Korso Brings an AI Intelligence Layer to the Back Office of Manufacturing

Ask any plant manager where the bottlenecks are and the answer rarely involves the factory floor itself. It is the paperwork around it – the RFQs that need processing, the quotes that need chasing, the suppliers that need reminding, and the production schedules that need rewiring every time reality intrudes. Korso, a Y Combinator Spring 2026 company founded in San Francisco by Daichi Hiraoka, Alex Liu, and Martin Pan, is building AI agents for manufacturing operations, positioning itself as the intelligence layer for modern manufacturing.

Korso’s agents plug directly into the ERP and CRM systems that manufacturers already run – systems that are notoriously messy, heavily customized, and legendarily painful to integrate with. Rather than asking customers to rip anything out, Korso’s agents learn existing workflows and act on behalf of the team: triaging incoming requests, following up on delayed orders, notifying customers, and escalating to humans only when judgment is actually required. The explicit goal is not to replace enterprise systems but to make them work harder, so operations teams can spend less time on coordination and more time on the work that genuinely moves the business.

Because manufacturing is unforgiving of errors – a missed shipment is not a mislabeled email – Korso has engineered the product around verification, scoping, and auditability from day one. Every agent action passes through verification layers. Tool access is scoped. Critical operations go through dry-run validation and secondary review, and full audit trails record what was done, why, and when. The company’s orchestration layer is explicitly built for the messy realities of real factories: long-running workflows that can span weeks, checkpointing that never loses progress, and integrations hardened against the quirks of ERP systems that most AI products break against almost immediately.

The founding team is building with operational credibility. One co-founder previously worked at General Motors and trained at the University of Pennsylvania, bringing firsthand exposure to the scale and constraints of industrial operations. The trio is focused entirely on building the intelligence layer for manufacturing – a positioning that is both narrow enough to be credible and broad enough to accommodate years of product expansion.

The market context is favorable. U.S. and allied manufacturing has re-entered the policy spotlight, with reshoring, defense-adjacent production, and advanced manufacturing all driving a wave of investment. At the same time, the back-office teams supporting these operations are chronically under-resourced and drowning in email threads, spreadsheets, and portal logins. AI agents for manufacturing operations that can responsibly reduce that coordination load – while respecting the zero-tolerance error culture of industrial work – are exactly the kind of product that enterprise buyers in this space have been waiting for.