Greetings earthlings,

Emma here, co-founder of Spark AI. If you want to really know what’s going on in the marketing and creative world, you have to go to Cannes Lions, the annual sun-soaked festival of creativity on the Côte d’Azur. Jules and I went along to ensure our fingers are on the pulse of the industry.

«Quelle surprise», AI was the hot topic at Cannes this year. The most overheard word everywhere - from panels, to the beach, to restaurants and bars - was definitely ‘agentic’. And the overarching sentiment was 1) that craft is more important than ever, while 2) creative AI is now simply a given. And yes, those two things can go hand in hand. The question now is not whether you will use it - it’s how will you use it to gain advantage, either through creating your own IP, creative effectiveness or pure craft.

From global CMOs, to the CEOs of the giant holdcos, to the indie creatives, you could feel the same tensions - between speed and craft, between expression and automation, between bias and reality. Nobody has these things solved - everyone is working it out. But you most definitely need to be in the race.

In this article Jules and I have pooled our thinking to bring you the seven signals we are taking away from Cannes this year on the future of creative AI. These are the things that spark new thinking - not the same-same generic nonsense about AI that plagues all our LinkedIn feeds.

If you couldn’t be there, this one’s for you. And if you were there, I’d love to hear if this chimes?

Inside this edition

• Seven signals from Cannes

  • 1. Creators with AI will change the industry

  • 2. Is the future of LLMs… video?

  • 3. New players and industry moves

  • 4. AI is shifting team structures

  • 5. Leadership, pragmatism and purpose

  • 6. Think → Prompt → Think (again)

  • 7. Build in creative cynicism

• And now, back home…

  • Real stories of AI in agencies

  • Testing, testing... ChatGPT Record

Seven signals from Cannes

1. Creators with AI will change the industry

One of the things that surprised me at Cannes this year was the focus on Creators and the influencer economy. Influencers now make up 50% of the marketing budget for Unilever, who spend one of the biggest marketing budgets in the world, and for B2C brands they are now an essential part of the mix.

Creators are innovating with AI both to make things and to monitor how their content performs - to maximise ROI for themselves and the brands that hire them.

Creators are now using AI to automate scripts, editing, community engagement suggestions to maximise their venue through things like:

• Auto-generated viral video scripts

• Fan insights (when to sell, what to price at, what fans want)

• Audience segmentation + monetisation tactics

Lucy Guo (who happens to be the youngest female billionaire) envisioned a near future where creator managers use AI agents to:

• Analyse past deals, preferences, values

• Automatically find brand matches and negotiate terms

• Manage hundreds of creators, not just a handful

So it got me thinking - what can agencies learn?

1. Stop thinking of AI as a “production hack” and start seeing it as a creative opportunity.

2. Speed isn’t the enemy of strategy. It just means your team needs frameworks to think clearly fast.

3. If you really want to be in the moment responding to cultural events - how can brands sign off on content that’s generated in minutes not days? They can’t. But they can sign off on the guardrails they operate in.

2. Is the future of LLMs… video?

It wasn’t just AI that was everywhere at Cannes, it was AI generated video. Hearing Google, Meta, Luma, and HUMAIN talk about their vision for AI video expanded our minds as to what is can be.

For what AI video workflows look like right now, Jules went to the Google beach and saw how their AI Creative Sandbox makes it easy it is to build sketch to image to video workflows using Gemini 2.5 Flash, Imagen 4 and VEO-3. Meanwhile their ViGenAiR product (yes they need to work on their product names!) then automatically cuts down your long form video for every social channel. Today’s workflow, semi automated. Agencies and brands that aren’t leaning hard into this for the right kinds of projects are missing a trick.

But looking further ahead, where is video generation headed?

The younger you go, the more that video is the preferred medium for conveying information. If you listen to Amit Jain the founder of Luma, it’s because it’s the most natural medium for our brain to consume - and I guess he’s right. So what does that mean if we extrapolate a few years ahead?

As younger cohorts make video the default medium of communication, it’s not just that we’ll use AI to create more video — it’s that video itself becomes a programmable medium. Want today’s headlines served up? How about a personalised video generated on the spot, covering the topics you’re interested in to the depth you like in the language of your choice. Want to know how to fix your dishwasher? How about an explainer video created in a minute? Want to explore the latest fashions from Cos? How about a fully fleshed out brand world for you to explore.

The infrastructure requirements will be huge. Right now a single 10 second video gen requires 200,000 tokens, hundreds of times more than most text queries. Companies like Luma and HUMAIN (who also announced a partnership this week) are building that future.

So what’s your organisation’s plan for dealing with video AI?

• Are you building AI enabled video pipelines?

• Are you training your teams to prompt across modes?

• Do you know how to evaluate what a good AI-generated video looks like?

• Are your creative reviews ready for a world where a script, a video, a voiceover, and a cut can be generated in minutes?

This is where training stops being optional. It’s transformation - or a slow path to irrelevance.

3. New players and industry moves

The FT Live panel featuring Martin Sorrell, Amit Jain (Luma AI), Hannah Elsakr (Adobe), and Steve Plimsoll (HUMAIN) - see photo above - sent a clear sign that the industry is in flux in nobody is agreeing on what’s going to happen.

• Sorrell was blunt: what used to take 2 months and $2M now takes 12 days and $200K.

• Luma AI aims to generate 8 billion hours of video a day — hinting at a future where LLMs may be video-first, not text-first.

• Steve Plimsoll called out agencies’ outdated processes: “We’re still using 1950s media workflows in a 2025 world.”

• Hannah Elsakr was bullish that AI will create many more new creative roles in a similar way to cameras created photographers as a profession.

And behind HUMAIN? Saudi funding, with a clear ambition: to become the third global AI superpower. It’s a signal of where creative power might shift next.

4. AI is shifting what good team structures look like

One of the undercurrents of the AI conversation at Cannes revealed that creative team structures are noticeably changing.

The old boundaries between strategy, creative, data and production are dissolving in many agencies. As AI agents grow in capability, they’re nudging teams to work more fluidly - and in smaller, faster teams. And where agencies are developing their own AI tooling, those having the most success are having their subject matter experts lead the tech development briefing - like Havas with Converged.ai described here (Kate Ross MD of eight&four talked about this in our webinar with her too).

We also heard about new skills - that people who understand creative but can also build AI enabled pipelines are in demand. Your next hire might not be a designer or copywriter, but someone who can help supercharge the ones you already have.

And several organisations talked about collapsing silos, embedding AI specialists into creative squads, and rethinking how briefs are shaped and iterated.

The message here is that agencies thriving now aren’t only adopting AI tools. They’re redesigning how their teams work around them. In our AI maturity model we take you through on the Spark AI Accelerator we describe how companies will move from Adoption (where everyone is AI fluent) to Optimisation (where what you spend your time on changes).

We’re seeing some organisations are already there. Are you?

5. Leadership, pragmatism and purpose

Arthur Sadoun, CEO of Publicis, issued a clear call to agency leadership: this isn’t the time for more flash. It’s time to get useful.

In a moment of economic pressure and industry uncertainty, clients aren’t asking for spectacle. They’re asking for strategic clarity. Impact. Pragmatism.

Sadoun also made one thing unmistakably clear: people come first for him. His number one strategy and secret behind Publicis’s success this year? Attract and retain the best talent in the industry. Not because it looks good - but because the companies that move first on AI will win the best people. And without the right people, no platform or tool will save you.

So the question is: what kind of company are you building?

6. Think → Prompt → Think (again)

In a previous edition of Spark, we introduced a simple framework — Think → Prompt → Think — and it showed up again and again at Cannes, in spirit if not in name. It’s deceptively simple, but powerfully protective.

• Think first. Know what you're trying to get from the tool. Get clear on your creative intention.

• Prompt deliberately. One precise prompt replaces a dozen loose ones.

• Think again. Judge the result. Refine with purpose.

Originally, we shared this as a sustainability tool. Better prompts = fewer tokens = less compute. But it turns out it’s also a way to protect your cognition.

Two pieces of new research: Your Brain on ChatGPT by MIT and this Human Behaviour study in Nature show what’s really at stake when we use AI without thinking. When we rely too heavily on AI, we don’t just shortcut effort, we short-circuit our own thinking. We lose nuance, originality, brilliant ideas and critical depth. And there’s real evidence that’s happening now. Use the Think → Prompt → Think mantra.

(You’ll likely need at least 6–8 follow-ups asking it to: clarify the tone, add audience relevance, align with values, and rewrite in different formats.)

7. Build in creative cynicism

One interesting example of creative application came from Dentsu. Caitlin Ryan shared her process for building an AI creative director — and why she intentionally coded it to be cynical.

Yes, cynical.

Why? Because perfect outputs aren’t helpful if they’re just plausible. We don’t need AI to flatter us - we need it to challenge us. Her point: creativity isn’t always optimism. Sometimes it’s doubt, mischief, a refusal to accept the obvious answer.

This same spirit underpins the “think again” step in our method above. Post-output, bring your most skeptical self. Critique it like you would a junior’s first draft. Push it until it breaks - and then rebuild. The goal isn’t polish. It’s rigour.

And now, back home…

1. Real stories of AI in agencies

In a fit of high productivity I filmed three webinars this month, all interviewing agency leaders on how they are adopting AI - the highs the lows, considerations and sage advice for others. Catch up here, it’s really worth your while:

What its really like to accelerate AI adoption in your agency

Hear from the leaders of Analogue and mark-making* about how they are shifting their agencies from experimenters to AI leaders.
📺 Watch now

Why distinctiveness is your survival strategy in the age of AI

Hear how UnitedUs and eight&four agencies have doubled down on distinctiveness as their competitive edge in a world with AI.
📺 Watch now

Where agencies really are with AI

Hear from Becky McOwen-Banks and I for a down to earth chat about what we are seeing actually working - and what isn’t.
📺 Watch now

2. Testing testing… ChatGPT Record

When ChatGPT Record dropped last week, we were excited. With record mode, ChatGPT can now transcribe and summarise audio recordings - think meetings, brainstorms, voice notes. These get saved as canvases in your chat history and can be reused to create things like project plans, emails, or even code. It can also reference past recordings for more contextual responses. This is ChatGPT trying to keeping up with Gemini, so we asked our talented AI coach EJ to take it for a spin in a real-world setting.

TL;DR: it’s not quite ready for prime time. In her tests the transcription was a bit buggy (especially via the browser rather than the app), and your laptop going to sleep during the meeting is not helpful! BUT the analysis was sharp, perhaps better than Gemini, with helpful follow-up questions and a surprisingly good attribution of who said what.

Early verdict: promising but buggy right now. We’ll retest in a few weeks.

Our advice: don’t get locked in to any year-long plans of meeting recorders. This is more evidence of the landscape changing all the time. You will want to keep your powder dry to jump on the best tools when they arrive.

That’s all for this week. See you back in Blighty.

Emma

Co-founder of Spark AI

P.S. if you can’t find me this week I’ll be laying in a dark room somewhere after the SXSW / London Tech Week / The AI Summit / Cannes Lions quadruple bill 🤪.

About Spark AI

We help creative and marketing leaders turn AI curiosity into confidence - through hands-on training, strategic transformation support and tailored AI roadmaps.

We teach AI in business at Oxford, speak the major conferences like SXSW and The AI Summit and our founders have been selected in the BIMA 100 2025, the UK’s most influential people in digital and tech.

Whether you're just getting started or scaling fast, Spark helps you confidently lead with AI on your terms.

 AI Fundamentals | AI Accelerator

Not a subscriber yet?