Greetings earthlings,

Here's your regular dose of creative AI insight from me: Emma, co-founder of Spark AI. This week we're diving into something that might surprise you: the agencies thriving with AI aren't the ones hopping between every shiny new AI tool. They're the ones who've approached AI as a business transformation and invested in their people learning to master them instead.

And speaking of learning, we've just wrapped the first episode of our AI Summer School series, where Jules demonstrated why mastering image prompting is still the gateway skill that unlocks everything else in creative AI. If you missed it, I've included the key takeaways below.

Meanwhile, it’s been a busy month of updates: OpenAI dropped ChatGPT 5, Google released a 3D world generator that could reshape game development, Midjourney's gone HD with video, and new research reveals which jobs AI is actually changing (spoiler: it's not what you think). Isn’t August meant to be quiet?! 😮‍💨

Here's what we're unpacking for you today:

• 1. ChatGPT-5 is finally here

• 2. Using AI to bring audacious ideas to life

• 3. Are you tuning into Spark AI Summer School?

• 4. AI news and market moves

  • Google's 3D world generator changes the game

  • Midjourney goes HD with video

  • NotebookLM gets video superpowers

• 5. Latest AI in Agencies benchmarking coming soon

• 6. Essential reading: Shift - AI for Agencies

1. ChatGPT-5 is finally here

OpenAI finally released ChatGPT-5 this week. After what feels like months of waiting, the response has been telling. While the technical improvements are solid (if not revolutionary) the real story is what happened when they changed how people actually use it. The backlash from paid users reveals something crucial about AI adoption that every agency should understand.

What's genuinely new for agency work:

GPT 5 has some solid technical advancements that make it even more powerful for work:

Fewer hallucinations. A constant bugbear for people using LLMs for research and data gathering, ChatGPT 5 has one-third the hallucination rate of earlier models. We recommend you still double check key facts before you build on them, but this alone makes it worth the wait in our opinion.

Multimodal creative collaboration. ChatGPT 5 can genuinely think across text, images, and audio simultaneously. Upload a rough sketch, describe your vision verbally, and reference a written brief – it understands the connections and can iterate accordingly. This isn't just about individual prompts anymore; it's about sustained creative dialogue.

Extended strategic thinking. The model can hold complex brand concepts and campaign strategies through much longer conversations. This means developing ideas through extended back-and-forth without losing the thread – more like working with a strategist who's read your entire brand bible.

Better brand context retention. Feed it guidelines, past campaigns, and strategic documents, and it maintains that context throughout the creative process, making brand-consistent output at scale genuinely feasible.

Coding ability: 5 has been optimised for coding tasks too. It can quickly build interactive prototypes and works better with large code bases. It doesn’t replace tools like Cursor, but digital agencies are going to find their dev teams just got more productive.

But the user response…

A key feature of ChatGPT 5 is the new "Model Router" that automatically selects the best model for each query. For novices this was a boon - 94% of users never selected different model and so hadn’t experienced the power of reasoning models like o3. But for power users, who were used to selecting their own models depending on the task (like we taught in our AI Accelerator) it felt like they’d had that control taken away from them. So now the model selector is back, and you can choose between GPT5 Auto, Fast and Thinking - so the complexity returns!

Along with the automatic model picker, OpenAI also retired their old model, GPT 4o. It turns out many people were extremely attached to 4o and treated it rather like a friend, confidante, therapist or trusted colleague. The new ‘personality’ of GPT5 wasn’t to their liking (no more “great idea!” for every question), and after a Reddit thread went viral OpenAI were forced to re-release 4o - telling us they would now give several months notice before retiring old models.

What does this tells us?

Basically, what we already know about technology adoption:

AI adoption is about people, not just capability. You and your clients and stakeholders will face the same challenges: team members who resist changes to familiar workflows, emotional attachments to current processes, and confusion about which tools to use when. These are human problems that require human solutions.

Your implementation strategy matters more than the tech. Having access to the most advanced AI doesn't guarantee success. The ChatGPT 5 launch shows the way you introduce and integrate AI tools can make or break adoption. A well thought through communication and change management approach is just as much a requirement for OpenAI as it is for rolling out AI within your team.

Practical considerations for you and your team:

1. Don’t just roll out tools - successful AI is about structured change management. I know, change management is hard. But you understand the specific culture, concerns, and capabilities of your team - build a structured adoption plan plan tailored to them to build confidence and competence gradually. We help you do this in our AI Accelerator programme.

2. The sophistication gap is widening significantly. Basic AI implementation will no longer impress clients. They'll expect agencies to demonstrate advanced capabilities and strategic application, not just tool usage.

3. Demonstrate sophisticated usage to your clients. Show clients integrated workflows that combine AI capabilities with strategic thinking. The value is in the orchestration and creative direction, not the individual tools.

4. Focus on outcomes over efficiency. When discussing AI with clients, emphasise creative breakthroughs and strategic advantages rather than cost savings. Position AI as expanding what's possible, not just reducing what's expensive.

Remember this when AI overwhelm strikes

When you get a dose of the inevitable AI overwhelm with all these new model releases running around remember these three things to stay grounded:

1. Your value isn't in having access to the latest models - it's in knowing how to integrate them into both your client work and your operations.

2. Strategic direction is more crucial than ever. As AI capabilities increase, the ability to direct them strategically becomes more valuable. Clients will pay for expertise in creative direction, not just execution.

3. Quality is now the differentiator, this is your superpower. When AI can handle more complex tasks, clients will focus on outcomes, not efficiency. "Can you make this faster?" becomes "Can you make this better than what we could do ourselves?"

2. How five agencies used AI to bring audacious ideas to life

Forbes recently featured agencies from Google's AI Lighthouse Program (their annual initiative that challenges leading agencies to create "impossible" ad campaigns using Google's AI tools) that have transformed their creative processes with AI. Okay, so it may read like advertorial for Google, but what struck me is not what the AI tools are capable of, and more that great creative is still fundamentally about the great ambition.

What each agency accomplished

BarkleyOKRP (Slice relaunch) used Gemini to identify cultural trends and then built an entire brand world - creating a fictional radio station complete with AI-generated songs, artists, and album art.

R/GA (Moncler campaign) built a custom tool called "Shot Flow" using Gemini as a coding partner. Shot Flow then orchestrated all Google's generative AI tools to create an AI generated film that maintained brand consistency - protecting Moncler's signature aesthetic.

McCann (Smirnoff experience) developed a custom AI co-host that learns guests' personalities and creates memorable brand interactions through personalised party themes: “The engine is designed to chat with guests, learn their personalities and quirks, and then fuse them into a completely bespoke party theme that evolves in real time as each guest RSVPs —recommending everything from the decor and playlist to a custom Smirnoff cocktail”.

Razorfish (Visit Orlando) created "The Morelandos" - 1,200 unique characters generated from thousands of Google reviews. Using Vertex AI, Gemini, and Imagen, they turned data points into an explorable universe showcasing Orlando's hidden gems.

Virgin Voyages team delivered personalisation at scale, creating distinct "postcard" ads tailored to individual interests and allowing users to upload their own photos into personalised, shareable content.

The crucial insight: start with what you do well

None of these agencies chose their most difficult problems to solve first with AI. Instead, they identified creative processes they already executed brilliantly - brand world-building, luxury consistency, experiential design, storytelling, personalisation - and developed AI workflows to enhance them. This approach gives several advantages:

• Clear baseline for comparison: They could easily measure whether AI made them better or just faster

• Reduced risk: Building on existing strengths meant less chance of failure

• Team confidence: Success with familiar processes built confidence for tackling harder challenges

• Quality benchmarks: They knew what "good" looked like before AI entered the picture

What made the difference: creative ambition matched with technical sophistication. 

Each agency combined their existing creative excellence with AI capabilities. They didn't let AI drive the creative vision - they used AI to make impossible creative visions achievable. Every example shows AI amplifying human creative judgment, cultural insight, and strategic thinking. The technology served the creative vision, never the reverse.

So the takeaway for me is the agencies standing out with how they are using AI aren't the ones with the most tools. They're the ones who've systematically enhanced what their people already do well - educated and upskilled them in AI and then built from that foundation of creative excellence.

👉 Read the full Forbes piece

3. Are you tuning into Spark AI Summer School?

Last Wednesday, Jules kicked off our AI Summer School series with a masterclass on image prompting. The principles Jules demonstrated apply to every AI interaction: clear creative intent, structured briefs, and iterative refinement. Whether you're generating images, writing copy, or developing strategy, the fundamentals are the same.

Key takeaways from Episode 1: AI for image prompting

Every detail matters. Jules showed how describing lighting, composition, and mood in your prompts isn't pedantic - it's professional creative direction. The difference between "a bottle on sand" and "a chilled bottle on pristine white sand, low angle shot with soft afternoon light filtering through palm fronds" is the difference between amateur and professional results.

Consistency comes from process, not luck. The agencies creating coherent visual campaigns with AI aren't getting lucky. They're using character references, style guides, and systematic prompting frameworks. Jules demonstrated how to build these systems.

Think like a creative director, not a user. The best AI prompts read like creative briefs. They include context, objectives, audience, and success criteria. They're documents you could hand to any creative collaborator.

👉 Catch up on Episode 1

Next up: Episode 2 on AI for Client Services

This Wednesday at 11am, our AI coach and client services expert Mette Davis takes over for Episode 2: Client onboarding and project set up. She'll show you exactly how to set up client projects for success from day one - using insights from our 3-month AI Accelerator programme.

You'll learn how to create briefing packs that keep everyone aligned, build onboarding documents that actually work, and set up projects in a Custom GPT or Gem for maximum clarity. Perfect for account managers, client services directors, and anyone who's tired of projects going off-track before they've even begun.

👉 Register up for Episode 2 - it’s free!

4. AI news and market moves

Google's 3D world generator changes the game

Google DeepMind released Genie 3, which generates interactive 3D environments in real time. Think of it as Midjourney for entire virtual worlds - you can walk through spaces that are created on the fly as you explore them.

Why this is very interesting:

AAA game development currently costs hundreds of millions and requires vast teams building environments by hand. If AI can automate even part of this process, it could democratise game development and open up entirely new forms of interactive storytelling.

For agencies, this hints at a future where creating immersive brand experiences doesn't require massive technical resources. Imagine prototyping virtual showrooms, interactive campaigns, or branded gaming experiences with the same ease as creating a mood board. Explore Genie 3.

Midjourney goes HD with video

Midjourney launched 1080p video generation, bringing cinema-quality motion to their image generation platform. Early tests show remarkably smooth, detailed video that rivals dedicated video AI tools and no longer requires upscaling. Read more from Midjourney (although they are as brief as usual!).

NotebookLM gets video superpowers

Google's NotebookLM can now generate video overviews of your documents, complete with AI hosts who present your content in a natural, conversational format. (I’m a huge fan of Notebook for sharing knowledge across the team, and digesting and analysing large reports). One more step towards video-first AI assistants that we predicted after Cannes (read more here).

For us, this opens up possibilities around:

• Briefing the team on a new client - tailored to different learning styles

• Training materials that explain complex strategies visually

• And so many more use cases we haven’t had time to imagine yet.

The technology is still early, but the direction is clear: AI is starting to make video content creation as straightforward as writing a document. Try NotebookLM's video features.

5. AI in Creative Agencies benchmarking - what to expect

Our first benchmarking report (covering October 2024 - March 2025) revealed some fascinating patterns about how agencies were really adopting AI. But things have moved on so much in such a short time! So this September, we're releasing an updated report covering March to now - and as we expected, the changes are significant.

We found that what it means to be an "AI pioneer" has shifted dramatically in just six months. This new report captures that evolution and shows exactly where the industry is heading alongside practical steps you can take.

👉 Download the current report and watch for the updated version this September. It's going to be a good one.

6. Essential reading inbound! Shift: AI for Agencies due for publication in October

Speaking of essential reading, my co-founder Jules' book launches in October! Shift: AI for Agencies is exactly what creative leaders have been asking for: a no-nonsense step-by-step playbook that shows exactly how agencies can use these tools strategically.

This isn't another book about prompt engineering or the future of work (although it covers that too). It's a super practical guide written by someone who's spent decades in creative industries and years helping agencies navigate AI adoption. Jules covers everything from how to deploy AI across your teams and build an AI forward culture, to tying AI into your strategy and talking to clients about it - all shaped by insights from working with dozens of agency leaders.

Think of it as having a smart, experienced friend walk you through what matters, what to ignore, and how to move fast without breaking what makes your agency unique.

👉 Pre-register for Chapter 1 on our website (scroll half way down the home page!) - and we'll send you the opening chapter as soon as it's ready.

Thinking of taking one of our programmes later this year?

We're booking our tailored AI programmes from autumn right now. So if Spark's even a twinkle in your eye, give us a wave early - we really hate saying no when our team's availability disappears. Book a slot in my diary if you'd like to have a chat and get on our radar.

That's all for this week. We love to have your feedback - please hit a button below and let us know how you found today's issue.

See you next time!

Emma
Co-founder of Spark AI

About Spark AI

We help creative and brand leaders turn AI curiosity into confidence - through structured AI guidance and upskilling for agencies and brand teams.Trusted by Oxford University Saïd Business School and backed by Innovate UK.

👉 Find out more 

Not a subscriber yet?