Artificial intelligence is becoming increasingly capable of completing workplace tasks, but teaching it how a company actually operates remains a major challenge. Many organisations rely on processes that exist only in employees’ daily routines, emails, spreadsheets and institutional knowledge. Munich based AI startup Marble Cognition wants to bridge this gap by allowing AI systems to learn how work gets done before taking over repetitive tasks. The company has raised $3.75 million in pre seed funding to expand its technology and bring its AI coworkers to more businesses.
Heartcore and Seedcamp Lead Investment
Founded in late 2025 by Felix Rösner and Christophe Schmit, Marble, also known as Marble Cognition, has secured $3.75 million in a pre seed funding round co led by Heartcore and Seedcamp.
The round also attracted angel investors, including a board member of Datadog, a founder of Personio and executives from McKinsey, Mistral AI and Lovable.
The company develops AI powered software designed to understand internal business processes and convert them into structured workflows that artificial intelligence agents can execute.
The new investment will support the expansion of Marble’s AI inference capabilities, increase its engineering team in Munich and help prepare the company for its launch in the United States.
The Problem With Undocumented Business Processes
Companies are increasingly exploring AI automation to improve productivity, reduce repetitive work and manage operational costs. However, implementing automation is not simply a matter of introducing an AI tool into an organisation.
Important operational knowledge is frequently scattered across different software applications and employees. Many workers understand how to complete their responsibilities through experience, but the exact sequence of actions, decisions and exceptions is rarely documented.
Marble highlights this as a major obstacle to enterprise AI adoption. According to the company, approximately 90% of core business processes at organisations it works with are not formally documented.
The company also points to McKinsey research estimating that existing technology could theoretically automate around 57% of working hours in the United States.
AI That Learns Before It Automates
Marble addresses this challenge through a desktop application that observes how employees complete everyday tasks across business software.
The technology captures workflow information, including screen activity, clicks and keystrokes within workplace applications. It then analyses these interactions and converts recurring activities into structured process maps that AI systems can understand.
Using this information, Marble develops AI agents called Marble Coworkers. These agents can perform business processes using the software and tools employees already work with, rather than requiring organisations to replace their existing systems.
The platform is designed to handle repetitive activities while keeping employees involved whenever a task requires professional judgement, approval or human decision making.
Felix Rösner, co founder and CEO of Marble, said companies want AI to take over real work, but often cannot explain precisely how that work is performed. He explained that Marble’s Coworkers first observe existing workflows before attempting to carry them out.
Privacy and Employee Control
Since the platform analyses workplace activity, privacy and data protection form an important part of its approach.
Marble says information is anonymised before it becomes accessible to employees or AI models. Its application does not monitor private applications, private browser tabs or activities outside working hours.
Employees can also pause monitoring whenever necessary. The company says its analysis focuses on aggregated workflow information rather than evaluating individual employee performance.
These safeguards are intended to help organisations understand their operational processes while maintaining boundaries around employee privacy.
Established Customers and US Expansion
Marble reports that its technology is already being used by organisations including EGYM, Mutares and Enam, alongside technology companies and private equity backed businesses.
The company aims to build on these early deployments as it expands its engineering capabilities and prepares to enter the US market.
With its $3.75 million investment, Marble is betting that understanding how businesses work will become an essential foundation for successful AI automation. Rather than simply building more intelligent agents, the startup wants to give them the operational knowledge required to perform meaningful work inside organisations.
