AI models

Choose an AI model by your task, not by a leaderboard

New models appear every few weeks, and each one tops some ranking. For a company the question is narrower: which model handles our emails, documents or questions reliably, where is the data processed and what does it cost per case? We test that on your cases.

See the criteria →

30 minutes, no obligation, German or English. Bring the task and the tools your team already pays for. Clicking loads Cal.com (USA); the call runs on Google Meet. Prefer to write first? Ask by email. We send the draft data processing agreement and our terms on request.

Define
The task, what a correct result is, and the rules for your data.How it works · step 1/3
Test
Candidate models on the same real cases, ideally scored without knowing which model wrote what.How it works · step 2/3
Decide
Quality, cost per case, speed and data location side by side, with a recommendation in writing.How it works · step 3/3

In short

The best AI model is the one that does your task reliably, under your data rules, at a cost that fits the volume. Rankings do not answer that. We compare models on your own cases: quality, cost per case, speed, languages, where the data is processed and what the contract says. In typical EU deployments we use models through AWS Bedrock and Azure OpenAI in EU regions, and local models where data must stay in-house. The written proposal and terms govern scope, credit and any refund.

The criteria

Six questions decide, not one score

Public benchmarks measure general abilities on standard tests. Your task has its own documents, language and rules. A model that leads a ranking can still be the wrong choice for you.

Quality on your task

Measured on twenty or more real cases with a clear definition of correct. A starting sample, repeated after model updates.

  • Cost per case

    Price per processed email or document, not price per million tokens. Volume and length change the answer.

  • Where data is processed

    EU region, EU data boundary or your own servers. Resource location and processing location are not always the same; we check both.

  • Contract and data use

    Business plans and API contracts differ from consumer plans in data use, retention and admin control.

  • Languages and tone

    German, Swiss German or English, formal or informal: models differ more here than rankings suggest.

  • Lock-in

    We build workflows so the model can be changed without rebuilding everything, because the best choice can change over time.

Your own AI benchmark: test twenty real cases →

Data and contracts

The plan matters as much as the model

The same model can be used under very different terms. A consumer account, a business plan and an API contract in an EU region are three different decisions.

Business plans
ChatGPT Business and Enterprise, Claude Team and Enterprise and Gemini in Google Workspace each come with their own data terms. We compare them dated, from the vendors' own pages.
API in EU regions
For automated workflows we use models through AWS Bedrock and Azure OpenAI in EU regions, with a data processing agreement.
Local models
Where data must stay in-house, open models run on your own servers. Their licences are checked before use.
No training on your data
Model providers are configured and contractually bound so that your data is not used to train AI models.
How draft plus sign-off works: our approach in detail →

How it starts

A model test as part of the trial run

  1. Intro call

    Free

    You describe the task and the tools you pay for. We tell you which models are worth testing.

  2. Trial run

    from €2,500 · credited against implementation; refunded if we advise against automation

    In the trial run we can test candidate models on your real cases and recommend one in writing, with cost per case. The written proposal and terms govern scope, credit and any refund.

  3. Implementation and care

    One workflow about €3,000 to €5,000 · care from €250 per month per workflow

    The workflow is built so the model can be changed. In monthly care we run the workflow, fix errors and adapt it as your processes change. Re-testing after a model change can be part of the agreed care.

The implementation of one contained workflow costs about €3,000 to €5,000. Projects with several processes and systems cost more, and the fixed fee is in your proposal before work starts. Ongoing care starts at €250 per month per workflow. 30 min · remote · no obligation. And if AI isn't the right tool, we'll say so.

Before you choose

Questions about AI models

01Is ChatGPT, Claude or Gemini better for office work?

It depends on the task, the language and your data rules. Our comparison article shows the plans and terms side by side; for quality we test on your cases.

02Can we trust public benchmarks?

They show general ability on standard tests. They do not show how a model handles your documents and rules. Use them to pick candidates, not to decide.

03Do we need the most expensive model?

Often not. For sorting, extracting and standard replies, smaller and cheaper models are frequently good enough. The test shows it.

04What happens when a new model comes out?

A pinned model version gives you control, but providers can still change their infrastructure and products. If a new model is worth testing, we run the same test cases and you decide whether to switch.

05Can we use different models for different tasks?

Yes, and it is often sensible: a small model for sorting, a larger one for difficult drafts, a local one for sensitive data. The implementation of one contained workflow costs about €3,000 to €5,000. Projects with several processes and systems cost more, and the fixed fee is in your proposal before work starts. Ongoing care starts at €250 per month per workflow. Read the article with the cost examples. Calculate with your own numbers.

Robert Van Ysendyck

Robert Van YsendyckFounder & Managing Director

Bring that task to a 30-minute call. You will know afterwards which models are worth testing and what the test would cost.