Innovation Day London 2026 took place in a setting that matched the ambition of the conversation. The Conservatory at the Barbican (a space more often associated with quiet reflection than business transformation) became the backdrop for a day centered on one question: what does it actually take to make AI real?
Organized by Empathy Lab and EPAM, the event brought together business leaders, technologists and operators, not to explore hypothetical futures, but to exchange what is already working, what isn’t and where the real friction lies.
Across keynotes, a panel discussion and a series of immersive demos, a consistent theme emerged: AI is no longer a question of access, it is a question of execution.
Lifting the conversation beyond use cases
The day opened with a keynote from David Billings (CSO at Empathy Lab), who deliberately resisted the temptation to start with use cases or tools.
Instead, he set the context: the infrastructure of the internet is being reshaped for a world where machines are not just supporting actors, but increasingly autonomous participants. This shift, he argued, runs deeper than the current cycle of applications and copilots.
The challenge for organizations is not simply to adopt AI, but to understand what it changes structurally. In that context, focusing on isolated use cases or headcount efficiencies risks missing where the real value sits.
The opportunity lies in how systems connect, how decisions flow across functions, and how organisations move from experimentation to coordinated execution.
Read the full article
Scaling content as a system, not a craft
That shift from experimentation to execution came into sharp focus in the keynote from Roland Butler (Head of Product at Zalando).
Speaking from within Content Solutions, the unit responsible for all fashion content across the platform, Butler laid out what it means to deploy AI at an industrial scale. Not in pilots, but in live environments serving millions of customers.
The core challenge doesn’t lie in generating content but in delivering relevance, inspiration, and consistency across thousands of assets every day, without scaling cost at the same rate.
Zalando’s approach reflects this reality:
- proving new formats in controlled environments before scaling them;
- building proprietary tooling to orchestrate multiple models rather than relying on a single vendor;
- shifting creative teams from asset production to system design.
What stood out is that AI is not treated as a creative shortcut but as infrastructure instead. A way to redesign how content is produced, distributed and optimised across the organisation. The result is not just higher output, but a different production model entirely.
Read the full article
From ambition to constraint: the reality inside organizations
If the keynotes established the direction, the panel discussion made the constraints explicit. Leaders from LSEG, Bank of Ireland, Willis Towers Watson and Ensek offered a view from inside organizations where AI is already being deployed, often under regulatory, operational or data constraints.
Across industries, the same pattern emerged: AI is accelerating parts of the system, but not the system as a whole. Coding is faster. Prototyping is easier. But bottlenecks remain:
- evaluation and validation in regulated environments;
- data quality and availability;
- organizational alignment across teams.
In many cases, progress is uneven not because the technology is lacking, but because organizations are not yet structured to absorb it.
The panel also highlighted a shift in interaction models. Users are beginning to delegate tasks to AI agents, and in some cases, systems are starting to interact directly with each other without human intervention. This raises new questions about product design, accountability, and control.
At the same time, one part of the transformation remains consistent across all organizations: people. Adoption is uneven, confidence is still being built, and meaningful change requires more than simply introducing new tools.
Read the full article
Making AI tangible: from concept to experience
Beyond the talks and discussions, the day was designed to make AI tangible. A series of demos brought the conversation into lived experience, which moved AI-conversations away from abstract capability towards concrete and direct impact.
Some demos focused on customer-facing experiences. A virtual try-on developed for Zalando demonstrated how AI can reshape product discovery into something more intuitive and immersive. An Aston Martin x Hugo Boss Formula One experience offered a different kind of engagement, turning brand interaction into something closer to participation.
Others focused on how AI accelerates decision-making, and operational performance and transformation within organizations, like our photorealistic Digital Twin for remote industrial maintenance and an AI platform that accelerates complex 'Know Your Customer' (KYC) reviews from days to minutes. Tools such as Synthetic Audiences and Neurotech illustrated how data, models and behavioral insights can be combined to inform strategy, test hypotheses and reduce uncertainty.
What these demos made clear is that AI is not a single layer added to existing systems. It sits across the entire value chain, from experience to experimentation to execution.
Returning to first principles
The closing keynote, delivered by Dr. Daniel Hulme (AI entrepreneur, academic and Chief AI Officer) brought the conversation back to fundamentals. Rather than building on the day’s examples, he reframed the underlying problem. For all the focus on data, insights and analytics over the past decade, organizations do not seem to lack information. Their real struggle is with decision-making.
AI, in that sense, is not valuable because it produces more output. It is valuable if it improves how decisions are made, especially in complex environments where human intuition falls short.
Hulme also introduced a more rigorous definition of AI as adaptive, goal-directed behavior. As systems that learn and improve over time. By that standard, much of what is currently deployed remains closer to automation than intelligence.
He ended on a broader reflection. As AI continues to reduce friction in production and access to resources, the longer-term question shifts from capability to consequence. Not just what organizations can do with AI, but what that means for how people work, create and contribute.
Read the full article
What this day revealed
Taken together, Innovation Day avoided a single narrative about AI and instead surfaced a pattern.
The technology is advancing quickly. Access is widespread. Experiments are everywhere.
And yet, the real challenge is not any of those. It sits in the transition between:
- pilots and production
- tools and systems
- capability and coordination
Organizations are done figuring out whether AI matters. They are figuring out how to make it work reliably, at scale and across functions. That is a more complex problem. But it is also where the real differentiation begins. And in a setting like the Barbican Conservatory (surrounded by something designed to grow slowly, deliberately and over time) the message landed with a certain clarity.
Making AI real is about more than speed alone. It’s about how everything connects.
Contributor in this article
