Munich-based software company Marble Cognition has raised $3.75 million in a pre-seed funding round to develop autonomous agents that map operational workflows and automate corporate tasks. Venture capital firms Heartcore and Seedcamp co-led the investment, backed by prominent European technology executives and angel investors.
Key points
- Marble Cognition closed a $3.75 million pre-seed round co-led by Heartcore and Seedcamp.
- The platform uses a desktop application to record screen activity and clicks, turning unwritten business habits into structured process maps.
- The startup plans to expand its Munich engineering team and prepare for entry into the United States.
- Software deployments are underway at German firms including EGYM, Mutares and Enam.
The problem with unwritten business processes
Enterprises have spent two years testing generative software, yet routine operations remain manual. Much of that friction stems from missing documentation. Crucial procedural details rarely live in software manuals. Instead, staff pass instructions through email chains, shared spreadsheets and unrecorded institutional memory.
Consultancy McKinsey calculated that modern digital systems could automate roughly 57 per cent of working hours across the United States. Marble found a practical barrier across its early customer base: approximately 90 per cent of core operational processes have never been formally documented.
Without clean workflow instructions, standard language models cannot reliably execute administrative tasks. Enterprise operators find themselves forced to interview staff and write software manuals manually before configuring automation pipelines. Marble aims to bypass that manual discovery stage entirely.
How Marble Coworkers learn on the job
Founded in late 2025 by Felix Rösner and Christophe Schmit, the German startup builds desktop software that watches staff work across their regular business applications. The system monitors screen activity, mouse clicks and keystrokes while employees handle routine duties.
The software aggregates these interactions and converts them into structured workflow maps that machine learning models can understand. Once the platform recognises the path of a business process, it configures autonomous software agents, named Marble Coworkers, to replicate the sequence.
“Our Coworkers watch how the work happens before they take it over,” said Felix Rösner, co-founder of Marble. The agents interact directly with existing corporate tools, reserving flagged steps for human approval whenever an operational judgement call is required.
Similar efforts across enterprise automation have drawn heavy funding over the past year. In the United States, Serval raised $150 million to expand its enterprise AI agents, while teams such as Crux Labs raised $6m to capture human judgment in operational software pipelines.
Data privacy and desktop monitoring safeguards
Continuous desktop recording often triggers pushback from staff and trade unions, particularly in Germany’s strictly regulated enterprise market. Marble designed its capture tool around privacy boundaries to reassure compliance teams.
The desktop agent strips personal information from captured sessions before human reviewers or machine models see the logs. The company analyses activity in aggregate rather than tracking individual worker performance, preventing managers from using the product as surveillance software.
Employees retain manual control over the tracking process. Staff can pause screen recording at any point during their shifts. The software ignores private browser tabs, personal applications and digital activity taking place outside scheduled working hours.
Backers and transatlantic expansion
The $3.75 million (€3.35 million) pre-seed round drew backing from prominent European venture houses and operator angels. Venture capital firms Heartcore and Seedcamp co-led the transaction.
Angel investors joined the syndicate, including a board member from cloud observability firm Datadog and a founder of German HR platform Personio. Executives from McKinsey, French AI lab Mistral AI and web design engine Lovable also put personal capital into the round.
Several commercial teams already run the software. Fitness technology provider EGYM, private equity investment firm Mutares and manufacturing company Enam have deployed Marble to review internal workflows, alongside other private equity portfolio businesses.
Marble will spend the fresh capital on scaling inference capacity for its foundation model pipelines and hiring software engineers at its Munich headquarters. The founders also intend to use the funding to establish a commercial presence in the United States.
