Munich startup Marble Cognition has raised $3.75 million in pre-seed funding to build AI coworkers that learn how a company works before they take work over.
The funding round
Heartcore and Seedcamp co-led the round, which the Munich company announced on 2 November 2026 alongside its first product launch.
Angel investors in the round include a board member of Datadog, a founder of Personio, and executives from McKinsey, Mistral AI and Lovable.
Marble was founded in late 2025 by Felix Rösner and Christophe Schmit, who met on Munich’s Center for Digital Technology and Management programme.
How the coworkers learn
Marble starts from a simple problem, since most business processes are never written down and instead live in inboxes, spreadsheets and the heads of employees.
Its desktop application observes how employees work in business applications, recording screen activity, clicks and keystrokes. It turns those observations into structured process maps that AI systems can read and carry out like code.
The resulting Marble Coworkers then run processes end to end in the tools teams already use, with people staying in the loop where judgement or approval matters.
McKinsey estimates that existing technology could theoretically automate about 57 per cent of working hours in the United States. Marble argues that undocumented processes, not the technology itself, are the main blocker.
Privacy controls
Marble says everything it captures is anonymised before any person or AI model sees it, and data is analysed in aggregate rather than at an individual level.
The application does not monitor private applications or private browser tabs, and it excludes activity outside working hours. Employees can pause recording at any time, and processing can run inside a customer’s private cloud or on premises.
Early results and what comes next
Marble says its technology is already in use at organisations including EGYM, Mutares and Enam. At one private equity backed financial services firm in the United States, a Marble Coworker prepared settlement offers for human review.
That deployment cut case handling time by 90 per cent, Marble says, and the team handled 50 per cent more cases in the first month without new hires.
Felix Rösner, co-founder of Marble, described the approach in the company announcement with a simple summary of the problem it solves.
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.
Reshma Sohoni, partner at Seedcamp, said mapping the work is the step that companies usually miss before they automate.
The hard part has always been mapping out how the work gets done, so you can actually automate it.
Marble will use the new funding to scale its inference capacity, expand its engineering team in Munich and support its launch in the United States.
