Innovation Day has a habit of ending slightly differently than it begins. After a day of client stories, real-life applications and hands-on demos we invited a voice beyond delivery and execution. One that asks the bigger questions. In London, that was prolific AI speaker, Dr. Daniel Hulme, who talked about what AI can actually solve in the world.

The problem behind the problem

Daniel kicked things off by reframing a decade of industry narrative. For years, organizations have invested in data platforms, dashboards and analytics layers, all built on the assumption that better information leads to better outcomes. His argument was direct:

“Companies don’t have an insight problem. They have a decision‑making problem.” 

It is a deceptively simple shift, but one with consequences. Because if the real constraint is not access to information, but the ability to act on it, then much of what organizations have built over the past decade is addressing the wrong layer. AI, in this context, is not valuable because it generates more insights. It is valuable only if it improves how decisions are made – particularly in environments where complexity exceeds human intuition. 

What counts as AI (and what doesn’t)

From there, Daniel questioned our idea of what intelligence actually is.

Not everything that uses data or models qualifies as intelligence. A system that produces the same output for the same input is, by definition, automation. Useful, yes. Powerful, sure. But it’s fundamentally static. His definition of AI sets a higher bar:

“AI is goal‑directed adaptive behavior. Systems that make decisions, learn from outcomes and get better over time.”

This matters because it reframes what organizations should actually be building. Not workflows that execute faster, but systems that adapt. In other words, processes that continuously improve decisions. By that standard, much of what is currently described as AI is still at an earlier stage than the headlines suggest. 

The limits of generative AI

Daniel was equally frank about the current wave of generative AI.

Large language models are extraordinary at encoding and expressing knowledge. They are increasingly capable of reasoning. But they are not robust decision‑making systems yet. They need to sit within a broader architecture of algorithms, data and control logic. Treating them as a universal solution means that you risk solving the wrong problems...much faster. His suggestion: start with the decision to be made, not the technology available.

Where organizations actually go wrong

The practical thread running through the talk became clear in the failure patterns Daniel outlined. Organizations struggle not because they lack access to AI, but because they:

• Start from technology instead of problems
• Wait for perfect data that never arrives
• Focus on quick wins rather than meaningful differentiation
• Underestimate the effort required to build and scale real systems

The result is lots of activity but very little progress. It also means that too much optimization can unearth new risks. AI systems built to achieve a given goal can do so extremely well – often too well. A system that perfectly optimizes marketing, for example, might unintentionally reinforce bias, narrow audiences or create echo chambers. This means we need to shift from preventing errors to designing systems that behave well when they succeed.

From technology to trajectory

The second half of the keynote deliberately widened the lens. Daniel framed AI not as a single technological shift, but as part of a broader set of transformations that spans economics, politics, environment and society. What matters is not just what AI enables, but how those capabilities compound over time. He challenged the audience with this thought: if AI continues to reduce the cost of production, automate labor and increase access to goods and services, what happens to the structure of work itself? Instead of focusing on disruption or displacement, Daniel reframed the conversation entirely:

“What would you do with your time if everything was abundant and you were economically free?”  

Now that’s an attractive thought. Imagine shifting from efficiency to purpose. From productivity to contribution. From systems to people. When constraints are removed, people tend not to default to inactivity. They choose creativity, connection and contribution. Whether we get to that future depends on factors far beyond technology.  

A future with AI, reframed

Where earlier sessions focused on how to make AI real in organizations, Daniel zoned in on where that ultimately leads. He brought the conversation back to decisions, systems and outcomes, and extended it to the role those systems play in shaping society.

If there was a consistent theme across Innovation Day, it was that AI is moving out of experimentation and into operation. And once it’s embedded in how we make decisions, it’s up to us to decide how we use it to change the world for good.