Beyond the atrophy trap: three rules for building an AI marketing organisation

In his Mad//Fest session, David Billings argued that many organisations are approaching AI transformation the wrong way. Their focus is on automating tasks, reducing costs and increasing output. The result is often greater efficiency, but not necessarily greater capability. Over time, businesses risk giving away the expertise, judgment and operating logic that make them distinctive.

David closed his session with three practical rules for leaders who want to capture the benefits of AI without losing control of the capabilities that drive growth. Together, they offer a useful framework for any organisation rethinking its AI marketing strategy, governance model or operating system.

Rule one: own your intelligence

Most organisations spend considerable time deciding which models to use. Far fewer examine where their institutional knowledge is actually stored. According to David, that's a mistake.

“Every prompt, every human correction, and every guardrail must live in a central repository that you own.”

David Billings, Chief Strategy Officer, Empathy Lab

Brand guidelines, historical performance data, prompt libraries, approval rules and the countless human corrections that improve AI outputs over time are rapidly becoming valuable business assets. If those assets sit inside a vendor-controlled ecosystem, organisations may be creating dependency without realising it.

His recommendation is straightforward: own the repository where that intelligence lives. Access can be shared with technology partners and agencies when required, but ownership should never leave the business. This is how you preserve the knowledge that accumulates every time your teams teach a system how your brand thinks, how your customers behave and what good looks like.

Rule two: never marry a model

The AI market evolves at extraordinary speed and "models are commoditising by the week”, David argued. Models that are best-in-class today may be overtaken in months. Costs change. Capabilities improve. New entrants emerge.

For that reason, David argues that organisations should avoid becoming dependent on any single model provider. If your architecture requires you to use one model and one model only, you are limiting your future options.

Instead, organisations should strive for model agnosticism. The intelligence layer should remain stable while the execution layer remains flexible. Models can then be selected according to the task at hand and replaced whenever a faster, cheaper or better alternative appears.

David pointed to the work Empathy Lab has delivered with Reckitt as an example of this approach. Rather than relying on a single model, the platform routes tasks across multiple models and can adapt as new capabilities become available. The intelligence remains intact while the technology underneath continues to evolve.  

“Your architecture must be designed to swap out the underlying hands the second a faster, cheaper one hits the market, without disrupting the brain.”

David Billings, Chief Strategy Officer, Empathy Lab

This is one of the defining characteristics of a modern Growth OS. Value does not sit in a particular model. It sits in the orchestration layer that coordinates data, judgment, governance and execution. If you’re interested to know more about the Growth OS, download our free white paper here.

Rule three: hire for judgment

The final rule is arguably the most provocative: "You are no longer managing production. You are managing curation."

For years, marketing organisations optimised around production. Success often depended on the ability to create more campaigns, more assets and more content. AI changes that equation. When machines can generate thousands of outputs, the scarce resource is no longer production capacity. The scarce resource becomes…judgment. Someone still needs to decide which ideas deserve attention, which outputs are genuinely on-brand and which opportunities are worth pursuing.

That reality has significant implications for talent strategies. According to David, organisations should place greater value on editors, curators and systems thinkers. They need people who can recognise quality at scale, communicate their reasoning clearly and translate expertise into repeatable operating principles. They need people whose judgment can improve not just a single piece of work, but the system that produces the next thousand pieces of work.  

“You need people whose taste is so impeccably high that they can supervise a machine generating a thousand assets and instantly spot the three that are actually brilliant.”

David Billings, Chief Strategy Officer, Empathy Lab

The skills becoming more valuable are not only creative skills. They are editorial skills. Strategic skills. The ability to apply judgment consistently and at scale.

Building for the long term

At first glance, these three rules appear to address different challenges: technology governance, platform architecture and talent. In reality, they all point to the same underlying principle: ownership.

Own your intelligence. Own your architecture. Own the capabilities that make your organisation unique. Or, as David summarised during the session: "Own the brain. Rent the hands."

The models, platforms and tools will continue to evolve. What creates long-term advantage is the ability to preserve institutional knowledge, apply human judgment effectively and orchestrate technology around the needs of the business. For leaders navigating AI transformation, competitive advantage will come from building systems where expertise, data and judgment compound over time.