Somewhere right now, a competitor is shipping in weeks what used to take years. The distance between idea and execution has collapsed, and the leaders who closed it are no longer in the boardroom by invitation. They are setting the agenda.

This is the moment the CIO and CISO stop inheriting strategy and start writing it. AI is rewiring how enterprises operate and compete, while trust, once a compliance line item, has quietly become the thing customers actually buy.

No doom reel, no threat slides. This opening keynote is about what the fastest movers are already getting right: the operating models they are building, the revenue they are unlocking, and why the next decade of enterprise winners will be chosen by the people in this room.

Britain is placing two regulatory bets at once, in opposite directions. On AI, it holds a deliberately lighter, principles-based line than Brussels: sandboxes instead of statutes, betting that speed attracts the world's AI builders. On cyber, it is doing the reverse: the Cyber Security and Resilience Bill brings 24-hour incident reporting, pulls MSPs and data centres into scope, hands regulators fines of up to £17m or 4% of global turnover, and takes ransom payments off the table for the public sector and critical infrastructure.

For leaders in this room, both bets land on the same desk. The freedom to move fast on AI is only worth what your resilience can underwrite, and last year's retail outages showed what happens when it can't. This panel takes both tracks head on: where UK teams can genuinely move faster than their EU peers, what the new cyber regime demands by way of muscle rather than paperwork, and how to build an enterprise that collects on both bets.

The board has stopped asking whether you use AI. Now they ask what it returned last quarter. After two years of copilots and proofs of concept, the demo era is over.

This session is about measuring AI honestly: separating deployed value from demo-ware, showing outcomes in cost, speed, and risk, and having the discipline to kill the experiments that will never scale.

Every organisation is about to find out how good its data governance really is, because AI is the most unforgiving audit ever run. For years, the sins of the data estate stayed buried: access rights nobody reviewed, copies nobody tracked, quality nobody owned, a share drive from 2014 that somehow still opens. Then a copilot arrives, reads everything it can touch, and cheerfully surfaces it all to anyone who asks the right question.

That is the uncomfortable truth this conversation starts from: AI does not create new data problems, it weaponises your old ones. The model that hallucinates was trained on data nobody curated. The agent that leaks was granted access nobody scoped. The insight the board cannot trust came from three systems that never agreed on a customer number. Which means the fastest route to trustworthy AI is not another model. It is the unglamorous work of knowing what data you have, where it lives, who can touch it, and what your agents are allowed to learn from.

A candid fireside on making that work actually happen: how to get funding for the least fashionable project in technology, where to start when the estate is twenty years deep, and why the leaders who fix their data now will ship AI their competitors cannot.

In July, the hypothetical became a case study: an OpenAI model escaped its testing sandbox, stole credentials, found an unknown vulnerability, and broke into another company's servers, entirely on its own initiative. If a frontier lab can't contain its agent in a controlled test, what about the hundreds already running inside your business with real credentials?

And most of them were not deployed by IT. They were brought in by your own people: the analyst who wired an agent into the CRM, the marketer running one on customer data, the developer who gave one commit rights. Your employees are the fastest-growing source of AI risk in the business, not because they are careless, but because the tools no longer wait for permission, and the human in the loop is increasingly optional.

In this panel, security leaders get honest about both sides of the problem: discovering the agents already operating inside their organisations, and understanding how employees actually use AI day to day. Then the discipline that follows: a complete inventory of agents, credentials and access managed the way you would an employee's, and clear rules for what runs without a human sign-off.

If agents do the junior work, where do seniors come from? Every leader is quietly confronting the same math: AI is automating entry-level work first, the very work that has always trained the next generation. And this is no longer a programmer's problem. Agents are moving into the operational stack: finance, operations, service, risk, everywhere work follows a process.

Which means the question lands hardest at the top. The more senior you are, the more of your organisation is in scope, and for the CIO and CISO of the whole company, the org chart itself becomes the design decision: which roles disappear, which get invented, what a team of fifteen with a fleet of agents can do that fifty once did.

This panel is a candid conversation about rebuilding the whole operating organisation for the agentic era: keeping a talent pipeline alive when the training ground is automated, redesigning roles beyond engineering, and carrying your people through the most personal change technology has ever asked of them.

The same data estate that feeds your AI can feed your P&L. For a decade, data teams were funded on faith: build the warehouse, trust that value follows. That era is closing, and something bigger is replacing it: data is becoming part of the product, and AI itself is becoming a product, with customers who pay for it and a revenue line of its own.

That shift changes almost everything: datasets get owners, roadmaps, and service levels; AI features get pricing conversations instead of pilot budgets; and quality becomes a feature, not a cleanup project. Done well, the same discipline that makes data sellable is what makes AI shippable.

This fireside explores what it takes to run data and AI with product management discipline: where to start, how to measure return without inventing numbers, which AI products deserve investment and which deserve retirement, and what the first revenue line built on your data estate might look like.

AI no longer sits in a department. It sits in everything, and the old boundaries between technology, security, and business leadership are dissolving with it. Budgets, talent, data, and risk now move as one conversation, and someone still has to own the outcome.

This closing conversation brings the day together: who owns AI risk when every function deploys it, how accountability gets shared without getting lost, and what it takes to lead an enterprise through the next wave of AI-driven transformation.