Munich's Marble Cognition raises €3.35M pre-seed for AI coworkers

Munich-based Marble Cognition has raised €3.35 million ($3.75 million) in a pre-seed round to build AI coworkers that learn how a company's work actually gets done before automating it. The round, co-led by Heartcore and Seedcamp, included angels with ties to Datadog, Personio, McKinsey, Mistral AI and Lovable.

The company will spend the money on scaling inference, growing its Munich engineering team and launching in the US. That is an ambitious list for a pre-seed cheque of roughly $3.75 million; inference costs alone can consume such budgets quickly.

Who is behind Marble?

Felix Rösner and Christophe Schmit founded Marble in late 2025 after meeting at Munich's CDTM programme. The startup says its system observes how teams work, converts every process into a structured map that AI can read like code, and then deploys "Marble Coworkers" — AI agents that take over manual white-collar work end to end, with humans kept in the loop where judgement matters.

The founders' core claim is contrarian to the current agent-building boom: with software and AI agents both becoming commodities, the real value lies in knowing exactly what to automate. That is hard, the company notes, because every company runs on processes nobody has written down.

What is the problem Marble says it solves?

Marble cites a striking gap between automation theory and practice. "McKinsey estimates that today's technology could, in theory, automate about 57% of US work hours. Yet in almost every company Marble works with, around 90% of core processes live only in inboxes, spreadsheets and the heads of the people doing them. That, not the technology, is the biggest blocker to AI transformation. What is missing is an exact picture of how the work is done," the startup said in its press release.

The claim that 90% of core processes are undocumented comes from Marble itself, not an independent audit. It is a useful framing for the company's pitch, and readers should treat it as such.

How does the technology actually work?

Marble argues that traditional automation bends a standard product to fit a company's processes, whereas it productises the understanding of the process itself. The mechanics raise obvious privacy questions, which the company addresses directly:

• Marble anonymises everything it captures before any person or AI model sees it, even inside the customer's own cloud.
• Its desktop app records screen activity, clicks and keystrokes in business apps only — never private apps, private tabs or anything outside working hours.
• Employees can pause recording at any time.
• Data is analysed only in aggregate, not based on individuals.

These are the company's own assurances; no third-party security assessment is cited in the announcement.

"Every company we talk to wants AI to take over real work. But nobody can tell the AI how that work is actually done. So we solve that first. Our Coworkers watch how the work happens before they take it over. They do it the way your best people do," said Rösner.

Who is using it, and what results are claimed?

Named customers include EGYM, Mutares and Enam, alongside tech unicorns, PE-backed organisations and AI roll-ups backed by top-tier funds — a roster the company did not fully disclose.

Marble reported one concrete result: at a PE-owned US financial services company, a Marble coworker cut handling time per case by 90%. In the first month, the team handled 50% more cases without hiring anyone new. The coworker prepares every settlement offer for a specialist to approve, and Marble learned the process by watching the team rather than from documentation.

A single named-vendor case study is thin evidence, but a 90% reduction in handling time — roughly a tenfold speed-up per case — is the kind of number that will attract enterprise scrutiny.

Why did the investors write the cheque?

Reshma Sohoni, General Partner at Seedcamp, framed the investment as a continuation of a pattern. "Seedcamp has a long tradition of backing founders who change how we work, from UiPath and Synthesia to Marble today. Most companies want to automate more, but the hard part has always been mapping out how the work gets done. Marble fills that gap, giving companies vital context for AI transformations that could define their future. We're proud to back them from the start," Sohoni said.

The comparison to UiPath, a robotic process automation giant, sets a high bar for a company founded only months ago. If Marble's process-mapping approach holds up at scale, the US launch will be its first serious test.