That's not obvious from how the industry talks about itself. Every year, the conversation at Cannes Lions tells you something about where an industry really is, and this year the loudest signal wasn't excitement. It was fatigue.
Sit through enough sessions and you start hearing the same panels, the same phrases, the same promises about how AI will change everything. Advertising Week called it AI's 'grown up' moment. Others were blunter, calling it a reality check. McKinsey & Company’s own read on the festival found that nearly 60% of marketers now use AI multiple times a week, yet only around one in ten have redesigned how their teams work as a result.
That gap is the real story. Not whether AI works. Whether we've built the organisations that can use it properly.
I've spent the past few years building Dragonfly AI, and talking to founders, marketers and CMOs across categories along the way. The pattern is consistent: the fatigue people feel isn't with AI itself. It's with AI theatre – the panels, the buzzwords, the co-pilot metaphors that have long since stopped meaning anything.
For a while, the industry needed those metaphors. When generative tools first appeared, the job was to stop people treating AI as a threat and get them comfortable sitting alongside it. That job is largely done. What's left is harder, and far less quotable from a stage: the operational work of actually embedding these tools into how a marketing organisation runs day to day.
This is where I think a lot of conversations about AI in marketing go wrong. They start with return on investment (ROI), as though ROI is the natural next step once you've decided to buy a tool. It isn't. The real next step is readiness – is your organisation actually set up to operate differently than it did before?
I heard this put well recently by Anastasia Leng, founder and CEO of CreativeX, on our podcast, The Creative Edge. Her point was that the technology is usually the easy part – you buy it, and it does what it does.
The harder problem is getting thousands of marketers across a large organisation to change behaviour they've held for years. That takes clean data, clearly defined roles, and people who understand why the change is happening and what it means for them. Skip that step, and you're not creating efficiency. You're scaling ambiguity, just faster than before.
I've felt a smaller version of this myself. Even at Dragonfly AI, as we've joined up new tools and data sources internally, we've hit the same wall: you assume the numbers are ready to use, then discover the underlying data was saved in the wrong format, or without the context that made it meaningful. The tool doesn't know your business. Someone still has to teach it. Multiply that across an organisation of thousands of marketers rather than dozens, and the scale of the readiness problem becomes obvious.
That's an uncomfortable truth for an industry that has, in many cases, just been through a brutal couple of years. Insight and marketing teams across CPG have taken real cuts.
Asking an already-stretched team to also be the test bed for a new way of working, without addressing readiness first, is how AI investments quietly fail.
There's a second shift worth naming, because it changes the shape of the job itself. For decades, marketing has run on the unit of the campaign: brief, create, launch, measure, stop, repeat. Generative tools make a different model possible – something closer to always-on, evergreen marketing, where content is continuously produced, tested and refined rather than released in discrete bursts.
That's a genuinely exciting shift. It's also one that almost no organisation is structurally ready for. An evergreen model doesn't work if approval processes, governance and measurement are all still built around the campaign as the unit of work. It needs new infrastructure, not just new tools bolted onto an old workflow.
You can see the same pattern one level up the stack. Cannes announcements this year, covered widely by titles including Digiday and Adweek, were dominated by agentic AI and interoperability rather than generative content itself, with some analysts projecting that AI agents could mediate trillions of dollars in global commerce within the next few years.
Whether or not that figure holds, the direction is clear: the infrastructure connecting tools, data and decisions is becoming the competitive battleground, not any single piece of creative output.
None of this points towards marketers becoming less important. If anything, the opposite. Where language models genuinely earn their keep is production – generating variation at a scale no team could manage manually, across every channel, continuously. That's a real and valuable capability. But it isn't creative judgement, and it isn't the same as understanding the customer.
I keep coming back to something Anastasia said about empathy, because I think the industry has quietly got this backwards. There's a narrative that AI will do the work and humans will step in at the end to approve it, adding a bit of humanity before it goes out.
That's not how empathy actually works. Empathy comes from a deep, ongoing understanding of who your customer is and what they're navigating, not from a final sign-off. If the person prompting the model doesn't have that understanding, the output will be fluent, confident and completely disconnected from what the customer actually cares about.
So where does that leave marketing leaders? There's a useful precedent worth remembering here. When Microsoft first introduced Excel to its own finance teams in the 1980s, the reaction wasn't celebration, it was closer to panic that the job was about to disappear. Decades later, finance teams are larger than ever. The tool changed how the work got done. It didn't remove the need for people who understood the numbers.
I'd expect something similar with AI in marketing, but only for the organisations that treat readiness, not ROI, as the first question. The technology was never really the hard part. The uncomfortable question every marketing leader should be asking isn't which tool to buy next, it's whether their organisation is honest enough with itself to admit it isn't ready yet.