Marble Cognition raises $3.75M for its AI agent workers that watch and learn how people work

Munich-based startup Marble Cognition has secured €3.35 million ($3.75 million) in a pre-Seed funding round to develop AI coworkers that can learn and automate business processes. The investment was led by Heartcore and Seedcamp, with additional support from angels including a Datadog board member, a Personio founder, and executives from McKinsey, Mistral AI, and Lovable. The funds will be utilized for scaling inference, expanding the Munich engineering team, and preparing for a U.S. launch.

Reshma Sohoni, a General Partner at Seedcamp, expressed their support for Marble Cognition, stating that the startup "fills the gap" by providing vital context for AI transformations in businesses. Marble Cognition was founded by Felix Rösner and Christophe Schmit in late 2025, who met during Munich's CDTM programme. The company's approach involves learning how work is executed within a company, converting it into systems AI can understand, and deploying Marble Coworkers – AI agents that automate manual white-collar tasks.

Marble emphasizes that while the potential to automate work using existing technology is high, the real challenge lies in understanding the intricacies of each company's unique processes. McKinsey estimates that technology could theoretically automate 57% of US work hours, but in most companies, 90% of core processes are undocumented, residing in inboxes, spreadsheets, and employees' heads.

Marble addresses this by observing team workflows, transforming each process into a structured map that AI can comprehend, and then building Marble Coworkers that execute the process end-to-end within existing tools, with human oversight where necessary.

The startup prioritizes privacy and security, anonymizing all captured data before any person or AI model accesses it. Their desktop app records screen activity, clicks, and keystrokes within business applications, excluding private tabs and activities outside working hours. Employees can pause data collection at any time, and analyses occur at an aggregate level, not at an individual level.

Notable customers include EGYM, Mutares, and Enam, along with other tech unicorns, PE-backed organizations, and AI roll-ups supported by leading funds. Marble claims substantial success, with one PE-owned US financial services company reporting a 90% reduction in handling time per case and a 50% increase in case handling without additional hires. The process was learned by observing the team rather than relying on documentation.