Skip to content
Odyssey Wire OW markOWOdysseyWire

Tracing AI through its unexpected turns.

Marble Cognition raises $3.75M for agents that record keystrokes and clicks
AI ventures

Marble Cognition raises $3.75M for agents that record keystrokes and clicks

Munich startup secures pre-seed funding to launch desktop apps that map workflows by monitoring employee actions for AI automation.

Depth
$3.75 millionMarble Cognition raised $3.75 million in pre-seed funding · reported
1 outlet · newest source 317h ago

Key points

  • Marble Cognition raised $3.75 million in pre-seed funding.
  • The startup’s desktop application captures keystrokes and screenshots.
  • Co-founders argue AI agents need granular operational context to automate work effectively.

SiliconANGLE reported that Marble Cognition raised $3.75 million in pre-seed funding to launch its platform. The Munich-based startup aims to help enterprises understand how to use AI agents by mapping workflows through direct observation of employee behavior.Marble Cognition raised money to build software that watches how employees work so AI can learn to do the tasks.Marble Cognition’s pre-seed round funds a desktop agent that ingests keystroke and screenshot telemetry to generate executable process graphs.

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

aifleet-siliconangle

The claim

Marble Cognition argues that AI agents are becoming commoditized but lack the specific operational context needed for effective automation. According to SiliconANGLE, co-founder Felix Rösner stated that most organizations currently use AI primarily for writing software. He believes real value comes from providing agents with granular details about what needs to be built.

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

aifleet-siliconangle

The response

The startup addresses the difficulty of mapping messy business processes that live in emails and spreadsheets. SiliconANGLE noted that while McKinsey suggests 57% of US work hours could theoretically be automated, most companies have not achieved this. Marble’s approach is to capture these scattered processes directly from the source.

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

aifleet-siliconangle

What the evidence means

The material indicates that more than 90% of business processes exist in unstructured formats or employee minds. Marble Cognition seeks to transform this data into structured process maps that AI agents can read and execute. The funding round was led by Heartcore and Seedcamp, with participation from angel investors including executives from McKinsey, Mistral AI and Datadog.

In practice

Users install a dedicated desktop application that monitors how employees work across different tools. The software captures screenshots, clicks and keystrokes to create structured maps. These maps allow AI agents to understand and execute tasks. The startup’s "Coworkers" watch work happen before taking it over, mimicking the methods of top performers. Access requires installing this monitoring software on employee devices.

Why it matters

  • Enterprise IT buyers must evaluate desktop applications that record keystrokes and screenshots for compliance and privacy implications.
  • Organizations seeking to automate 57% of work hours, as suggested by McKinsey, need a method to map unstructured processes.
  • The involvement of investors from Mistral AI and Datadog signals industry interest in workflow automation tools.
  • Current AI agent usage is largely limited to software writing; this tool targets broader operational automation.

Topics: AI agents, workflow automation, enterprise software, funding

Updated

Then / now / next

Then

AI agents were primarily used for writing software without deep operational context.

Now

Marble Cognition launches a tool to map workflows by monitoring employee actions.

Next

Enterprises will need to assess the feasibility of installing monitoring software to enable agent automation.

Story timeline

  1. Then AI agents were primarily used for writing software without deep operational context.
  2. 02 Nov 2026 Marble Cognition raises $3.75M for its AI agent workers that watch and learn how people work
  3. Now Marble Cognition launches a tool to map workflows by monitoring employee actions.
  4. Next Enterprises will need to assess the feasibility of installing monitoring software to enable agent automation.

The strongest doubt

Counterpoint from the sources

SiliconANGLE noted that building a picture of corporate workflows is notoriously difficult because few processes follow neat blueprints.

Ask the article

Local keyword search of this article; no generated answers.

Did this article give you the evidence you needed?

No vote saved on this device.
Your choice stays on this device. No public tally.

Questions readers ask

How does Marble Cognition capture workflow data?

Its desktop application captures screenshots, clicks and keystrokes to create structured process maps.

Who led the funding round for Marble Cognition?

Heartcore and Seedcamp led the $3.75 million pre-seed round.

What problem does the startup aim to solve?

It aims to provide AI agents with the granular operational context needed to automate messy business processes.

Connected stories

This storyAdobe Max: Local MCP lets third-party AI edit native Creative Cloud files in 2027 betaAdobe Max: Local MCP lets third-Lumbra secures $25M Series A for AI agent orchestration in secure intelligence environmentsLumbra secures $25M Series A forUber discloses minimal AI agent usage in Q3 2026 earnings dataUber discloses minimal AI agent Survey shows half of US adults use AI weekly, but agents remain nicheSurvey shows half of US adults u

Reading receipt

  • Marble Cognition raised $3.75 million in pre-seed funding.
  • The startup’s desktop application captures keystrokes and screenshots.
  • Co-founders argue AI agents need granular operational context to automate work effectively.

Still open: How will enterprises handle the privacy implications of recording employee keystrokes and screenshots?

Still open: What specific industries or roles will be the first to adopt this workflow mapping technology?

Time spent: 0:00 · Depth: Full

Sources

  1. SiliconANGLE AI

Related stories

Background reading