Procurement is the part of a manufacturing business that nobody puts on a conference stage. It is suppliers, purchase orders, compliance documents, certificates of conformity, and the follow-up email chasing the supplier who has not answered the previous follow-up email. It is the plumbing that keeps a factory fed, and like plumbing it gets noticed only when it backs up: a missing part, a late certificate, a line that stops. Compri, a Milan startup founded in 2024, just raised €3.2 million to argue that this unglamorous corner is exactly where AI earns its keep, and not, for once, where it writes marketing copy.

The reason the argument lands now is that procurement stopped being a quiet cost centre and turned into a strategic one. A decade of stable supply chains made purchasing a clerical job. Then came a pandemic, a war on Europe’s eastern edge, tariffs, shipping shocks and a general sense that the parts you need might simply not arrive. Buyers who used to negotiate price now manage geopolitical risk, qualify second and third suppliers, and document compliance for a thickening stack of regulations. The work multiplied. The headcount, in most mid-sized manufacturers, did not.

A digital workforce, not another dashboard

Compri’s pitch is deliberately not “another piece of software to log into.” It sells what it frames as a digital workforce: AI agents that pull data from the systems a buyer already lives in, ERP platforms, emails, spreadsheets, PDFs and external databases, and then do the repetitive work themselves. Supplier follow-ups, document collection, compliance monitoring, order-confirmation checks: the tasks that eat a buyer’s day and require diligence rather than judgment. The company says the platform already serves more than 40 customers and cuts manual procurement work by up to half. That second number is the one that sells itself to a finance director, because it converts directly into either fewer hires or far more supplier volume handled by the same team.

The starting market is a deliberate choice, not an accident of geography. Italy is the second-largest manufacturing economy in Europe, behind Germany, and it is built on thousands of specialised mid-sized companies, the famous industrial districts that make everything from machine tools to luxury components. Most of them still run purchasing on Excel and rigid legacy ERP systems that were installed once and never loved. That is a large, under-digitised base sitting right on Compri’s doorstep, and the company has not stayed home: it already lists operations across Italy, the United States, South America, the Nordics, Germany, the Netherlands, Spain and France.

The honest worry is adoption, not capability

The seed was led by Picus Capital, with Shapers, the Italian Founders Fund, DFF Ventures and private investors joining, taking Compri’s total funding past €5 million. Picus is the early-stage firm with Rocket Internet roots that has a long history of backing unglamorous B2B software early, which is roughly the profile of bet being made here.

The risk is not whether the technology works. It is whether the customers who need it most will actually buy it. “AI for SMEs” is a category that has quietly buried a lot of startups, because the smaller and more traditional the manufacturer, the slower it is to adopt anything new, and the harder it is to reach with an efficient sales motion. A factory that has run procurement the same way for thirty years does not change because a Milan startup has elegant agents. It changes when the pain of not changing, a stockout, a failed audit, a buyer who quit and took the supplier relationships with them, becomes louder than the inconvenience of learning something new. Compri’s job is to find the buyers for whom that pain is already loud.

What is in its favour is the shape of the bet. The founders went after a real, boring, expensive problem on their own manufacturing-floor doorstep, the agents do work that is tedious rather than creative (which is exactly where today’s AI is most reliable), and there are paying customers from close to day one rather than a pilot and a promise. Forty customers and a halving of manual work is not a revolution. It is a wedge into a function the rest of the software industry has been too bored to fix. Boring infrastructure is where Europe quietly compounds.