Courtesy of Octonomy
Many AI assistants run into problems when they face complex or unstructured data, like the kind that underpins industries like manufacturing, insurance and pharmaceuticals.
While they might be fine with simple FAQs or text-based queries, processing more complicated data results in more hallucinations and less accuracy.
That’s the problem German startup Octonomy, which has just raised $20m in seed funding, is trying to solve, with its more generic solution for automating complex, knowledge-intensive enterprise workflows.
The round was led by Macquarie Capital Venture Capital, with participation from Capnamic, NRW.Bank, and theTechVision Fund, and brings its total funding up to $25m since it was founded 15 months ago.
“I think our perfect ICP [Ideal Customer Profile] is the company that has just failed,” Octonomy co-founder and CEO Sushel Bijganath tells Pathfounders. “They have tried with whatever tool you can think of at the moment in the market, and for the simple use cases it works actually pretty well. The moment it gets a little bit more complex, people say, ‘Yeah, we hit our boundaries.’”
Now based across Germany, India and, as of August, the US, the startup raised $5m seed from Capnamic when it first started, though this wasn’t publicly announced until February this year when the company officially launched.
Bijganath co-founded the company alongside Thorsten Grote, Markus Hanslik, and Thomas Bollig, and the 70-person team includes veterans from Meta, Amazon, Aleph Alpha, Personio, Staffbase, and SoSafe.
“The biggest problem we are solving with Octonomy is to avoid hallucination on complex documentation,” Bijganath says. “You can try to do this with OpenAI, with Anthropic, with any kind of AI solution that already is a unicorn in the US and they will all get, let’s say on the documents that we are talking about, maybe get 50% if they’re good and we get to 95%. And that makes a big difference.”
It typically takes five to 20 days to onboard a new client, and the goal is to provide an expert system to troubleshoot where there might not be the manpower. Industries with highly technical products such as manufacturing, where there might be documents that are tens of thousands of pages long, are worst hit by AI hallucination.
So while Octonomy’s product has so far been quite broad, as it grows it will likely focus on machinery and manufacturing and companies with particularly complex products and services.
Particularly with aging workforces in these industries and a younger demographic entering the workforce that expects to use AI tools while they work, this is especially needed.
“Industry has not made use of the potential of modern AI technology,” Capnamic founding partner Jörg Binnenbrücker says. “[Octonomy] translates expert knowledge into scalable, operational intelligence and leverage experience for productivity. These are exactly the kinds of technologies that industry needs to transfer AI from the research stage to value creation.”