Marble raises USD $3.75 million to automate office work

Marble Cognition has raised USD $3.75 million in a pre-seed funding round co-led by Heartcore and Seedcamp.

The Munich-based company is building AI coworkers that learn how tasks are carried out inside businesses by observing employees' work in the software tools they already use. It then uses that information to automate repetitive white-collar processes, while keeping humans involved where judgement is needed.

Customers include EGYM, Mutares and Enam, alongside unnamed technology unicorns, private equity-backed companies and AI roll-up groups in Europe and the US. In one example at a US financial services company owned by private equity investors, Marble said one of its AI coworkers reduced handling time per case by 90% and helped the team process 50% more cases in its first month without adding staff.

Marble argues that the main constraint on wider AI use in office work is not a lack of models or software tools, but the absence of a detailed record of how work is actually done inside companies. Many internal processes are not formally documented and instead are spread across email inboxes, spreadsheets and employee knowledge.

Its software is designed to build a map of those workflows by recording screens, clicks and keystrokes in business applications. The system excludes private apps, private browser tabs and activity outside working hours, and workers can pause recording at any time.

Marble added that data is anonymised before any person or AI model sees it, including when the system is deployed in a customer's own cloud environment. Analysis is carried out in aggregate rather than on an individual basis, according to the company.

Investors in the round also included angel backers described by Marble as a board member of Datadog, a founder of Personio and executives from McKinsey, Mistral AI and Lovable.

How it works

Marble said its system first observes how teams complete work, then converts that information into a structured process map that AI can interpret. It then deploys what it calls Marble Coworkers to execute those processes end to end in the same workplace software used by employees.

This differs from traditional automation software, which often requires companies to reshape their processes around a standard product. Marble's focus is on capturing and structuring the process itself before automating it.

At the financial services client cited by the company, the AI coworker prepares settlement offers for specialist review. Marble said it learned that process by watching the team's work rather than relying on written documentation.

"Every company we talk to wants AI to take over real work. But nobody can tell the AI how that work is actually done," said Felix Rösner, Co-founder of Marble. "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."

Investor view

The funding comes as investors continue to back companies trying to move AI from assistant tools into more direct task execution inside large organisations. Marble's pitch sits in a crowded automation market, but its emphasis is on gathering operational context from day-to-day employee behaviour before assigning work to software agents.

Reshma Sohoni, General Partner at Seedcamp, said: "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."

Max Niederhofer, Partner at Heartcore, said: "The scarce input for automation is not the models, which are commoditising. It is building a precise record of how work gets done. Marble learns a whole company's processes, anonymises the data, derives SOPs and deploys agentic AI against them. It is what Meta has been doing for the past year and now it is available for anyone. What Felix and Christophe are building is a key technology for the AI-enabled workplace."

Founders and plans

Marble was founded in late 2025 by Felix Rösner and Christophe Schmit, who met at CDTM. Rösner previously worked at 11x, an AI sales agent platform backed by Benchmark and Andreessen Horowitz. Schmit was a researcher at MIT and also worked at Enam, an AI startup backed by Elad Gil.

The company will use the new funding to expand its engineering team in Munich, increase inference capacity and support its move into the US market.