Services

The services in detail

Before publicationScientific review

A study, a report, a manuscript or an educational programme is to be published, submitted or offered.

Beforehand it needs an independent look at the data basis, method, sources and conclusions — according to the rules of good scientific practice and without regard to the desired outcome. For educational programmes and certifications the same applies to teaching content, sources and the qualifications promised: what the promise rests on and what the certificate actually establishes.

The report states what is substantiated, what is insufficiently substantiated and what cannot be decided on the available data. It does not replace peer review, but can precede it — and it remains confidential until you decide otherwise.

Before the decisionSecond Opinion

As a rule, an assessment of a project already exists: from the provider itself, from a commissioned consultancy or from within your own organisation. None of these parties is disinterested. Before the decision, a judgement is therefore needed that is subject to no interest in the outcome.

AI projects largely resist assessment from the outside: central processing steps remain opaque, and for most systems there is no reliable long-term experience. We state what is substantiated — and what no defensible statement can be made about at the present state of knowledge.

The provider of the service under examination is not involved. Without your express release we make no contact; your engagement, your documents and our findings stay between you and us.

The return on an independent second opinion lies in the case that does not occur: in the far-reaching wrong decision that is not taken.

After the decisionAnalysis and expert reports

A service exists or is in operation: a published study, a lecture given, a completed training programme, a certification issued or a productive AI system.

For studies we examine the data basis, method, sources and conclusions according to the rules of good scientific practice. For lectures, training programmes and certifications we appraise content, sources and the qualifications promised. For AI systems the examination covers the preparation of the data holdings, the architecture of knowledge management, the systems in use and the data quality of the connected sources.

What is under examination is the service delivered, not your organisation — even where it sits in your infrastructure.

The report records the extent to which the promised services have been delivered. Subject matter, criteria, methods and the form of the report are agreed in advance. It distinguishes substantiated findings, professional appraisal and remaining uncertainty, and states explicitly where no defensible conclusion is possible.

Documents not submitted, sources not accessible, insufficient evidence and demonstrably fictitious research are different findings. These differences are set out in the report.

The engagement remains confidential, including towards service providers already contracted. Any contact with them is agreed with you beforehand.

Where data is sensitiveOn site with a local AI

Nothing leaves your own four walls.

The system is in production and holds data that must not reach the outside: trade secrets, personnel and customer data, unpublished research data, ongoing projects.

We examine with a locally operated AI without an internet connection, in an agreed room on your premises. The data examined remains there; no transfer to external AI services takes place.

Scope, storage, handover of results and deletion are agreed in advance. A purely local examination cannot fully reproduce the actual behaviour of external cloud services. This limit is part of defining the engagement.

For the examination itself, without exception: no internet connection, no transfer. Separate from this is a second use of the same setup: should you later wish to hand over selected parts of your holdings to external models, wherever they are operated, those parts can be delimited and anonymised on site — and only they then leave the premises.

What we examine

The five fields apply before a decision or publication as well as after it — beforehand as a question to the promise, afterwards as a question to its fulfilment.

Performance claims

What a provider or a study claims — and how much of it is covered by verifiable findings.

Knowledge and data foundations

The quality and soundness of the foundations a provider invokes for its service.

Sources, method and data paths

Whether cited sources can be verified, whether the method is appropriate, and how transparent the declared data processing is.

Usability in daily work

Whether the service offered suits its intended purpose.

Cost and benefit

The plausibility of price, total effort and claimed benefit.