← All case studies
Atlas Ventures Venture Capital 5 weeks

Atlas · A multi-agent research system for a venture firm

A planner-worker-critic agent system that scouts companies, reads filings, and synthesizes weekly memos, replacing 2 analysts of grunt research with one clean inbox.

AI Agents VCLangGraphMulti-agentPython

01/ The challenge

Atlas was hiring junior analysts faster than they could train them, and most analyst time was spent on company scouting and basic diligence, work that hadn't changed in 5 years.

02/ The approach

A three-agent system: a planner that generates research questions, workers that fetch and read filings/news/web, and a critic that scores and rewrites outputs. Wired into their CRM and slack with daily digests.

03/ The outcome

Analyst FTEs replaced
1.5
Build to launch
5 wks
Eval-driven improvement
Weekly

Replaces ~1.5 analyst FTEs of routine work. Partners spend their time on judgement and meetings, not literature review. The system continues to improve with weekly evals.

“The agent system gave our partners back judgement time. Five weeks, end to end, and the weekly evals keep it improving without us touching it.”

Maya Iyengar
Partner · Atlas Ventures

Velo · An AI co-pilot for ops teams

Built an AI ops co-pilot that drafts SOP responses, triages tickets, and writes weekly performance digests for a 50-person operations team, cutting average ticket time by 62%.