Why, How, Prove, Scale, Ground: The Sovereign AI Series — Five Webinars, and the Reasoning Behind Them
Between 3 September and 29 October we are running five 45-minute sessions, each for a different person in the same building: the executive who has to decide, the engineer who has to run it, the security officer who has to sign it off, the leader who has to make it how the organisation works — and the data leader whose lakehouse it all has to stand on. Here is what each one covers, who it is for, and why we ordered them this way.
Every conversation we have with a European public body or regulated enterprise about AI eventually involves five people who rarely sit in the same meeting: the person who has to decide whether to move, the person who would have to run whatever they move to, the person who has to be able to defend the choice to a board, an auditor or a regulator, the person who has to turn one team's success into how the whole organisation works — and the person who owns the data all of it stands on. Each has a different question. So this autumn we are running five sessions instead of one, in the order the questions actually get asked — why, then how, then prove, then scale, then ground — all at 11:00 Brussels time, all 45 minutes, all with a live system rather than a deck.
Why five, and why in this order
A single "platform webinar" tries to answer all five questions at once and answers none of them well. The executive doesn't want the demo yet; the engineer doesn't want the market context; the security officer wants evidence, not enthusiasm; the transformation lead wants to know what changes for the other nine hundred people; the data leader wants to know what happens to the lakehouse they just finished. Splitting the series lets each session go deep for one audience while still building on the previous one: the decision frames what gets built, what gets built is what gets audited, what gets audited is what can be rolled out, and what gets rolled out only works if the data underneath carries meaning. Each session ends by pointing at the next, so a CIO who attends the first can send their architect to the second, their CISO to the third, their head of transformation to the fourth and their CDO to the fifth — and all five will have heard the same properties described five ways.
There is also a reason for the timing. The first week of September is when Europe's public sector is back and planning; the middle of September is when engineering teams start the work; 1 October opens European Cybersecurity Month, with NIS2 enforcement live and the Cyber Resilience Act's reporting obligations applying from 11 September; mid-October is when 2027 budgets get written — the moment to ask whether next year's AI line is more copilot seats or something that changes how the organisation runs; and the end of October is when data teams plan next year's platform work. We are meeting each audience when the question is on their desk.
3 September — Why now: Palantir alternatives for Europe
For the person who has to decide. Nathan Bijnens, with a live segment from Xavier Geerinck.
This summer changed the question from "should we?" to "which one, and how?". France's DGSI and Germany's BfV moved towards European platforms, the Bundeswehr ruled vendor-operated systems out of its defence cloud, and Paris and Berlin put a "sovereign digital backbone" on paper. In 45 minutes we cover what a Foundry-class platform actually is (four layers, not one product), the five sovereignty tests any alternative must pass — jurisdiction, keys, operating staff, disconnected mode, exit — applied live to a real estate, a fair one-slide view of the European field, and where Scrydon fits and, just as importantly, where it doesn't. You leave with a procurement checklist. The pre-read is our Palantir Foundry alternative page.
17 September — How: inside the AI OS, live on a cluster
For the person who has to run it. Xavier Geerinck.
No slides after minute five. An agent gets an identity and scoped permissions; it calls tools over governed MCP; it retrieves from the ontology rather than raw tables; it runs inside the sandbox and is stopped when it steps outside policy; and everything lands in an audit trail a reviewer will actually accept. Then the finale: the same stack brought up on a disconnected network, and what changes when there is no network — model updates, telemetry, identity, evidence export. If you have read The Harness Holds, this is that argument shown rather than written. Properties and refusals, demonstrated.
1 October — Prove: AI agents inside a NIS2 essential entity
For the person who has to sign it off. Cornelia Kutterer and Xavier Geerinck.
NIS2 is being enforced, the CRA's reporting duties are now live, and the AI Act's post-Omnibus timeline is fixed. Cornelia sets out what those regimes actually require of AI systems in an essential entity and where the shadow-AI inventory has to start; then the shared-responsibility line — what the platform supplies, what the entity owns — mirrored on our compliance pages; then Xavier shows agent governance, DLP guardrails and the audit trail live; and we close with the decision table for when air-gapped is the answer. Two disciplines on one stage, because that is how the sign-off actually happens.
13 October — Scale: from personal AI to organisational AI
For the person who has to make it how the organisation works. Nathan Bijnens and Xavier Geerinck.
Most organisations now have personal AI — a copilot in every seat — and most have noticed that the productivity of individuals went up while the outcomes of processes did not. That is not a model problem. Personal AI raises the productivity of a person; organisational AI changes the outcome of a process, and the gap between the two is four missing things: shared context (an ontology, not each person's chat history), governed action (agents that act on systems under policy rather than suggest in a sidebar), identity and permissions that follow the work across teams, and evidence a board or regulator will accept. Xavier shows the contrast live — the same question to a personal assistant and to the AI OS grounded in the ontology, then one end-to-end process run by agents with people in the loop — and we close on what it takes organisationally: the AI Centre of Excellence, who owns the ontology, and where sovereignty enters. This is the session for any sector, not only the regulated ones. The pre-read is what organisational AI is.
29 October — Ground: beyond the lakehouse, the ontology as the data layer AI & humans can actually use
For the person who owns the data. Xavier Geerinck and Nathan Bijnens.
The lakehouse gave you governed storage on open formats — table-shaped, batch-minded, read by analysts. It was necessary. It is not sufficient, because agents, and increasingly analysts, reason over meaning: entities, relationships, state and rules that people, applications and AI read from and write back to. That is an operational ontology, and it is not the same thing as a semantic layer or a knowledge graph — this is the session where we finally put the four definitions on one slide. Xavier shows the same business question answered by vector RAG over documents and by ontology RAG over the model, side by side, with provenance; then an agent writing back through the ontology to an operational system under policy. We close with the migration path from an existing lakehouse: what you keep, what you add, what you never move — coexistence, not replacement. This is the session for CDOs and data architects in any sector; the pre-read is what Palantir got right with ontologies.
What we will and won't do
We will show a real system, name the trade-offs, and say where we are not the right choice. We won't read slides, and we won't collect your details to sell you something afterwards — a person will follow up only if you ask us to. Every session will get a replay and a transcript on this site the week after, so the argument stays available to the colleague who couldn't make it.
If the questions above are live in your organisation and 45 minutes is too long to wait, the shorter route is a 30-minute sovereignty briefing — the same conversation, one to one.
Prefer to write? Email hello [at] scrydon.com and we will get back to you.