Experian's CPO on taking agentic AI from ambition to reality

Financial services firms are still raising their AI budgets, yet far fewer say they are ready to run agentic systems at scale. For one product chief at Experian, the gap has less to do with the models than with the foundations beneath them.

David Gallihawk is chief product officer for business information, verticals and platforms at Experian, the credit reference and data analytics company. In this contributed piece he sets out his own view of what separates ambition from execution in agentic AI.

Agentic AI is no longer ‘new’, and as businesses become increasingly au fait with integrating agentic systems into their operations, Deloitte offers timely research that reminds us of the gap between ambition and execution. While nearly three in four (74 per cent) of leaders anticipate nearly half of their business processes being redesigned around agents in the next four years, only 5 per cent think they’ll be AI-ready.

2027 will be here in no time, and as businesses look at their performance and priorities for the year ahead, the conversation is shifting. Previously we were focused on outputs from operational efficiencies. Now, agent capabilities are driving outcomes with real, practical and measurable results.

Our own research reflects this: while financial services (FS) firms keep increasing investment into AI, nearly half also say that integrating data into these workflows is a challenge. Another third cite weak data lineage and a further third siloed data.

The question is no longer ‘will we adopt AI?’ Instead, it’s how to do so effectively, turning ambition into reality.

Starting with the ‘why’

FS organisations have spent years gathering and managing huge volumes of data. And they’re no stranger to automation and AI.

But as AI evolves and scales, the next step in the journey is connecting these data assets to critical business questions and decisions that matter. Firms need to know why they’re implementing AI for it to be truly effective: is it seeking to improve operations, speed up delivery, or drive consistency in strategic planning?

The shift in mindset here is important. AI, particularly agentic AI, only adds value when it’s tied to clearly defined outcomes. Whether that’s better customer service, early risk detection, improved resilience or personalisation, starting with a use case ensures AI remains a strategic business resource, not just another technological input.

Iteration through testing

Potential starts with experimentation, and that’s where sandbox environments come in. As opportunities expand, these testing grounds are becoming a core part of AI adoption, giving organisations a controlled setting to see how agents behave, learn, stall, and where human oversight might be needed.

But good sandbox environments don’t just generate insight. They help operationalise systems, turning tested outputs into outcomes through reusable frameworks, workflows and code for scalable application. Work that once took weeks can be completed in days or hours, letting teams test, refine and inform their decisions faster.

Creating foundational clarity for scale

We now know that AI is about more than the model. It needs trusted data, clear context, strong governance and the ability to turn insight into action. Without these elements, systems can move quickly but risk accuracy and reliability.

This is particularly true in highly regulated sectors like FS, where confidence and trust in guardrails and consistency are essential to adoption. And as frontier models become ubiquitous in day-to-day work, agents raise the stakes, requiring stronger design, orchestration and visibility as to how decisions are supported, monitored and refined.

The nature of FS is one in which you can’t act on answers you can’t explain. That means investment in the foundations mentioned above is essential to move us from a world of ambiguity to clarity while balancing compliance, innovation and speed.

Delivering on AI’s potential

We’ve heard about AI’s potential ad nauseam. Now, the industry wants to see proof, and the best examples come from organisations using AI to solve real problems.

We recently worked with a high street bank to do just that, partnering to strengthen their customer support during moments of financial vulnerability. In identifying changes in behaviour and circumstances sooner, the bank was able to proactively step in rather than waiting for problems to escalate. The insight generated, and the action enabled by AI, is helping to actively mitigate risk while helping customers in need.

That’s why the real opportunity today is in applying the way AI can deliver demonstrable value. That hinges on continued investment in the tools, environments and expertise that help organisations design, test and deploy systems and agents responsibly.

What this all boils down to is simple: those businesses that focus on executional foundations first will be the ones who pull ahead in the AI race. Success will depend less on the sophistication of software and more on the ability to make agentic solutions practical and applicable, delivering good outcomes for both consumers and businesses.

Only then will ambition become reality.