Dutch AI model Metis takes on US giants in automated decisions

Dutch AI company GreenPT has released Metis, a decision-making AI model that it positions as a European answer to American offerings such as Jev, from US firm TypeSafe AI. The launch comes as European businesses look for AI tools that keep their data inside the EU.

Metis works differently from the chatbots most people know. Instead of writing paragraphs of text, it answers structured questions — yes or no, multiple choice, or a rating on a scale — and returns a confidence score for every possible answer. Companies can then let the system act automatically when it is confident and send borderline cases to a human.

Addressing daily choices

GreenPT says the approach suits the many routine choices organisations make every day, such as sorting incoming requests, flagging risks or setting priorities. Because the output is already structured, it can plug straight into existing software without anyone having to interpret free-form text.

The company gave a customer-support example. Presented with a ticket about failed payments, Metis judged in a single pass that a live service was probably down (95% confidence), that the technical team should take it (93%), and that it needed immediate attention (96%). According to GreenPT, the whole assessment took 72 milliseconds and used 0.0047 watt-hours of electricity.

"Metis gives you probabilities you can act on immediately for clear cases," said Robert Keus, GreenPT's co-founder and CEO.

Faster and lighter

GreenPT credits the speed to Metis's compact design. Rather than producing an answer word by word, as text-generating models do, it reads the input once and calculates the probabilities in one step. The company says this makes Metis up to five times faster than comparable chat-based systems, with typical response times below 100 milliseconds and up to 32 questions handled at once. These figures come from GreenPT and have not been independently verified.

Sustainability is central to the pitch. Every response through GreenPT's API includes a report of the energy used, so customers can track the footprint of each request.

Data stays in Europe

Metis runs only from GreenPT's EU endpoint, in European data centres powered by renewable energy. The company says it does not use customer data to train its models — a point aimed at organisations wary of sending sensitive information to US-based providers.

GreenPT, founded in 2025 and led by Keus and co-founder and CTO Cas Burggraaf, sees a large market. It points to TypeSafe AI, which reportedly reached an annual revenue run rate of $100 million within a week of launching Jev, as evidence of strong demand for this kind of tool.

Whether European buyers will choose a home-grown option over established US players remains to be seen. But with data sovereignty and AI energy use climbing the agenda in Brussels and boardrooms alike, GreenPT is betting that Metis arrives at the right moment.