We take on a limited number of mandates. Three formats, one principle: you stay capable of acting — even without us.
Zensation researches collective intelligence on cognitive architecture — humans and machines together, on a structure that retains knowledge. Out of this research have come systems that work in companies' everyday operations: AI with real memory, run on your infrastructure, in your ownership.
Working with us buys you neither a licence nor a dependency. Every mandate is set up so that at the end you can operate the result yourself, develop it further, or hand it over to a third party. Fully transferable — that is not a gesture, it is the acceptance criterion of every project.
Three formats, one principle: you know in advance what you get, what it costs and when it ends.
AI Sovereignty Audit
The entry point
2 weeks · €9,500 (fixed price, plus VAT)
Before you invest in an AI system, you should know whether it will pay off. That is exactly what the audit clarifies. In two weeks we examine three things: your processes (where knowledge is created, where it is lost), your data (what exists, what is usable, what must never leave the house) and your organisation's absorption capacity — the honest question of how much change it can take on today. Added to this is the sovereignty check: what sits with third parties today, what should sit with you, and what would be transferable if it came to that?
Outcome: A written report with a prioritised roadmap — what is worth doing, what is not, and in which order. “It is not worth it for you at the moment” is a possible result too. The report is written so that it remains usable if you continue with someone else afterwards. That is intentional.
4–6 weeks · €24,000–42,000 (plus VAT, depending on scope)
A sprint proves on a real process what a memory system can do in your company — not on a demo. Together we pick a process with real knowledge loss: quoting, service, procurement, or the experience of a department whose key people are about to leave. On top of it we set up a pilot with ZenBrain and ZenAI: on your infrastructure, with your data, in your ownership.
The technical foundation is published and openly verifiable: 15 memory mechanisms (9 foundational, 6 PMA), an open package with 20 modules — independently judged, in our benchmarks: LongMemEval 47.7% vs. 31–43%.
Outcome: A running pilot plus a joint evaluation with measurements — not our assessment, but your numbers: what did the system retain, what did it find, what did it save the specialist department? Then you decide: keep operating it yourself, expand it, or end it. All three paths are open; documentation and handover are part of the sprint.
Retainer · €3,000–6,000/month (plus VAT) · maximum of 4 mandates
For executive teams that develop their AI strategy over months with scientific support: ongoing advice on architecture decisions, data sovereignty and make-or-buy. We assess the offers that land on your desk vendor-neutrally — we sell nothing you would have to subscribe to afterwards — and, on request, help prepare conversations with the works council, IT and shareholders.
We run a maximum of four advisory mandates in parallel — not as a scarcity gesture, but because a research lab with one researcher cannot seriously support more. When all slots are taken, it says so here.
Outcome: A second, independent voice in the room when decisions are due. Cancellable at the end of any month.
Audits and pilot projects of this kind are frequently eligible for digitalisation and innovation funding programmes in Germany, Austria and Switzerland. Whether your project qualifies is something we clarify in the first call; in the proposal we itemise the funding-relevant components separately and supply the technical project description for your application.
How you can check that this holds up
We do not keep a logo wall. What we disclose instead is fully verifiable: the published paper (arXiv:2604.23878), the open package with 20 modules — and the playground, where you can examine the memory system yourself in the browser before you ever talk to us. A research lab proves with results, not with reference slides.
47.7% vs. 31–43%LongMemEval — independently judged, in our benchmarks
15memory mechanisms (9 foundational, 6 PMA), published
20 modulesopen package — verifiable live in the browser
Sources: Open Access preprint arXiv:2604.23878 · DOI 10.5281/zenodo.19353663 · live playground.
Three questions. You answer the form below — two minutes, no registration.
A 30-minute call. You talk directly to Alexander Bering, not to a sales team. Together we check whether a mandate makes sense for both sides. If it does not, we say so — and, where possible, where you should turn instead.
Audit. The usual entry point. Two weeks later there is a report on the table you can work with — with us or without us.
Frequently asked questions
We have a works council. What does that mean for a project?
That we involve it early — which we recommend anyway. Our systems analyse processes and documents, not employees; monitoring the performance or behaviour of individuals is not part of our work. For committee discussions we supply comprehensible material.
Do our data have to leave the house?
No. Operation on your own infrastructure, or in a European environment under your control, is the normal case, not the exception. What may sit where is explicitly part of the audit.
Who will actually work on our project?
Alexander Bering, personally. There is no junior team to delegate to — which is why capacity is limited, and why the offers are cut so that this depth remains affordable.
And what if it turns out that it is not (yet) worth it for us?
Then that is exactly what the report will say, with reasons and with the conditions under which it would become worthwhile. A research lab has no revenue target that would require a different result.
For companies
Three questions, two minutes. You will receive a personal reply within two working days — no automated funnel.