Marble Cognition raises $3.75M to map and automate corporate workflows

Bridging the automation gap

While artificial intelligence agents are becoming common in the corporate world, many organizations struggle to deploy them effectively. Most companies currently use AI primarily for writing software, but they lack the specific operational context needed to automate broader business tasks.

Marble Cognition, a Munich-based startup, aims to solve this by providing AI with the granular details of how work actually happens. The company recently raised $3.75 million in pre-seed funding to launch its platform, which focuses on mapping out workflows that do not follow neat blueprints.

The startup was co-founded by German AI engineers Felix Rösner and Christophe Schmit. They identified a major hurdle: over 90% of business processes are hidden in messy email threads, scattered spreadsheets, and the minds of employees, making them difficult for standard AI to grasp.

How Coworkers learn the job

The company's solution involves a dedicated desktop application designed to monitor how employees interact with different tools. These agents, which the company calls 'Coworkers,' watch the work happen by capturing screenshots, clicks, and keystrokes from the user's screen.

The system then transforms this raw activity data into a structured process map. This map acts as a set of instructions that the AI agents can read and execute, much like a computer program or code. It allows the AI to learn the specific steps of a job by observing a human expert.

Once the workflow is mapped, the Coworkers take over the most repetitive parts of the process end-to-end. Humans remain in the loop to make important decisions, but the AI handles the routine execution across the organization's existing software stack.

Privacy and data security

To address concerns regarding employee privacy, Marble Cognition has built several safeguards into its data collection process. All information recorded by the desktop application is anonymized before it is used for any purpose.

The company also allows for a review process where employees or managers can check the data before it touches any AI models. Furthermore, all model processing can be hosted within a customer's private cloud or on-premises environment to ensure data remains secure.

The software is also designed to be selective about what it records. It can identify non-business-related computer use, such as browsing social media during a lunch break, and will exclude that data from the collection process.

Real-world efficiency gains

Early adopters of the technology have already reported significant improvements in their operational speed. For example, a private equity-backed U.S. financial services firm used the agents to create settlement offers for human review.

This specific application allowed the firm to slash its case-handling times by 90%. Because of this efficiency, the company's human staffers were able to process around 50% more claims without the need to hire additional staff.

Other early customers include EGYM GmbH, Mutares SE & Co. KGaA, and Enam GmbH. These partnerships demonstrate how mapping out work can lead to measurable productivity gains in complex industries.

The future of AI context

The funding round saw participation from several notable investors, including executives from McKinsey & Co., Mistral AI, and a board member of Datadog Inc. Seedcamp General Partner Reshma Sohoni, who previously backed UiPath, highlighted the importance of this technology.

Sohoni noted that the hardest part of automation has always been mapping out how work gets done so that it can actually be automated. Marble Cognition aims to fill this gap by providing the vital context required for large-scale AI transformations.

By turning human actions into structured data, the startup hopes to help companies move beyond simple chatbots. The goal is to enable autonomous workers that can handle real work by following the same patterns as a company's best employees.