AI transparency

1. Starting point

AI is a technique here, and we use it when it solves the problem. DownForce builds systems for questions that have no off-the-shelf answer: data and integration, forecasting and planning, margin and pricing, evaluative models, regulation and compliance. Some of them contain an AI model, others an ordinary calculation. The client's question decides what goes in.

2. Three rules

• The decision stays with a human. A system of ours shows where something deviates and why, and puts it down there. That follows from how it is built and does not rest on the terms alone.
• Every finding traces back to its source. For every figure and every flag you can follow through to the data it rests on. Without that trail, we do not show it.
• Client data is never used to train or tune models. Not for our own purposes and not for product development. That is also stated in article 4 of the data processing agreement of Ramirez BV.

3. Evaluative AI is one of the three

Evaluative AI, systems that test a document or a file and flag what does not hold up, is one of the three domains of DownForce. Those systems assess documents, not people. They award no scores and no final judgement, they make themselves known to whoever works with them, and they deliver their findings to a human who decides. Esmond, the own platform of Ramirez BV, is built that way.

4. Esmond and the AI Regulation

On the basis of that design, Esmond does not in our assessment qualify as a high-risk AI system under Annex III of Regulation (EU) 2024/1689 (AI Act): the system is used by the student and not deployed by an educational institution to determine learning outcomes, and it assigns no grade or score. That assessment relies on the European Commission’s draft guidelines on Article 6, which are not final at this time; we revisit it on every product change and on new guidance. The system is built to the stricter expectations for AI in education: human oversight from start to finish, an audit trail per remark, and removal of findings that are not verifiable in the document.

5. What a system of ours does not do

We do not declare that something is compliant, correct or approved. A system from DownForce replaces no accountant, auditor, lawyer or supervisory authority, and the evidence it produces is no assurance or audit report. It makes visible what is there, with the source alongside, so that you can decide in full knowledge.

6. On this website

This website contains no AI systems you interact with: no chatbot, no profiling, no automated decision-making. In producing text and images we use AI as a tool. Every publication is reviewed and settled by a human.

7. Questions

Questions about our use of AI, for instance in the context of a supplier assessment or a tender, are answered at [email protected].