Magentic raises £13M to put AI 'digital workers' into factory procurement

Magentic, a London startup, has raised a £13M Series A to point AI agents at one of the least glamorous and most expensive corners of a big company: buying things. Its software runs what it calls “digital workers” — multi-agent AI systems it nicknames Mages — that take on procurement and supply-chain work for large manufacturers, from choosing suppliers and negotiating contracts to running orders and clearing invoices. The round was led by Felicis, with Sequoia Capital and The Westly Group also taking part.

The pitch rests on a real, unsexy problem: procurement at a global manufacturer is a swamp of fragmented data spread across ERP systems, spreadsheets and PDFs, and the work of making sense of it is slow, manual and enormous — “billions of rows of data, tens of billions in spend,” as the company puts it. Magentic’s agents sit on top of that mess, working through Microsoft Teams, email and a company’s own systems, and taking tasks end to end while keeping humans in control of the decisions that matter. The claimed results are concrete rather than hypothetical: early customers are said to include three of the world’s ten largest beverage companies, with one processing over a million orders a year through the system and another crediting it with $4M of savings.

The founders give it credibility. Chief executive Robin Van Aeken is a former McKinsey and OpenAI hand, and chief technology officer Odhran O’Donoghue also came from OpenAI — a background that helps explain both the multi-agent architecture and the willingness of investors like Sequoia to back a procurement tool. That an increasing number of people leaving the frontier labs are starting companies in London, rather than only San Francisco, is a quietly encouraging pattern for the UK.

It also fits a thread we keep pulling on: AI agents being aimed squarely at real, back-office enterprise work rather than consumer novelty — after governing enterprise AI and, this same week, commercial risk and disputes, now procurement. The caution is the one that applies to all of it: procurement means real money and binding contracts, so “humans in control” is not a nicety but the whole basis of trust, and the test is whether the agents stay reliable on messy, high-stakes supplier data over time. But the direction of travel is clear, and Magentic is a serious entry in it.