The Machine of Micromanagement
This week in AI & marketing
Before I worked in leadership positions in marketing, aka CMO-things, I had many bosses. With some rare exceptions, all of them were lunatics, to be honest. Not because they were horrible people (well, some of them were outright psychos, but that’s just startup world, I guess), but because they were the worst micromanagers.
See, if there is chaos around you, and let’s be honest, most of the work we do as marketers is quite chaotic, your natural instinct is to seek control. If you can’t control the big things, obviously, so what are you going to do? You control the little things.
I am not even joking, how many times have I been in meetings about single Facebook ad copies? Or the CEO of the agency started typo-correcting an email campaign’s copy live, alongside the creative team. Most people would say, yeah, he pulled up his sleeves, wasn’t afraid of the work! Yeah, I would say it’s lunacy, micromanagement level 10,000, the most unproductive thing you can do as a leader, the absolute worst way to use your time at a company that you are leading. Celebrating this type of behavior is madness.
Thanks to AI, you don’t have to do this anymore.
The thousand reversible, informed, time-consuming decisions underneath a big-level strategy, the small minor details that are either driven by data, or niche specialist knowledge, or both, have all been automated in the last few months.
The strategy itself? No, that’s still you. That’s worth your time. That’s worth your meeting with others. That’s worth your team.
But if in the second half of 2026, you’ll either get pulled or set up a new meeting about grandiose things like a Facebook ad copy discussion, my honest advice is a simple word: run.
Let AI handle this. Focus on the things that matter. This week, entire campaign functions got automated, see below.
Seedance 2.5 might have just moved AI video into a new league
A funny thing has happened in AI video over the past year. Every new model claims to be “cinematic,” every demo gets called “Hollywood-quality,” and every launch is apparently the one that changes everything.
Most of them don’t. ByteDance’s new Seedance 2.5, however, might actually deserve some of the hype:
Early reviewers have been unusually enthusiastic, describing it as “game-changing” and “unbelievably powerful.” More importantly, they’re all pointing to the same thing: consistency. The model promises 30-second videos generated in a single pass, far more stable characters and products across longer scenes, and support for up to 50 (!) reference images. If those claims survive independent testing, they’re much more significant than another incremental jump in visual fidelity.
Length isn’t really the story here.
The real bottleneck for commercial AI video has never been image quality. We’ve already crossed the threshold where a six-second AI clip can fool almost anyone. The problem has been keeping the same product, the same character, the same lighting and the same camera logic alive for more than a few seconds without everything quietly melting into nonsense.
Marketers should care much more about consistency than duration. A 30-second clip is nice. A 30-second product commercial where your bottle doesn’t mysteriously change shape halfway through is considerably nicer.
There are still plenty of unanswered questions. Seedance 2.5 is currently in enterprise beta, pricing hasn’t been announced, and many of the most exciting features still come from ByteDance’s own presentations rather than independent benchmarks.
Still, the direction feels increasingly obvious. AI video is slowly escaping the “look what I made!” phase and entering the “let’s actually ship this campaign” phase. Some companies will still spend the next year generating astronauts eating spaghetti.
There’s one problem with OpenAI’s ad strategy
Funny story: every company that reaches enough scale eventually rediscovers advertising.
Google did. Meta did. Amazon did. Netflix resisted for years before eventually joining the party. Now OpenAI is now making it very clear that ChatGPT ads aren’t a side experiment.
At Cannes, OpenAI shared a few encouraging early numbers. Apparently, users are dismissing ads about half as often as they did when the pilot launched earlier this year, and roughly 20% of ChatGPT conversations already have commercial intent. They’ve also lowered entry costs, expanded partnerships with companies like Criteo and StackAdapt, and continue building the infrastructure advertisers expect.
So far, so good.
Then comes the forecast.
OpenAI reportedly believes advertising could generate more than $100 billion annually by 2030. That’s where things become considerably harder to believe.
WPP recently projected the entire AI search and chatbot advertising market to be worth roughly that amount by the end of the decade. Not OpenAI’s slice. The whole market—including Google, which, last time we checked, isn’t planning to quietly leave.
Forecasting a market that barely exists is admittedly difficult. Nobody knows how people will behave once conversational interfaces become the default way of discovering products. Maybe AI ads become the next search ads. Maybe they don’t.
Whether OpenAI reaches $100 billion is almost beside the point. The interesting question is whether conversational advertising becomes the next major customer acquisition channel. And that answer probably won’t come from a keynote in Cannes.
It’ll come from your performance dashboard.
Meta wants to become your marketing department
It’s easy to dismiss Meta’s Cannes announcements as another pile of… incremental AI features. Even if the list is long (brand Memory, better copy generation, creator updates, messaging agents, creative strategy space, business agents, creative approvals, translation, voiceovers, phew that was long), individually, none of them feel particularly earth-shattering.
Together, though, they’re telling a much bigger story.
Meta is trying to eliminate the gaps between every stage of digital marketing. You start with a product catalog and a campaign brief. AI generates the creative. It adapts it to your brand guidelines. It identifies suitable creators. It launches campaigns. It optimizes delivery. It follows up with customers through Messenger or WhatsApp.
Increasingly, the entire customer journey stays inside Meta’s ecosystem. That’s a very different ambition from simply making better ads.
We’ve spent years thinking about marketing as a collection of disconnected tools. One platform for creatives, another for influencer management, another for media buying, another for customer support. Yet another for reporting.
Meta is betting that the winning product isn’t the best AI image generator or the smartest chatbot. It’s the workflow.
That’s also why we wouldn’t obsess too much over Meta’s headline ROI figures. Yes, they’re reporting stronger returns from AI-powered campaigns, and some of those numbers are certainly impressive. But internal benchmarks always deserve a healthy dose of skepticism.
The more interesting competitive advantage is structural.
Every additional AI feature becomes more valuable because it feeds the next one. Every creator interaction strengthens campaign creation. Every campaign strengthens recommendations. Every customer conversation improves future targeting. Every new business that joins the platform makes the whole system slightly harder to leave.
Google is building something similar. Shopify is trying too, as you will see in just a second. Increasingly, the AI race isn’t about building the smartest model, which Meta definitely does not have. It’s about owning the entire workflow before someone else does.
Shopify’s Campaign Autopilot is exactly what it sounds like
Remember when building an online store meant… building an online store?
Somewhere along the way, e-commerce owners accidentally became media buyers, lifecycle marketers, email specialists, copywriters and attribution analysts. Shopify clearly thinks that’s ridiculous.
Its new Campaign Autopilot is essentially an in-house marketing agency that lives inside your Shopify admin. You set a budget, define a few guardrails, connect your channels, and the AI plans campaigns, allocates spend across Meta, Shop and email, launches them, and keeps optimizing over time. More channels—including ChatGPT Ads, Microsoft Advertising and Snapchat—are already on the roadmap.
Notice what’s missing.
Shopify isn’t trying to become another AI image generator. It isn’t promising Oscar-worthy copywriting or miraculous creative ideas.
The emphasis is on orchestration. Taking dozens of small marketing decisions that happen every day and quietly making them on your behalf.
Of course, there are limits. Campaign Autopilot still relies on your existing product catalogue and assets rather than inventing entirely new creative, and Shopify is refreshingly honest that results aren’t guaranteed. The system needs time to learn your business, and you still decide how much autonomy you’re comfortable giving it.
But the trajectory feels familiar.
First, AI helped you write.
Then it helped you design.
Now it’s starting to decide.
Adobe stops chasing AI benchmarks
Adobe has spent the last two years fighting an increasingly difficult battle.
Not because Firefly is bad—it isn’t—but because every month seems to bring another AI image model that’s faster, cheaper or simply more exciting. Competing feature-for-feature against startups is a game very few incumbents win.
Adobe seems to have accepted that.
Instead of trying to build one magical AI that does everything better than everyone else, it’s making AI disappear into the software people already use. Firefly AI Assistant is now available across Premiere, Photoshop, Illustrator, InDesign and Frame.io, while the new Firefly Studio keeps projects, assets, characters and creative context together from the first idea all the way to production.
That’s a much smarter place to compete.
Most professional designers don’t actually spend their day generating pretty pictures. They spend it searching for assets, adapting formats, maintaining brand consistency, reviewing feedback, preparing variants and trying not to lose their minds after opening their seventeenth application before lunch.
Adobe is automating that work.
And that’s probably the bigger opportunity across the entire AI industry.
The models themselves are becoming increasingly interchangeable. Today’s breakthrough quickly becomes tomorrow’s commodity. The harder problem—and the one customers are much happier to pay for—is making all those models fit naturally into the workflows they already have.
Some companies will keep competing on benchmark scores. Adobe seems increasingly interested in competing on something much harder to copy: not making AI better, but making creative work feel a little less like administrative work.
Share with a colleague. Earn paid membership months.