Meet Metis: a decision model that weighs every answer

Some decisions are easy. Most of the ones your business makes every day are not. Can this order go through? Which team takes this ticket? How risky is this case?

Metis is GreenPT’s decision model. You give it a case and your rules, then ask your questions. Instead of writing an essay, it weighs every possible answer and tells you how sure it is.

Watch the introduction

Watch the 45-second introduction, with English narration and captions in English, Dutch and German.

Transcript: Some decisions are easy. Your business makes thousands of them, every day. And not all of them are easy. Can this order go through? What should we do next? How risky is it? Meet Metis, from GreenPT. Give Metis the case and your rules, then ask your questions. It doesn’t write an essay. It weighs every possible answer, and tells you how sure it is. Confident? Act on it. Unsure? Pass it to a person. Metis is a small model, so it’s fast, and light on energy. Metis. Wise advice, for every decision. One of the few decision models hosted in the EU, on sustainable infrastructure.

What is a decision model?

A decision model is an AI model that answers a question with a fixed set of possible answers. You describe a case and list the options; the model writes nothing at all. It reads everything once and returns a probability for every option, so you see which answer it favours and how sure it is.

A chat model works differently. It takes your question and writes an answer, token by token, which is what you want for a draft or an explanation. But a large share of business work needs an answer to a question with a fixed set of options, and the output you want is a set of numbers you can act on, not a paragraph someone has to read.

That sounds almost too simple. It is also what makes it so useful:

• The answer is fast. Writing text takes time. Weighing a fixed set of options takes a single pass.
• The answer is structured. You get probabilities your own systems can use directly, with nothing to parse.
• The answer tells you how sure it is. Every answer comes with its own probabilities, so you can decide when to trust it.

Three kinds of question

Metis answers three kinds of question about a case:

• Yes or no: is a rule met, is something true?
• One option from a set: which team, which category, which next step?
• A level on a scale: how urgent, how risky, how severe?

Here is a real example from the Metis documentation. A customer writes in: since 09:00, every checkout on their web shop fails, and their status page shows the payment service as down. We asked Metis three questions about that ticket in one request:

Question	Metis’s answer
Is a live service down?	Yes, 95%
Which team should take this ticket?	Technical, 93%
How urgent is it?	Immediately, 96%

Metis returns the full picture, not only the winner. For the team question, it also gave billing 6%, account 1% and sales 1%.

Confident? Act on it. Unsure? Pass it to a person.

The probabilities are the point. When Metis is confident, your process can act on the answer automatically. When it is not, the case goes to a person. You decide where that line sits, question by question.

That makes Metis a good fit for the routine decisions that pile up in a queue: routing tickets, checking a request against a policy, sorting documents, or choosing the next action in a workflow. The clear cases move on their own, and your team spends its time on the ones that need judgement.

Small, fast and light on energy

Metis is a small model, and it does far less work per answer than a chat model. It reads your case and all your questions, up to 32 of them, in one pass, and it generates no tokens.

In the support-ticket example above, Metis answered all three questions in 72 ms, using 0.0047 Wh. Those figures come from the impact report GreenPT returns with every response, so you can check the energy use of each request yourself.

How we use Metis ourselves

We built Metis because we needed it. In our Mac app and in browser use, handing routine choices to Metis has made those decisions five times faster.

We are also testing Metis in our model router. Many tasks don’t need a large model, and Metis can decide in one pass whether a smaller, more efficient model is enough. Every task that runs on a smaller model saves energy.

Hosted in the EU, on sustainable infrastructure

Metis is one of the few decision models hosted in the EU. It is served from GreenPT’s EU endpoint only, so your cases stay in Europe, and our data centres run on renewable energy. We don’t train on your data.

What Metis does not do

Metis applies the rules and facts you give it. It does not look anything up, and it is not legal, tax or investment advice. It returns probabilities, not explanations: when you want the reasoning behind a decision, ask a chat model. It also does not calculate reliably, so put computed figures in the case, such as an order of €61,800 against a limit of €24,000, rather than only the numbers they come from.

Try Metis

Use your GreenPT API key and send a case with your questions to https://api.greenpt.ai/v1/systemone, with metis as the model. If you already use the TypeSafe SDK, it works with GreenPT once you set the base URL and the model.

The Metis guide has the complete request, every answer type and the limits. Metis is named after the Greek figure of wise counsel: wise advice, for every decision.