Insights from Big Data Day London: Agents aren’t enough – why data still matters in the age of AI

This week, I went to Big Data Day in London – and unsurprisingly, agents were one of the hottest topics of discussion. The term means different things to different people, so one of my main goals was to better understand how the industry is defining and using them.

Agents: More than just a buzzword

From Google’s perspective, agents are LLMs augmented with tools. For example, a fraud investigation agent could access transaction information and even issue a service ticket automatically. But they are also an architecture in which multiple agents work together through technologies like A2A and MCP.

One point I particularly appreciated was Google’s focus on providing agents with useful memory and context. Right now, the lack of continuity is one of the biggest frustrations for AI users, so it was encouraging to see practical solutions emerging.

Beyond the tech giants, I noticed a wide range of agent frameworks being showcased – some generic, others highly specialised. A good example is agent frameworks designed specifically for data analysis, which make the technology directly applicable to day-to-day business tasks.

A business lens on agents

While the technical discussions were fascinating, what also stood out was a talk by Bibby Financial Services’ Chief Strategy Officer, Lucile Flamand. She brought a business perspective, highlighting how agents are interesting because they can support:

• Risk decisioning
• Chatbots for customer engagement
• Training for employees and customers

This struck me as a reminder that the real excitement isn’t just  the technology itself, but how it can be applied to everyday business activities.

3 ways people are thinking of agents

At the moment, agents tend to fall into three broad categories:

1. LLMs equipped with tools
2. Multiple LLMs working together
3. LLMs applied directly to business problems

Data still matters

Of course, Big Data London wasn’t just about agents. There were excellent sessions that reinforced why data management and governance remain as critical as ever.

• John Lewis – James Finlason and Dylan Saxby shared how they built a data catalogue that really works for the business, emphasising the importance of collecting and acting on user feedback. It was a refreshing reminder that data has users – and so do data catalogues.
• Lego – Chloe Thompson gave a talk that resonated deeply with me. Her focus on modularisation and reuse in MLOps pipelines felt like a natural extension of the brand’s DNA.
• Sportradar – Jeremy Jack discussed the challenges of governing a platform handling 1TB of data daily. His advice on finding business data stewards was pragmatic – emphasis the role is a step up the data management tree. Useful advice in a world where every function wants control of its data, but few want to provide resources to support stewardship – a truth I’ve seen play out in many organisations.
• Penguin Random House – Kerry Philips explained how they improved data quality through consistent monitoring and reporting approaches. A simple but powerful message: you can’t improve what you don’t measure.

And finally, I even tried to get into a session on DuckDB at scale – but the room was full. If nothing else, that’s a sign that lightweight databases still have an important place in the data ecosystem.

Closing thoughts

Big Data London was a powerful reminder of where the industry is headed. While agents may be grabbing the headlines, the foundations of data governance, quality, and usability remain just as vital.

For me, the biggest takeaway was that agents are not a single, monolithic concept. They are evolving along multiple paths – as tools, as collaborators, and as business enablers. The real challenge and opportunity will be figuring out how to apply them in ways that deliver meaningful value for people, not just technology teams.