Marble raises $3.75m for AI coworkers
Marble Cognition raised $3.75m to build AI coworkers that learn internal business processes by observing how employees perform their work.
Before a company can automate a process, it needs to understand how that process actually works. Marble Cognition is starting with the gap between formal documentation and everyday practice.
What happened
Marble raised a $3.75m pre-seed round co-led by Heartcore and Seedcamp. The funding will support inference capacity, growth of its engineering team in Munich and a US launch.
The company develops software that observes workflows across business applications, converts that activity into process maps and uses the resulting context to build AI coworkers. Its intended approach keeps people involved in work requiring judgement or approval.
Why it matters
Important details often sit in emails, spreadsheets and the experience of employees. A written procedure may omit exceptions, missing information or the small decisions needed to complete a task.
Learning from actual workflows could provide more useful context for automation. The challenge is deciding which observed actions represent a good process and which reflect workarounds or avoidable errors.
The bigger picture
Process discovery and task execution are becoming more closely connected in enterprise AI. A product that understands how work moves across applications could help companies choose what to automate and assess what changes afterwards.
Observation also involves sensitive workplace activity, making access controls and clear boundaries important product requirements. Marble’s progress will depend on useful, reliable automation within those boundaries, rather than on collecting as much employee activity as possible.