Business Insider, “the only thing that stands out is real human-to-human interaction.”