The past six months have brought a decisive shift in how the creative industry is engaging with AI. The conversation has moved from pilots and proofs of concept to grappling with production-scale agentic deployment. The announcements have come thick and fast, the ambition is real, and the momentum is undeniable. But what is becoming increasingly clear is that the speed of adoption and the readiness of enterprise architecture beneath it are not moving at the same pace. And as the industry prepares to gather at Cannes Lions, the organisations that have spent the past year exploring agentic capabilities are confronting a harder problem: not whether to deploy agents into production, but how to do so safely, reliably, and at enterprise scale.
The challenge is no longer model capability, or even the capability of the agents themselves. It is the complexity of operationalising autonomous systems across fragmented systems, governance, security, and deployment environments. And as Cannes Lions approaches, it is that gap – between agentic ambition and production reality – that is the conversation worth paying close attention to in Cannes this year.
AI agents have already arrived in post-production, broadcasting, and advertising
The scale of what is already underway is evident across the industry. Avid and Google Cloud recently announced a multi-year partnership to embed agentic AI directly into professional post-production workflows – moving video editing from a largely manual process toward intelligent, autonomous orchestration of media discovery and production at scale. And Nvidia, Adobe, and WPP are collaborating to deploy agents simultaneously across creative production and content activation, with autonomous systems continuously generating, adapting, and activating content across millions of audience and channel combinations.
The same advancements are progressing across broadcast too – agents are now handling content routing, metadata verification, and archive management without constant human intervention, unlocking value from content libraries that were previously inaccessible or underutilised. And in advertising, agentic buying and selling systems are beginning to negotiate and transact autonomously across media surfaces, with AI agents carrying brand objectives, standards, and budgets at machine speed.
These recent announcements are solidifying that the creative industry is at the centre of this adoption, and Cannes Lions represents the largest gathering of senior industry leaders since many of these developments landed.
The questions creative teams aren’t asking loudly enough
But despite all of that innovation, the speed of adoption continues to outpace the architecture beneath it, and brands are beginning to discover that deploying agentic AI responsibly is fundamentally different from any other technology transition they have navigated.
Take the integration challenge alone. Creative workflows don’t run on a single system. They span production tools, asset management platforms, rights databases, media activation channels, and client data environments simultaneously. Agentic systems operating across this landscape must maintain context, make decisions, and coordinate across all of it – without the predictable, linear behaviour organisations have built their operations around. For media leaders, the exposure this creates isn’t just technical complexity. It’s the strategic vulnerability of running commercial operations on foundations that weren’t built to support them.
The governance challenge cuts closer still. When autonomous systems are interpreting brand guidelines, making creative decisions, and executing media buys at machine speed, accountability becomes genuinely difficult to locate. For organisations where brand integrity and client trust are core commercial assets, that isn’t a process problem to be managed – it’s a strategic one. Where does human oversight begin and autonomous execution end? And who carries accountability for decisions made faster than any review process can follow?
What strikes me is that even within organisations that are asking the right questions, these challenges aren’t being stress-tested loudly enough yet. Underneath both sit a data ownership problem the creative industry has yet to fully reckon with. In environments where client brand data, audience data, and creative IP are all in play, poorly designed systems can expose sensitive information across platforms in ways that are difficult to detect and often impossible to reverse. That is a reputational risk as much as an operational one – and for the organisations building competitive advantage on audience trust, it has to be treated as such.
Getting this right demands more than ambition
Deploying agentic AI safely and reliably in this environment requires architectural sophistication from day one. Retrofitting autonomous systems onto infrastructure designed for human-operated workflows will not work – instead, enterprise architecture must be rebuilt around autonomy from the ground up.
Before anything else, getting this right demands a shift in how the problem is framed. The most effective agentic deployments we are seeing aren’t built around broad autonomous systems asked to do everything. They are designed around well-defined workflows, where AI handles specific tasks with precision, context is preserved across steps, and human oversight is embedded at the points it matters most.
And fundamentally, this is the gap that existing fragmented AI stacks cannot close. When teams are assembling their AI stack from disconnected tooling, each requiring separate integration, monitoring, and maintenance, governance becomes almost impossible to enforce end-to-end. Fixing this means replacing that fragmentation with a coherent foundation built for production from the start.
Importantly, it also demands understanding the entire AI stack as an integrated system – from the foundational infrastructure underneath through to how agents interact with creative workflows, content pipelines, rights environments, and activation channels. For media teams, that full-stack view is what will differentiate between AI agents that create competitive advantage and those that contribute to exposure.
The organisations that will come out ahead are those that architect agentic systems thoughtfully, with security, scalability, and governance embedded as foundational requirements — creating environments where autonomy operates within well-designed boundaries and AI augments rather than replaces human judgement.
The question worth asking at Cannes this year
As Cannes Lions approaches, the creative industry will gather to celebrate ambition – and increasingly, to interrogate the infrastructure beneath it. The question now is whether the foundations are in place to engage with agentic AI responsibly, at scale, and with the governance that protects what matters most.
That is a conversation I am looking forward to having at Cannes – especially as I bring our newly launched Stelia AI OS into these conversations with organisations moving past experimentation and in need of the production-grade foundation to execute with confidence.
