Antonym: Ad-land's Nokia Moment
"They're borrowing your watch to tell the time, not realising that the watch has stopped."
This week, the marketing services industry has experienced what many predicted for 2027 – but it’s arrived ahead of schedule: a crash in confidence and share prices.
This is a structural upheaval rather than part of a cyclical downturn. Ad spend will rise slightly year on year, but the foundations beneath the big agency holding groups are being quietly washed away.
Pain and volatility is being felt across the sector, but WPP has been the most visible example.
WPP’s stock plunged 16% on 9 July to a 16-year low, following an emergency profit warning. Three flagship clients – Mars, Coca-Cola, and Paramount – are gone. Reactions on Fishbowl (the anonymous, work-focused social network) and in trade publications were brutal. There was talk of “a crisis of client confidence” and grim predictions that this is just the beginning.
David Jones, CEO of Brandtech, called it WPP’s “Kodak moment” – referring, of course, to the camera giant’s demise in the face of digital photography. Yet Nokia might be a better comparison: remember when the company hired a senior Microsoft leader to rescue it after losing its lead to Apple’s iPhone? This week, WPP did the same. Out goes Mark Read. In comes Cindy Rose, an expert in AI-driven business transformation.
This is not simply a symbolic handover to technology. It’s a signal that agencies now require genuine AI-literate leadership to have any hope of survival.
Was this all caused by AI? No, but AI has certainly put its foot on the accelerator.Will AI fix it? Not on its own. But leadership that understands—and acts on—AI’s implications might.
Recently, in the Emergency Edition of Antonym, we warned that leadership teams who delegate AI integration to committees and working groups will be outpaced by those who make it their urgent, central mission. The message applies to every sector, but advertising, media, and education are right in the frontline.
There will come a day when it’s too late to act. Move before that day and you’re still in the game. Miss it, and you risk a slow decline or sudden collapse.
You’ll only know you’ve missed that day in hindsight. For some, it’s already behind them. Agencies who failed to rebuild their models with urgency are now in distress. Even the quick movers aren’t immune to disruption’s second and third waves – but at least they are in the game: heads up, looking for space.
After a cruel summer, September will be the beginning of a new chapter for the sector. Rose starts her tenure as CEO on September 1st and will be taking WPP – maybe the rest of the industry – back to school.
Upping your game
This week I ran two versions of the same webinar on AI for leaders for teams in the same company. Part of the webinar was about using meeting transcripts for more than actions and minutes, for instance performance evaluation. So, practicing what I preach, I used AI to help learn and improve from the experience. Here are my notes in the What? / So what? / What next? critical reflection format.
What?
After the first session I asked Claude 4 Opus (a reasoning model) to look at the transcript and act as an expert communications and presentation coach to assess the performance of the speaker. (I wonder if it was more objective because I didn’t say it was me. So it avoided some of the chatbot sycophancy.)
The advice was practical, including
Cut down on the ums and errs (you can do this by speaking more slowly and calmly).
Have back-ups of demos as screen-recordings. ChatGPT started running slow at one point, which doesn’t matter sometimes when you’re working on something but ruins the flow during a presentation.
Have a check-list for technical set-up and recording. (Obvious, and immediately fixed.)
Prompt: Act as an expert in executive presentation and learning performance coach: Analyse this transcript of a webinar and appraise the performance of the speaker.
So what?
In an ideal world, we’d hold a retrospective and feedback session after any piece of work. In the real world, there isn’t time.
Asking for an objective evaluation of your work once you’ve finished it from an expert point of view seems to be a useful way of getting a feedback loop on your own performance and getting out of your head to think about that performance.
Humility required. I’m good at some aspects of public speaking, and this compensates for persistent flaws. It was a little bit nerve-wracking to get the verdict and and reflect on it, even though it was a machine and not a person judging me.
This was very fast, effective and practical. The reward was a sense of confidence and feeling calmer afterwards and hopefully a better experience for the client.
The assessment rang true. Because the assessment was based on a transcript you it misses some of the tone and visual cues, but it was useful enough to help.
What next?
This was a fast, simple way to learn and iterate on a presentation that was working well but could definitely be improved. It made me think about where things might be going as AI systems get more powerful and we learn to use them in new ways.
Faster, cheaper AI and innovation in meeting tools means more people will get access to this kind of feedback. Who writes the prompts and puts the values into the system will be important.
Adding video: Video processing by LLMs is coming online, so more data about presentations will be assessed, including body language and even emotional states.
Some feedback will become realtime. Smart glasses, watches, phones and windows on laptops will be able to give speakers personalised feedback or cues.
Team performance will be data we can all use to improve.
Negative potential: This personal empowerment of insights about how well you’re doing can be cast in a darker hue in cultures without psychological safety or respect. Imagine a nervous person going to present to a senior team knowing that everyone’s watching realtime critiques of what they are saying? The tech is already there to do that.
Building with AI as a team
We had some incredible progress this week on building tools and products at Brilliant Noise. I'll write more about it soon, but in the meantime, this interview with one of the co-founders of Canva echoes our experience. Most of us aren't confident coders, but we are developing confidence as a team in what is possible with building and developing products.
How else do you think AI will change how you build products at Canva? The key changes we're seeing include:
Shared language development. Teams speak more of a common vernacular around problems because they’re consulting AI.
Universal prototyping ability. Everyone can create working demos, not just engineers.
Faster testing. Ideas become tangible prototypes you can test immediately rather than through long documentation cycles.
Cross-functional problem solving. We’re seeing less role-based silos and more collaboration between PMs, designers, and engineers.
AI Illiteracy Watch
A signal of beginner levels of AI literacy is a fixation on productivity. When we are seeing the technology through the frame of what has been, it seems like increasing output and cutting jobs is the whole game. But productivity is the sugar rush that can make starting AI transformation attractive.
Focusing on AI literacy – the ability to understand AI’s strengths and limitations – for leaders and organisations as a whole (everybody needs it, but you have to get leaders learning early) means limiting frames, stupid decisions and assumptions like this can be left behind faster. In the meantime, there are no shortage of tunnel-visioned suppliers and consultants willing to help them begin removing the human intelligence and culture (the things that connect humans together to act in concert as teams and organisations). Call this type of strategy “nose restructuring to maximise facial spite yields” in consultant-speak.
Ethan Mollick posted an example of this on LinkedIn yesterday – bad advice from McKinsey:
McKinsey's new report on AI agents shows the same mindset I see in many firm's IT departments: a focus on making small, obsolete models do basic work (look at their suggested models!) rather than realizing that smarter models can do much higher-end work today (and those models are getting cheaper & better).
Their "Foundational models for agents" sidebar recommends Llama 3 8B, Mistral Small, Gemini Nano—models that are already outdated. They're worried about "minimal compute footprint" and costs while missing that frontier models are rapidly becoming more efficient.
The report celebrates agents doing password resets and expense approvals when they are capable of high end analysis (see, for example, Deep Research). I worry that companies are so focused on wringing efficiency from yesterday's technology (and reducing headcount) that they're missing how today's models could transform their actual competitive advantage, and they are going to be very surprised when a competitor figures it out first.
The old joke about consultants borrowing your watch to tell you the time could be amended in this case: "They're borrowing your watch to tell the time, not realising that the watch has stopped."
Meanwhile, OpenAI has started offering consulting to enterprise customers… PwC’s price cuts and missteps like McKinsey’s make one wonder how far behind the advertising sector’s trajectory consultancy services are.
That’s all for this week…
Thank you for reading. If you are caught in the turbulence of disruption, may you find a way to ride the waves. Be critically curious, learn and read as much as you can. Develop your AI literacy through regular practice.
See you next week,
Antony
Thank you sir. Of course I should be using AI for performance feedback. I have transcripts for pretty much every meeting, presentation, workshop I run. A gold mine for personal development.