You pay only 1/5 of the savings that actually happen. Change one address in your app and see the saving of every request in euros. The same plug also gives you protection and sovereignty.
10-day pilot for free, you keep all the savings.
Your data trains only your own savings model.
Savings sends every request first to the cheapest model that can handle it, and shows the saving in euros, request by request.
You save €0/mo (18%), or €0 a year.
If the answer fails the quality gate, the request moves up to a bigger model.
You see the reference price and the actual price side by side.
Change the address and key in your app and set the model to auto. Savings are on from the first request.
Shield detects personal identity codes, account numbers and customer data before the request leaves for the model. You decide what happens to that data, and the request still gets an answer.
Write a reply to the complaint from the customer whose personal identity code is 131052-308T.
Write a reply to the complaint from the customer whose personal identity code is HETU_1.
Dear customer (identity code 131052-308T), we apologise for what happened and will handle your complaint without delay.
The request runs on AIvantage’s own capacity in Finland, and the data never leaves the country.
The cloud model sees a placeholder such as HETU_1 instead of the real value. The real value is restored in the answer.
If neither will do, the request is not sent.
Server location alone doesn’t decide who can access the data. What matters is which country’s company runs the model.
The CLOUD Act obliges a US company to hand data over to the authorities, even if the server is in Frankfurt.
China’s intelligence law obliges companies to support state intelligence. DeepSeek stores its users’ data in China.
The requests you choose run on open models on our own capacity in Finland. An open model is not as good as the best cloud model, so other traffic continues to the cloud through Shield.
Janne has worked in IT for more than 20 years. He founded the digital agency Aucor Oy, grew it into a profitable expert firm and took it to an exit.
He holds an EMBA (Quantic, 2023), is a certified board professional (HHJ) and a certified turnaround consultant (TMC), and has written several business books. Pilot enquiries go straight to him.
Turn on savings and do three things: accept the terms, add your model provider’s API key and switch your application to AIvantage’s address and key and set the model to auto. The portal gives them ready to copy. It takes five minutes. Savings start from the first request, and all savings from the first 10 days are yours.
Yes, when you use Claude Code with an API key. You add AIvantage’s address, key and model to Claude Code’s settings, and the portal gives them to you ready to copy. Claude Code signed in with a Pro or Max subscription can’t be routed through AIvantage, because the subscription is tied to Anthropic’s own service.
Anthropic (Claude), OpenAI (GPT), Google (Gemini) and xAI (Grok). You bring your own API key for one or more of them, and the bill still comes straight from the provider. You can also use AIvantage’s own models, which run in Finland.
No. If your application uses an OpenAI- or Anthropic-style API, such as the OpenAI SDK, the Anthropic SDK or LangChain, you change the address and key and set the model to auto. The same works in any tool that lets you set the API address yourself.
You don’t have to trust blindly. Create a separate key for AIvantage in your model provider’s console. You can see in the provider’s own view what it is used for, and revoke it there in seconds. AIvantage stores the key encrypted (AES-256-GCM) and never shows it to anyone after it is saved. From prompts and responses we store only training excerpts for your own savings model. They are deleted after 180 days, and collection can be switched off. During the pilot Shield only observes, so nothing changes in production because of it.
Yes. Your app sends requests to the AIvantage router, which runs on a server of the Finnish provider UpCloud in Helsinki. From there the request continues to the model provider your rules allow. From prompts and responses we store only training excerpts for your own savings model. They are deleted after 180 days, and collection can be switched off. For reporting we store metadata such as time, model, tokens and price. The Shield log records which data types were found and what was done with them, but not the values themselves, and it is kept for 90 days by default. The mapping between masking placeholders and real values is held encrypted only in the router’s memory and is forgotten an hour after last use.
A cheaper model’s answer is returned only if it passes the quality gate. Otherwise the request moves to a bigger model. On top of that, a sample of requests also runs on the strongest model your rules allow, and the answers are compared side by side. The result appears in the pilot report. For a task where you want no compromises you can lock the model. Then no savings arise, and nothing is billed either.
10-day pilot for free, you keep all the savings.
Your data trains only your own savings model.