Deep dive with Martin, co-founder of Corso AI — a YC company building agentic workflows for factories and manufacturers, with a customer base concentrated in Asia and a deliberate focus on the parts of manufacturing that most software has skipped.

Corso started before the idea did. Martin and his co-founder Daichi wanted to build together and looked for the domain where they had the most experience. Martin had interned at General Motors. Daichi had done PCB work as an electrical engineer at a startup. They picked manufacturing because they knew it. Over winter break, Martin visited five or six factories in person to figure out what the operators actually needed. The consistent answer was manual data entry between email, Slack, Teams, and the company's ERP or CRM, plus follow-ups with suppliers and procurement chains. That is where Corso started.

The product, Atlas, integrates directly into a manufacturer's existing ERP. The onboarding is deliberately slow. Atlas does not start executing workflows on day one. It sits in the system, watches how emails come in, how responses go out, how the team assigns roles and shares information internally. Over the first few weeks it learns the company's patterns. Then it starts suggesting outputs, gets corrected by humans in the loop, and only takes autonomous action once the accuracy is near 100%. The efficiency gain Martin cites for procurement teams is 8 to 15 hours per week saved on manual work.

Corso is also building what Martin calls an intelligence layer on top of the workflows. Executives at manufacturers do not just want the manual work automated. They want insights the current tooling cannot give them. Martin's vision is a system where an incoming RFQ gets analyzed against historical pricing, current commodity prices, shipping bottlenecks, and geopolitical constraints in real time. The system would then suggest what the quote should actually be, informed by all of it.

The company's ICP is in Asia, which is why they did a pre-YC trip through Japan, Taiwan, and China visiting manufacturers before starting the batch. WhatsApp integration is live. WeChat integration is in progress, and Martin is direct that WeChat is the hardest platform they have ever built against. The team is moving to LA after YC because there are more factories to iterate with in person than in San Francisco.

Also covered: why Corso pivoted from building an AI-native ERP to integrating with legacy systems after talking to customers, why the pace of AI progress makes planning past three to six months feel pointless right now, and the long view on which parts of manufacturing procurement should stay human, and which should not.

—

Agentic Stories is the weekday briefing on the AI agent economy — governance, security, and deployment. Deep Dives drop on off-days with founders building in the space. New episodes Monday, Wednesday, Friday.

agenticstories.ai