The Munich startup is creating a ‘context graph,’ a constantly updated map showing how operational decisions are made within an enterprise, based on millions of real cases rather than potentially unwritten documentation.
In every enterprise AI deployment, there’s a particular friction point often encountered early on. An AI agent may be technically proficient, capable of reading documentation, following instructions, and executing steps.
However, it cannot replicate the judgment of an employee with fifteen years of experience who knows why standard procedures fail at specific times, like Tuesdays in the logistics department. This experiential knowledge was never documented, as there was no need, until now.
Interloom aims to tackle this issue. The Munich-based startup, operating in Munich, Berlin, and London, announced on March 19 that it has raised $16.5 million in a seed round led by DN Capital, along with participation from Bek Ventures and existing investor Air Street Capital.
This round marks a significant increase from the company’s initial $3 million seed, led by Air Street in March 2024 when the company emerged from stealth mode.
Interloom’s main product is the Context Graph: a constantly evolving model of how decisions are made within an organization, created by analyzing millions of real cases, support emails, service tickets, call transcripts, and work orders to identify patterns in problem-solving by expert workers.
Founder and CEO Fabian Jakobi emphasizes the challenge of tacit knowledge, a concept introduced by British-Hungarian philosopher Michael Polanyi, who observed that most expertise cannot be fully articulated by the expert who possesses it. Jakobi estimates around 70% of operational decisions are never formally documented.
Jakobi compares it to Google Maps: just as the navigation tool learns optimal routes from real-time traffic, Interloom creates a map of the paths experts use to solve problems, then uses that map to guide AI agents and new employees in similar situations. The system updates continuously, so every resolved case contributes to institutional memory instead of disappearing when someone leaves or retires.
This retirement risk supports the pitch. The press release highlights the statistic of 10,000 baby boomers leaving the US workforce daily, as widely documented by Pew Research.
The issue enterprises face is losing decades of institutional knowledge at the very time AI is expected to automate complex tasks. Without capturing that knowledge first, AI lacks a useful foundation.
Interloom’s early clients include Zurich Insurance, JLL, Fiege, Commerzbank, and Volkswagen, the last two confirmed independently by Fortune in its exclusive on the funding. At Commerzbank, Interloom analyzed millions of customer support emails against internal documentation, reportedly closing the gap between documented procedures and actual processes from around 50% to 5%.
At Zurich Insurance, Interloom won an internal AI competition against 2,000 competing AI startups for an underwriting use case, as described by Jakobi to Fortune.
The investors have their own validating logic. Guy Ward Thomas of DN Capital, leading the investment, was the first institutional backer of Cognigy, a German enterprise conversational AI platform supported by DN Capital from its Series A in 2019 and acquired by NICE in 2025 for $955 million, noted as Europe’s largest AI exit at the time.
Ward Thomas learned the critical role of organization-specific context in making AI agents function effectively. Mehmet Atici of Bek Ventures, an early supporter of UiPath, argues this wave of AI agent adoption is the next major phase in enterprise automation following RPA.