For all the industries to choose, it’s marketing where AI has emerged from an “innovation lab” side-project into how briefs are written, production workflow followed, approvals made and media bought. One post published this December on WPP iQ and based on a webinar with WPP and Stability AI Trends indicates how AI is being used in day-to-day operations, specially in marketing agencies.
Here, we are focusing on the practical realities of constituent constraints that determine whether AI brings change to daily work or adds a level of complexity/formality/tooling.
Brand accuracy is a repeatable capability
The AI used by marketing agencies treats brand accuracy as something that can be engineered. WPP and Stability AI observe that off-the-shelf models “aren’t trained on your brand’s visual identity”, so results can frequently look generic. The company’s cure is tuning, that is, training me on brand-specific data sets so I learn the brand playbook, including its style, look, and colors. Then, those factors can be consistently replicated.
Argos from WPP is a great example. After training a model specific to the retailer, the team explained that their model was detecting details beyond just characters, such as lighting and subtle shadows used in the brand’s 3D animations. Mimicking those subtler effects often involves time going into the void in production, courtesy of re-rendering and rounds of approval. As AI outputs become increasingly “ready,” teams spend less time improving and more time shaping the narrative and translating the media across different channels.
Cycle time compresses (and calendars shift)
Traditional 3D animation can be too slow for reactive marketing, WPP and Stability AI note. I mean, cultural moments require immediate content, not cycles measured in weeks or months. In its example Argos case, WPP trained the custom models with two 3D toy characters so that they learned their appearance and their behavior – for instance, how they hold themselves or objects.
The result was “high-quality images…produced in minutes rather than months.”
It’s a fast-moving process, but also one that adds rather than subtracts production touchpoints. Assuming creating the variants themselves is fast, the constraints are review, compliance, rights management, and distribution. The issues were always there, but the speed and treatability of AI in this context highlight how close we are to the possible, rather than getting alerts from systems that have become systematised and accepted into workflows. Agencies that want AI to transform daily routines have to change how tasks are done around it, not just add the new technology as a tool.
The “AI front end” is becoming increasingly critical.
“Creative teams waste time because interfaces to common tools are ‘disconnected, complex and confusing,’” as WPP and Stability AI put it in their announcement: they end up working on workarounds, they have to move assets between them all the time… a UI problem. In many cases, replies are custom front ends, specific to the brand and with complicated back-end workflows.
WPP describes WPP Open as a tool that translates WPP’s know-how into “globally available AI agents” that teams use to plan, produce media, and sell. Operational efficiencies are derived from smoother handovers between tools when work moves from briefs to production, from assets to activation, and from performance signals back into planning.
Self-serve capability changes agency operations
Client-facing AI-powered marketing platforms are on the rise. Operationally, that pushes agencies to focus on the parts of the workflow that their clients can’t self-serve with too much ease — like designing the brand system itself, crafting detail-oriented fine-tuning, and embedding governance.
We will see a shift in governance from policies to workflows.”
For AI to become an everyday part of our lives, governance must be embedded across every work environment. Dentsu talks about creating “walled gardens”, digital environments to prototype and develop a suite of AI-enabled solutions in security, and also spaces where the best ideas could be commercialised. This mitigates the risk of sensitive data exposure and enables experiments to flow into production systems.
Planning and insight compress too
The effects go beyond production operations alone. That’s compared with AI-powered content strategy and planning developed by Publicis Sapient, which “turns months of research into minutes of insight” by leveraging big language models alongside contextual knowledge and prompt libraries [PDF]. Research and briefs accelerate workflow, enabling more client work and allowing the agency to respond quickly to a changing culture and platform algorithms.
What changes for people
From these examples, the effect on marketing practitioners is a rebalancing and migration of job descriptions. Less time is spent on mechanical drafting, resizing and versioning, more time is dedicated to being a brand steward. New operational roles are emerging, with titles such as model trainer, workflow designer, and AI governance lead.
AI is making the most operational difference when agencies use their own models, have usable front ends that make adoption (particularly by clients) frictionless, and integrated platforms that tie planning, production, and execution together.
The obvious benefit is speed and scale, but the bigger shift is that marketing delivery begins to look more like a software-enabled supply chain–standardized with customization where necessary, flexible when needed and measurable.
