Redefining audience intelligence: can AI really deliver on its promise?

The current state of the AI promise

The question has shifted. It’s no longer whether to use AI, but whether marketers can truly trust the decisions it’s making on their behalf. What data should inform those decisions? How explainable are they? And how can marketers have confidence in outcomes increasingly driven by automation?

These questions sat at the heart of a panel I hosted at Adweek Europe, joined by Kara Osborne Gladwell, Global Product Architect Officer at dentsu Media Global, and Olya Dyachuk, Global Media & Data Director at The HEINEKEN Company.

We explored AI through three interconnected lenses:

• Experian as the data and identity partner fuelling AI engines
• dentsu as the architect designing AI driven media systems
• Heineken as a global brand applying AI at scale

The AI trust gap

We opened the session with a live audience poll, and the results were telling. Almost everyone in the room was already using AI in their day-to-day work. Almost no one fully trusted it.

That tension defines the moment marketers are in. The potential of AI is widely recognised – but confidence in automated decision making hasn’t yet caught up.

This trust gap is fast becoming one of the biggest constraints on AI-led marketing. Not capability, but confidence in the inputs, decisions and outcomes that automation produces at scale.

The discussion quickly moved from capability to credibility. What does it actually take to build AI systems that are trustworthy, transparent and commercially impactful?

From the Experian perspective, three things matter above all else:

Trust isn’t philosophical for us – it’s operational. It’s embedded in how we provide trusted data and identity that organisations rely on to make decisions across channels, markets and regulatory environments.

Quality data: the fuel AI depends on

A simple analogy resonated throughout the panel: AI is an engine – but engines are only as good as the fuel they run on.

As automation accelerates, signal quality matters far more than signal volume. Performance increasingly depends on whether models are trained on accurate, credible data – and whether they can differentiate meaningful signals from noise.

A more sophisticated model, or a more engaging interface, can increase speed. But only trusted data improves the quality of decisions AI makes.

Success isn’t just about having the right data sources. It’s about how effectively signals are curated, connected and continuously evaluated. The real value comes from understanding how signals reinforce – or contradict – each other, and interpreting them through context: who, what, where, when and why.

This matters because poor signals don’t just degrade performance. They introduce:

• Bias baked into automated decisions
• Hidden compliance and regulatory risk
• Brand and reputational damage that compounds rapidly

At Experian, our view is clear: if you can’t explain where a signal came from – or why it’s trustworthy – you shouldn’t let AI optimise against it. Data quality has always mattered. In an AI-led ecosystem, it’s non-negotiable.

Identity: anchoring assumptions to reality

Kara spoke about how dentsu is using synthetic audiences to explore niche segments and accelerate learning – whilst emphasising the need to ground them in identity to derive true meaning. That distinction is critical.

Identity anchors decisions to people, not abstractions or inferred proxies. It ensures that as campaigns move across channels, platforms and markets, systems retain a consistent understanding of who they are engaging, how those individuals behave, and which touchpoints truly drive outcomes.

The point of identity isn’t granularity for its own sake. It’s continuity – the ability to connect decisions across channels and understand why outcomes occurred, not just that they did. Identity also underpins explainability. When decisions can be tied back to known individuals or households, marketers can understand why outcomes occurred – not just what happened. Without identity, AI risks optimising toward assumptions instead of reality, gradually drifting away from the customers it’s meant to serve.

Humans stay in the driving seat

Human oversight emerged as a consistent theme – and for good reason.

AI may be the engine, but people must remain firmly at the wheel. That means governance: monitoring bias, hallucinations and misalignment with brand or business context. But it’s also about creativity.

Human oversight doesn’t slow AI down -it makes it more valuable. It ensures automation remains aligned to brand, culture and commercial intent, rather than optimising blindly toward short-term efficiency.

As Olya noted, if everyone hands planning over to the same algorithms, we risk creating a “sea of nothingness.” Cultural nuance, originality and bold thinking still require human imagination. She captured it perfectly with the concept of “AI + PI”- Artificial Intelligence powered by People Intelligence. The role of humans isn’t to defer to machines, but to challenge, guide and amplify them.

An AI framework that actually delivers

We closed the session by mapping a practical framework for AI that works in the real world – based on the principles we consistently see in systems that deliver meaningful audience intelligence across brands, agencies and markets: