What I think about AI, advertising, and why most of it is changing.
There's a $27 billion gap in the enterprise AI market. Nobody has built the infrastructure layer that sits underneath both agencies and AI tools.
The build cycle has collapsed from months to days. The bottleneck has moved. It's no longer 'can you build it?' It's 'can you get anyone to care?'
WPP cut 7,500 jobs last year. This year they're going bigger. The cuts land in a specific zone: the middle. The agency is changing shape.
Agencies don't sell creativity. They sell organized creative labor. AI won't kill agencies, but it will kill the way they operate.
Global AI marketing spend will exceed $107 billion by 2028. But most of that investment is being poured into the wrong places.
Eight out of ten CMOs say AI is their top strategic priority. AI accounts for 8 to 10 percent of direct marketing spend. There's a gap. A big one.
Vertical AI apps retain users at 2-3x the rate of horizontal ones. The era of "AI for everyone" is ending.
Canva has 190 million users. That's not 190 million designers. It's 190 million people who decided they don't need one. The profession is splitting.
The average mid-market company runs 21 separate AI subscriptions. The math doesn't work. And everyone knows it. Consolidation is coming.
75% of AI startup revenue flows directly to foundation model providers. It's a fundamentally broken business model.
Three of the world's largest advertising holding companies have made their bets on AI. The strategic clarity behind them is another matter entirely.
Everyone in marketing talks about personalization. Almost nobody does it well. Here's what the winners look like.
Most companies are trying to teach AI their brand by cramming a 500-word style guide into a system prompt. It's not working.
AI app median Day 30 retention is 4-6%. That's worse than the average mobile game. The pattern is depressingly consistent.
If you're a mid-career designer or copywriter, you're in the most dangerous position in the creative industry right now.
Two years ago, building an AI startup required venture capital. It's no longer correct in 2026. The math has fundamentally changed.
