Across the industry-agency ecosystem, everyone is investing. Platforms are being bought, agentic workflows are being designed, pilots are underway and at-scale
implementation roadmaps are being written. But one question keeps coming back to me: how often are we reaching for technology to solve what is also a workflow,
decision-making or people challenge?

The answer: more often than most are comfortable admitting. In numerous applications, AI has the potential to create genuine value across our sector. But there is also a noisy market where the gap between a confident demo and operational reality can be significant. The pace of change makes it genuinely difficult to stress-test what’s being offered or to distinguish genuine proprietary capability from a well-packaged foundational model or agent with a branded interface.

In our sector, the stakes of getting this wrong are specific and serious. Proprietary
compound data, embargoed clinical findings and patient information can end
up in third-party models before anyone has asked where that data goes or who else
benefits from it. The rise of shadow AI, employees using unapproved tools because
approved alternatives don’t meet their needs, makes this more urgent, not less.

The literacy gap also matters more than we think. In many organisations, the spectrum of readiness is striking: from people who’ve never typed a prompt to those already using AI tools daily. That spectrum shapes what can realistically be selected and deployed.

But literacy – knowing how to use AI – is only the starting point. What organisations need is fluency: the ability to think with AI, to judge where it adds value and where it doesn’t, and to integrate it into critical thinking and decision making rather than bolt it onto existing workflows. Most training programmes stop at literacy. Most adoption metrics measure literacy. The shift to fluency is where behaviour changes and it’s a point that most organisations haven’t yet reached.

Even when the right tool is selected, implementation is where value quietly leaks away. Most organisations aren’t deploying a single AI solution. They’re assembling an evolving multi-agent system, from enterprise-wide tools to function-specific platforms to specialist applications, each with its own selection logic, implementation needs
and interdependencies. Making them work together within real workflows is where things quietly break down.

Adoption moves fast where utility is obvious and stalls where it isn’t. The uncomfortable reality is that most pilots don’t deliver on their promises, not because the technology failed, but because the conditions for success were not right. There is heavy investment in selection and then the expectation is that the ‘messy middle’ will resolve itself – but it rarely does. The people side of transformation, building fluency, embedding into workflow, making use cases tangible and working with the culture you have rather than the one you wish you had, is consistently the part that gets compressed or skipped.

Perhaps the most important question is how we measure whether any of this is working. There has been a great deal of focus on efficiency: time saved; cost reduced; content produced faster.

A more useful lens is to think in terms of efficiency, value and innovation. Can it save time and money? Can it make the work genuinely better? And does it teach us something about how AI works that informs what we do next?

But the honesty mirror should also reflect some uncomfortable truths. If AI absorbs foundational content development, first-draft writing and basic data analysis, where do junior people learn the craft? That question is live on both the pharma and agency side and few organisations are addressing it. Meanwhile, regulatory frameworks like the EU AI Act are catching up and organisations that haven’t built governance into their operating model will find themselves reacting rather than leading.

Those who extract genuine value from AI will be the ones who recognised early that this was never about the technology. It was always about how work gets done, how people develop and how partnerships evolve. The prize isn’t doing the same work faster. It’s doing better work and more of it, because AI has created the capacity to be more ambitious.

What has your AI investment freed you up to do that you couldn’t do before? That’s the question worth considering.