Somewhere in a Milan office park, a team of software agents is ordering industrial parts.
Not suggesting orders. Not flagging them for review. Actually ordering them—chasing quotes from suppliers, negotiating delivery schedules, tracking shipments, maintaining compliance documentation. The agents work in teams of five, specialized by function, and they never sleep. They're the core product of Compri, a startup that thinks it's figured out what's wrong with procurement software: it makes humans do too much work.
This isn't the typical "AI will transform everything" pitch, though there's plenty of that going around. Compri's founders, Edoardo Arbizzi and Edoardo Gava, came to the problem through the unglamorous trenches of industrial procurement—Arbizzi at Ferrari and Accenture, Gava in venture capital after stints at Infineon and consulting. What they saw wasn't a technology gap. It was a time trap.
Procurement teams at manufacturing companies, they realized, spend substantial portions of their time on operational drudgery: emailing suppliers for quotes, updating spreadsheets, confirming delivery dates, chasing paperwork. The strategic work—supplier relationship management, cost optimization, risk mitigation—gets squeezed into whatever time remains. Which is often none.
The company just closed a €3.2 million seed round led by Picus Capital, with Shapers and existing backers Italian Founders Fund and DFF Ventures joining. That brings total funding north of €5 million since founding, and the startup now claims to be managing over €10 billion in procurement spend across 40 customers in eight countries. For a company over two years old, the traction is notable—perhaps more than the founders expected when they started pitching in 2024.
What Makes This Different
The phrase "AI workforce" gets thrown around loosely these days, but Compri's architecture is specific. The platform deploys more than 30 specialized agents organized into five functional teams: Procurement Intelligence, Intake, Sourcing, Compliance, and Order Management. Each agent handles discrete tasks within its domain. One requests quotes. Another follows up three days later if suppliers haven't responded. Another validates vendor certifications against regulatory requirements.
Traditional procurement software—SAP Ariba, Coupa, Ivalua—provides workflows and dashboards. Someone still has to execute the work. Compri's system executes the work, generating audit trails and escalating to humans only when thresholds are breached or judgment calls are needed.
The technology runs on a combination of OpenAI, Anthropic, and Gemini models, fine-tuned on procurement-specific data and hosted on EU-based AWS infrastructure. GDPR compliance and preparation for the EU's emerging AI Act were baked in from the start—a necessity for enterprise customers in regulated industries. According to the company, implementations typically go live within weeks, a sharp contrast to the months-long deployment cycles common with legacy enterprise software.
Integration depth matters here. The system plugs directly into existing ERPs (SAP, Oracle, Microsoft Dynamics 365, NetSuite) and pulls external data from sources like Dun & Bradstreet for vendor screening, commodity indices like S&P and LME for pricing intelligence, and ESG ratings from EcoVadis. It connects to email systems, document repositories, even communication tools like Slack. The goal isn't to replace a company's tech stack but to operate within it—invisibly, ideally.
From Idea to €10 Billion Under Management
Compri announced a €1.6 million pre-seed round in early 2025, co-led by Italian Founders Fund and DFF Ventures, with angels including Giacinto Carullo (former chief procurement officer at Leonardo, the Italian defense contractor), Gianluca Cocco (CEO of fintech startup Qomodo), and Enrico Giacomelli (founder and chairman of digital identity firm Namirial). That capital funded the initial product build and early pilots.
Growth since then has been brisk. By mid-2026, the company had grown to more than 30 employees, according to a press release from that period, with team members drawn from Bain, BCG, Apple, Scalapay, and Ferrari. The customer base had expanded to 40 companies, spanning automotive, energy, machinery, transportation, packaging, food and beverage, and aerospace and defense. Forbes Italia reported the company's headcount and customer expansion around that time.
Geographic reach now extends beyond Italy into the Nordics, Germany, Netherlands, Spain, France, South America, and—perhaps most significantly—the industrial Midwest of the United States, with early customers in Ohio and Michigan. The US expansion is worth watching. America's industrial procurement market dwarfs Europe's, and if Compri can gain traction there, the growth trajectory could steepen considerably.
Leadership includes CTO Tommaso Latini and Chief Product Officer Luca Fornaroli, both part of the core team that built the agent architecture.
Does It Actually Work?
Customer results, at least as Compri presents them, suggest the automation delivers tangible outcomes. Movinter, a publicly identified client, reportedly saved approximately €360,000 annually and recovered more than 2,000 hours per year. An unnamed packaging company improved on-time delivery rates from 80% to 95% over six months.
Across its customer base, the startup claims operational task time reductions of up to 80% and cost savings exceeding 10% on managed spend. The €10 billion figure—assuming it's accurate—is roughly equivalent to the annual procurement budget of a mid-sized Fortune 500 manufacturer, distributed across 40 customers rather than concentrated in one.
Whether those savings hold up at scale, across more complex procurement scenarios and tighter vendor relationships, remains an open question. Procurement isn't just transactional. It's relational. Long-term supplier partnerships, negotiation nuance, and organizational trust don't always translate smoothly to algorithmic execution. Compri's bet is that the repetitive 80% can be automated reliably, freeing humans to focus on the strategic 20%. That's plausible. It's also not a guarantee.
The company has been building community presence in the procurement world, exhibiting at SPS Italia, a major manufacturing trade fair, and recording its "Compri Bene" podcast live at the IKN Italy Procurement Forum with senior procurement leaders from companies like STIGA and Dormakaba. That kind of visibility helps in a sector where trust and reputation matter as much as feature lists.
What Comes Next
Picus Capital's decision to lead the seed round signals confidence in both the product and the market opportunity. The €3.2 million will fund continued product development and European expansion, according to the company's announcement. But the real test will be scaling in the US, where competition is fiercer and customer expectations are higher.
The competitive landscape is messy. Legacy platforms like SAP Ariba, Ivalua, and Coupa still dominate enterprise procurement, but those systems were architected before large language models existed and well before anyone seriously talked about agentic AI. They're workflow engines, not digital workers. That's a conceptual gap Compri is trying to exploit.
Still, big incumbents move slowly until they don't. If Compri's approach proves out, expect rapid imitation—or acquisition interest. The window for independent growth in enterprise software isn't infinite.
For now, the startup's pitch is clean: procurement teams don't need another dashboard to manage. They need colleagues who do the work. Even if those colleagues happen to be algorithms running on servers in Frankfurt.
Whether that vision scales will depend less on AI capability—which is table stakes at this point—and more on execution, trust-building, and navigating the messy realities of organizational change. Procurement workflows are embedded in culture, vendor relationships, and decades of habit. Automating them is possible. Making people trust that automation enough to step back? That's the harder problem.
