ADVISORY

An agent cannot answer for what it does. You can.

And increasingly you must: regulation now requires it, your insurers price it, and every outcome carries your name. Running agentic AI you can stand behind comes down to the Sovereignty Chain — three links, standing on one ground, none of which a vendor can own for you.

THE SOVEREIGNTY CHAIN · THREE LINKS, ONE GROUND

Semantic
You, not the vendor, own the meaning your agents reason from and act within.
Enforcement
A change in constraints takes effect on your schedule and fires before the action, not on the vendor's release cycle.
Execution
You can verify, independently of the provider, that an agent did what it was told.

THE GROUND · JURISDICTION

The platform running your agents is insulated from a third party's off-switch.

Lose any link and the chain collapses to whoever owns it. Lose the ground and all three go dark at once. My advisory work and my talks both run on this.

ADVISORY

I work with boards and executive teams as a non-executive director, board advisor, and fractional Chief Data Officer on the governance of trustworthy agentic AI: the meaning layer agents reason from and act within, the mandate that governs them, and the liability that comes with it.

Twenty-five years of data and AI in safety-critical industries sit behind it — upstream oil and gas and maritime, where a wrong call is a blowout or a life, not a bad quarter — from multi-agent decision systems grounded in formal ontologies two decades ago to agentic AI in maritime operations today. I am fluent in the technology; the work is strategic, not architectural.

What I advise on

I see agentic AI as this decade's defining value opportunity; the work is capturing it on a foundation you can trust. Every topic here is a facet of one problem: governing trustworthy agentic AI. No hype, and vendor-neutral throughout.

GenAI vs agentic AI

GenAI answers and you review it; an agent acts under your authority.

Rent the model, own the meaning

Own the meaning layer rather than rent it, inferred, from a data platform.

The value case

The cost of governing well vs the far larger cost of an agent acting wrongly under your name.

Use cases, workflows & MVO

Where to start with the minimum viable ontology, and how single agents join into connected workflows on shared meaning.

The meaning layer

The governed ontology agents reason from and act within, owned in an open form.

Governance at machine speed

Human-tempo controls break when agents act in seconds; control must fire before the action.

The board’s duty

Fiduciary oversight, insurability, and personal liability, now statutory under NIS2 and KRITIS.

Organisation design & operating modes

The CDO office and its three operating modes.

Roles & responsibilities

Who is accountable across board, CDO, CIO, COO, CRO, and Process Owner, with joint custody and two gates.

The common thread

The living meaning layer

An ontology is owned only if it is kept current — a stale definition is an ungoverned decision an agent still acts on.

Where it matters most

Sharpest where the stakes and the statute are highest. In critical infrastructure, energy, maritime, and transport, the NIS2 Directive and Germany's KRITIS-Dachgesetz now place personal, statutory liability on the board for what its systems are permitted to do, agents included, against a threat to the grid that is no longer remote. That is my heritage sector, and the ground where owning the whole Sovereignty Chain stops being a principle and becomes a duty with a deadline.

Available for non-executive director appointments, board advisory, and fractional CDO mandates focused on agentic AI governance.

SPEAKING

Keynotes, board briefings, and executive workshops

For boards, leadership teams, and the CDO, CAIO, and Responsible-AI circuit. I speak on any of the advisory topics above. The signature talks:

The Sovereignty Chain
What an enterprise must own to run agents it can answer for — the meaning, the constraints, the verification, and the ground they stand on — and what it costs to rent each.
Governing agents in critical infrastructure
NIS2, KRITIS, and the board's new personal liability for what its systems do, in the sectors where the threat is no longer remote.
Why agentic pilots stall in production
Not the model, the missing foundation, and why governance is what ships an agent rather than what slows it.
GenAI vs agentic AI
The one distinction every board must get right: GenAI answers and you review it; an agent acts under your authority.

Formats
Keynote, board briefing, half-day executive workshop, panel.
Audiences
Boards and executive teams; CDO, CAIO, and Responsible-AI summits; critical-infrastructure and DACH forums.

Let’s talk

For an advisory conversation or a speaking enquiry on agentic AI governance and the meaning layer.

Get in touch