Marble Receives $3.75 Million

The Munich-based startup Marble Cognition has raised $3.75 million in a pre-seed funding round. The round was led by Heartcore Capital and Seedcamp. Marble develops AI agents that first observe business workflows and can then take them over autonomously. Its clients include EGYM, Mutares, and Enam. The company plans to use the capital to expand its team in Munich and enter the U.S. market.

The idea is simple: many companies want to automate their processes with AI but often don’t know exactly how their day-to-day work actually unfolds. Much of this knowledge resides in employees’ minds, in emails, or in Excel spreadsheets. Marble aims to capture this knowledge and make it usable for automation.

Heartcore and Seedcamp Invest in Marble

With Heartcore Capital and Seedcamp, the startup—founded in late 2025—has secured two European early-stage investors.

Heartcore has backed companies such as GetYourGuide, Tink, and FINN. Seedcamp’s portfolio includes companies like Revolut, Wise, UiPath, and Synthesia. Business angels from the Datadog, Personio, McKinsey, Mistral AI, and Lovable ecosystems are also participating in the funding round. These include a member of Datadog’s board of directors, a Personio founder, and executives from the aforementioned companies. The investors see a competitive advantage primarily in understanding internal business processes. While AI agents are becoming increasingly easy to develop, the question remains as to which tasks can be meaningfully automated.

AI agents learn directly from day-to-day work

To achieve this, Marble relies on a desktop application that tracks how employees work with business software. It records screen activity, mouse clicks, and keystrokes and uses this data to create structured process models. The software then identifies recurring tasks and develops so-called “Marble Coworkers” that take over these activities within the existing applications.

According to Marble’s observations, approximately 90 percent of its clients’ core work processes are not fully documented in writing. This complicates the use of conventional automation solutions. The AI agents are therefore designed to first understand how tasks are actually performed. Employees remain involved in decisions that require human judgment.

90 Percent Less Processing Time for One Client

An example from the U.S. financial services industry illustrates how this works in practice. At a private-equity-backed company, a Marble Coworker handles the preparation of comparative offers, which are then approved by a specialist. According to the startup, this reduced the processing time per case by 90 percent. In the first month, the team was able to process 50 percent more cases without hiring additional staff.

The software had previously learned the process by observing the employees. It remains to be seen whether such results can also be achieved at other companies. However, Marble already lists EGYM, Mutares, and Enam as clients and, according to its own statements, is collaborating with other technology companies and private-equity-backed firms.

Founders Bring Experience from the AI Industry

Marble was founded in late 2025 by Felix Rösner and Christophe Schmit. The two met at the Center for Digital Technology and Management (CDTM). Rösner previously worked at 11x, a provider of AI-powered sales agents. Schmit worked as a researcher at the Massachusetts Institute of Technology (MIT) and subsequently gained experience at the AI startup Enam. Both had been exploring the use of AI agents in businesses even before founding the company.

Every company we speak with wants AI to take over real work. But no one can explain to the AI exactly how that work is actually done.

Felix Rösner, Founder

Marble aims to solve this problem by having the software first learn the workflows on its own.

Expanding into the U.S. with Fresh Capital

The funding is primarily intended to go toward further developing the technology. Plans include hiring additional staff for the Munich development team and expanding computing capacity. Marble is also preparing to enter the U.S. market.

This could open up a growing market for Marble. However, the key will be how reliably the AI agents perform and how quickly the technology can be deployed for new customers. Marble is banking on the idea that, in the future, it won’t be the development of individual AI agents that creates the most value, but rather the knowledge of how companies actually operate. The young startup must now prove whether this approach will succeed on a larger scale.