Munich-based startup Interloom has raised $16.5 million in seed funding to build what it calls "knowledge infrastructure" for AI agents - technology designed to capture the undocumented operational expertise that companies need to automate but have never written down.
The round was led by DN Capital, with participation from Bek Ventures and existing backer Air Street Capital. The company operates across Munich, Berlin, and London.
The Tacit Knowledge Problem
About 70% of operational decisions in most organizations have never been formally documented. This is the tacit knowledge problem: the experienced employee who knows exactly why the standard playbook does not work on Tuesdays in the logistics department has never written that knowledge down, because nobody needed them to.
Until now. AI agents need explicit instructions to operate. They cannot replicate the judgment of someone who has been doing a job for fifteen years and learned through thousands of edge cases that no manual covers.
"AI agents cannot replicate the judgment of the person who has been doing the job for fifteen years," Interloom's founders told Fortune. "That knowledge has never been written down, because nobody needed to write it down."
This gap between documented processes and actual operational reality represents one of the biggest barriers to AI agent deployment in enterprises.
How Interloom Works
Interloom's approach is to ingest millions of operational records - support emails, service tickets, call transcripts, work orders - and use them to build what it calls a "context graph."
The analogy the company uses is Google Maps: just as Google Maps learns optimal routes from real-time traffic data, Interloom builds a map of the paths operational experts actually take to solve problems. It then uses that map to guide AI agents and new employees facing similar situations.
The process works in three stages:
- Data ingestion - pulling in operational records from existing systems (CRM, ticketing, email, knowledge bases)
- Pattern extraction - identifying how problems actually get resolved, including deviations from official procedures
- Context graph construction - building a continuously updated map of operational reality that AI agents can reference
Enterprise Traction
Interloom's early customer base includes several large European enterprises:
| Customer | Use Case | Result |
|---|---|---|
| Commerzbank | Customer support email analysis | Reduced documentation-reality gap from 50% to 5% |
| Zurich Insurance | Claims processing knowledge capture | Active deployment |
| JLL | Property management operations | Active deployment |
| Fiege (logistics) | Warehouse operations knowledge | Active deployment |
| Volkswagen | Manufacturing process expertise | Active deployment |
The Commerzbank result is particularly striking. The bank analyzed millions of customer support emails against its internal documentation and found that roughly half of how work was actually being done was not reflected in official procedures. Interloom reduced that gap to 5%.
Why This Matters for Business Automation
Interloom addresses a problem that has plagued automation efforts for decades: the difference between how organizations say they work and how they actually work.
Traditional business process automation (BPA) and robotic process automation (RPA) rely on documented workflows. When those documents are incomplete or inaccurate - which they usually are - automation breaks. Processes fail on edge cases that experienced employees would have handled intuitively.
The RPA market, valued at approximately $13 billion in 2026, is Interloom's primary capture opportunity. But the startup is positioning itself as infrastructure for the broader AI agent ecosystem, not as a replacement for RPA specifically.
The Knowledge Transfer Challenge
Interloom's technology addresses a challenge that every organization faces when experienced employees leave, retire, or change roles: institutional knowledge walks out the door.
This problem is particularly acute in:
- Customer support - where veteran agents know shortcuts and solutions that are not in any manual
- Insurance claims - where adjuster expertise determines which claims get fast-tracked and which need investigation
- Manufacturing - where operators know the machine quirks that keep production running smoothly
- Professional services - where partner and manager expertise drives client outcomes
For companies that use virtual assistants for operational support, the tacit knowledge problem is directly relevant. VA onboarding has improved - average onboarding time has dropped to 9 days - but much of the knowledge that makes VAs effective still transfers informally through experience.
Implications for Virtual Assistant Services
Interloom's technology has direct applications for the virtual assistant industry:
Faster onboarding. Context graphs could dramatically reduce the time it takes for new virtual assistants to become productive. Instead of learning through trial and error, VAs could reference operational maps that show how similar tasks have been handled successfully.
Quality consistency. One of the persistent challenges in VA services is maintaining consistent quality across different assistants. Knowledge infrastructure that captures best practices automatically would help standardize service delivery.
AI-human collaboration. As VA providers increasingly integrate AI tools into their workflows, tools like Interloom could help bridge the gap between what AI agents know and what human experts know - creating more effective hybrid workflows.
The $16.5 million bet on Interloom is a bet that the future of business automation depends not on better algorithms, but on better access to the knowledge those algorithms need to work correctly.