Agentic AI: Who’s Steering the Future?

As AI moves from a passive tool into something that interprets, prioritises, and acts, the question of who remains in control has become one of the most consequential strategic challenges facing senior leaders today.

We brought together a room of senior executives at the Royal Automobile Club with two people who have worked through these questions inside government and business.

Lord Liam Booth-Smith, Head of EMEA Policy at Anthropic, served as Downing Street Chief of Staff from 2022 to 2024 and has spent the past year inside one of the world’s leading AI safety organisations.

Oliver Shah, Founder of Newcome Advisory and Former Associate Editor of The Sunday Times, has spent fifteen years covering the decisions that define British business.
Together, they offered a perspective that moved well past capability announcements and into the harder questions of adoption, governance, and responsibility.

The conversation was hosted and chaired by Lord Stuart Marks of Hale CBE, Chairman of L Marks, who opened with a framing that set the tone for the discussion: as AI systems begin to act, not just respond, the question is no longer what they can do, but who is ultimately steering.

“As AI becomes more agentic, moving from answering questions and taking actions and making decisions, are we steering, or are we being driven?”
–  Lord Stuart Marks of Hale CBE, Chairman, L Marks

The gap that matters most is not technical

AI capability is advancing faster than organisations can absorb it, and that gap is widening. The technology itself is not the limiting factor. What slows transformation is the absence of leadership clarity, and the tendency to treat AI as an IT decision rather than a strategic one.

Organisations that are moving are doing so because their leadership has made a deliberate choice to engage with it structurally.

The corollary to this is that the risk most organisations should be focused on is under-utilisation, not over-reliance. Decision-making inertia is a more immediate threat than the scenarios that tend to dominate public debate.

“Depending on who you talk to, AI is a tool… a weapon… a utility… or even a new species”

Governance is already behind

Across most organisations, employees are using AI tools today without oversight and without a clear understanding of what data they are sharing or where it is going. The frameworks that should govern this are still being written, which makes governance one of the most immediate risks for leadership teams, regardless of sector or scale. Awareness at the top of most organisations is still behind actual usage on the ground, and that gap creates exposure.

The transformation follows a sequence

The transformation follows a defined sequence. The first stage is capital reallocation toward AI, which is already underway across most sectors. The second is business model and operational restructuring, which follows as organisations work out what AI integration actually requires of them. The third stage is social and workforce disruption, as the effects of that restructuring become visible at scale. The fourth is political and regulatory response, which arrives once disruption is widespread enough to demand it. Most organisations are currently navigating stages one and two, which means the consequences of stages three and four are still ahead, and the decisions being made now will determine how much room there is to manoeuvre when they arrive.

Labour market impact will be gradual rather than sudden, concentrated first in white-collar roles, and differentiated by sector. The more useful frame is transition dynamics rather than end states, because the disruption will emerge incrementally and the organisations best placed to navigate it will be those that started preparing before it became urgent.

Enterprise adoption is where advantage is built

The most durable competitive advantage will come from deep integration of AI into core workflows and decision-making, not from consumer-facing applications. B2B adoption allows for the kind of sustained, iterative learning that creates organisational capability over time. The businesses that treat AI as a strategic asset, and invest accordingly in internal deployment and experimentation, will build a compound advantage over those waiting for regulatory consensus before they commit.

Legal and regulatory frameworks are not yet fit for purpose

Regulators are working within legacy frameworks that were not designed for technology of this kind, and global regulatory approaches have not yet settled into a coherent framework. Legal clarity will come, and when it does, adoption will accelerate. The key question that legal frameworks will eventually have to answer is what constitutes an acceptable error rate for AI decision-making relative to human performance. Until that question has an answer, liability will remain tied to human accountability, and certain categories of deployment will stay constrained.

The geopolitical dimension is inseparable from the strategic one

Countries are now treating AI capability as a measure of national power and strategic influence, and the question of who controls the underlying infrastructure has become a central geopolitical issue. Most nations will depend on AI systems built by others, which means the choices made now about partnership and adoption carry political and security implications, not just commercial ones.

What this means in practice

Ultimately, the conversation kept returning to ownership. AI providers cannot define business strategy for the organisations that use their tools. That work belongs to leadership teams, and it requires both bottom-up experimentation and top-down commitment to be effective. The organisations that will be best positioned in five years are those that start building internal capability and institutional knowledge now, before strategic clarity is fully formed, because the clarity tends to follow the practice rather than precede it.

The question Lord Marks posed at the outset remains the right one to sit with: are we steering this technology, or is it beginning to steer us? The answer will depend on the decisions that leaders make in the next few years, and whether those decisions are made deliberately, or deferred until the choice is no longer theirs.