Fabian Jakobi is the co-founder and CEO of Interloom. A German startup is using cutting-edge AI to automate business processes.
Large companies have invested heavily in recent years in software designed to automate routine back-office operations, often by cutting and pasting or dropping data from one software application to another. Simply fill in the database fields using the down menu.
This type of software “robot”, known as “Robotic Process Automation” or RPA, is not AI. Some macros are simply improved versions of Excel macros that record mouse movements and keyboard strokes. Others use “if-then” rules to allow the software to complete a workflow.
However, according to technology analysis firm Forrester Research, it is estimated that businesses now spend more than $6 billion annually on RPA software, and this number is growing at a double-digit rate. His UiPath, one of the leading companies in the RPA space, is valued at $13.5 billion. Appian, Blue Prism, and IBM also offer his RPA solutions.
Now Interloom, a small startup based in Munich, Germany, believes it can disrupt this entire market by re-architecting process automation on top of a new wave of large-scale language models and generative AI assistants. Air Street Capital, a London-based venture capital boutique specializing in early-stage AI investments, provided the company with his $3 million in seed funding to begin realizing its vision.
Fabian Jakobi, co-founder and managing director of Interloom, is a serial entrepreneur. In 2021, he sold his last company, Boxplot, to Hyperscience, a New York-based AI software company that specializes in extracting data from unstructured documents. Jacobi believes similar his AI techniques could be used in the future to extract information from video recordings, call logs, notes, etc., allowing AI software to understand how professionals actually do their jobs. I believe that you will be able to learn. Many parts of these tasks can then be automated using AI agents based on the same underlying AI methods that power today's large-scale language models.
This has the potential to automate much higher-value tasks than can be addressed with today's RPA. RPA is only useful for tasks with highly routine and repetitive workflows. According to Jakobi, his current RPA technology can automate about a third of his tasks in business. This is a limitation that helps explain why a report from consulting firms EY and Deloitte found that the vast majority of RPA projects fail completely or never reach their potential.
Rather than starting with an ideal workflow, Jacobi says companies can train AI software based on what they actually do in real-world situations. Rather than sticking to overly standardized and rigid templates, AI can intuitively determine the best workflow for a given situation.
He gives the example of drafting and sending a non-disclosure agreement as part of a commercial transaction. With today's RPA, a company could create a rule that any transaction over $100,000 requires the counterparty to send her NDA. The workflow of filling out her NDA template for this particular transaction is automatically handled by a software robot.
The problem with such rules, Jacobi says, is that they are too inflexible to capture real business logic. What happens to $98,500 worth of transactions? Most companies will want the robot to be covered by an NDA, even if it's below the standards they set for it. His latest LLM-based software excels at capturing tacit knowledge from historical data. Tacit knowledge includes a lot of professional judgment, such as when an NDA is necessary.
According to Interloom, tasks best suited for this type of automation include sourcing and risk assessment, customer onboarding, mortgage and insurance claims processing, and managing import and export logistics documentation.
Jacobi said humans will still be needed to perform quality control checks on Interroom's software robots, but in many processes, AI robots like the one they are building will help employees do it in a given time. We believe we can increase the amount of production we can produce. The time will be multiplied by 30.
Interroom, which currently employs only 10 people but is hiring rapidly, is targeting Germany's large Mittelstand, the medium-sized industrial companies that are the bulwarks of the German economy, as its first customer base. , with plans for global expansion including: I'll be heading to America soon.
“Every sector of the economy will eventually be reimagined to be AI-first,” said Nathan Benaich, founder and general partner at Air Street. “With experience in Boxplot and Hyperscience, the Interloom team has a uniquely deep understanding of automating complex workflows. This is why we are building his infrastructure to support processes that support AI-powered business productivity gains. We are in the perfect position to do so.”