Michael Polanyi, the British-Hungarian philosopher renowned for coining the term “tacit knowledge,” highlighted a significant aspect of expertise that remains largely unarticulated and undocumented. His assertion, “We know more than we can tell,” underscores the challenge facing companies today as they seek to integrate artificial intelligence (AI) into their operations. A significant portion of the knowledge required for effective AI solutions remains unwritten, creating a gap in workflow automation.
Interloom, a Munich-based startup, has recently secured $16.5 million in funding to tackle this very issue. The investment round, led by DN Capital with contributions from Bek Ventures and existing investor Air Street Capital, follows a $3 million seed round announced earlier in 2024. Although Interloom has not disclosed its valuation following this funding, the startup aims to revolutionize business process automation for the AI generation.
Fabian Jakobi, the founder and CEO of Interloom, emphasizes that the current enthusiasm for AI agents often neglects the bottleneck created by tacit knowledge. He points out that approximately 70% of operational decisions within organizations have yet to be formally documented. This knowledge, often gained through experience, enables seasoned professionals to navigate complex situations successfully, even when such information is absent from manuals or databases. Jakobi asserts, “The most important person at the bank is the person who knows whether the documentation is right or not.” This unrecognized expertise often has a direct impact on quality and operational efficiency.
Interloom’s solution involves the analysis of millions of operational records—ranging from support emails to service tickets—to create what it refers to as a “context graph.” This dynamic map serves to capture how problems are resolved within organizations, akin to how Google Maps adapts to real-time traffic conditions. By mapping out the routes taken by operational experts, Interloom aims to empower both AI agents and new employees with the insights necessary to succeed in complex environments.
Jakobi’s entrepreneurial background includes founding Boxplot, a venture sold to Hyperscience, which specializes in extracting data from unstructured documents. Interloom’s software is currently operational within major European corporations. For instance, at Commerzbank, it assessed the efficacy of existing internal documentation against millions of support emails, revealing inconsistencies and gaps. The company claims to have reduced this disparity from approximately 50% to a mere 5%. Interloom has also processed customer support tickets for Volkswagen and earned a company-wide AI competition win at Zurich Insurance, standing out among 2,000 competing startups.
Jakobi emphasizes that underwriting decisions in insurance firms reflect the unique risk appetites and institutional knowledge inherent to each organization, which general-purpose models may overlook. He criticizes traditional consulting firms for not possessing the nuanced understanding that an internal underwriter would have regarding their operations.
The ongoing dialogue within the industry stresses that AI agents need organization-specific context to be effective. Jakobi frames this issue as the “corporate memory” problem, noting that while compilers in software programs can verify code functionality, such assurances are absent in other domains, necessitating human expert evaluations.
Investors in Interloom share Jakobi’s perspective. Guy Ward Thomas of DN Capital acknowledges that the effectiveness of AI agents is heavily reliant on the expert decisions informing them. He has observed that agents lacking proper context frequently underperform. Mehmet Atici from Bek Ventures, who previously supported UiPath, aligns with this view, recognizing the transformative potential of AI in enterprise automation and its departure from traditional robotic process automation.
Interloom’s emergence comes at a critical time, coinciding with the “Great Retirement,” where approximately 10,000 Baby Boomers are leaving the workforce daily in the U.S. This mass exit risks the loss of valuable institutional knowledge, coinciding with companies’ efforts to deploy AI solutions on a larger scale.
Jakobi remains candid about the competitive landscape, identifying inertia within large organizations as his primary obstacle. As Interloom sets its sights on future product development, it plans to introduce a feature dubbed internally as “Chief of Staff,” aimed at providing managers with real-time visibility into AI agent performance while incorporating version control for these processes.
While Interloom isn’t the only player venturing into AI agent management, Jakobi is confident that his company’s unique “context graph” offers a competitive edge, particularly in providing the comprehensive insights necessary for navigating complex operational challenges—something he believes larger companies often fail to achieve.
