Andrew Chen on AI's Pre-Google Ads Moment and Why PMs Are Becoming ICs

AI Products Are Growing Because They Work at All

The most explosive AI products today share a strange characteristic: they have zero growth strategy. No notifications. Random acronym names. Millions of users anyway.

Andrew Chen calls this the "it works" feature. Just like the earliest days of mobile apps—flashlight apps, fart apps, Flappy Bird—the fact that an AI product functions at all is impressive enough to drive adoption. There's quantitative evidence for this pattern. An analysis of the App Store since 2009 shows massive volatility in the top 100 during early years, then gradual ossification as categories matured.

If you look at the current state of the art versus even 12 months ago, there's still definitely a step change, but the time before that is even more so. Being able to do it at all is amazing. Now we're getting smaller and smaller improvements.

— Andrew Chen

The question is whether novelty wears off. If it does, the companies that rode high growth and high churn may face a reckoning—unless they're the ones sitting on all the VC money to figure out retention. History suggests the first generation doesn't always win. Flipboard and Foursquare came before Instagram and Uber.

The M&A Wave Nobody's Talking About

Chen predicts a surge in small M&A in 2024, but not the headline-grabbing kind. This will be talent acquisitions, soft landings, and category consolidation—not transformative deals.

The math is simple. Companies that raised in 2021 cut burn by 25-30%, extended runway to three years, and raised flat rounds if they were lucky. Now those three years are almost up. According to Pilot, 60% of startups need to raise before the end of 2024. Many won't be able to.

Growth rates tell the story. A few years ago, investors wanted 3-5x annual growth and accepted seven-figure monthly burn. Today, companies have cut burn to near-zero but are growing 2-3x annually—or less. Without data to support claims they could grow faster if allowed to burn, they can't raise. That leaves three options: profitability, shutdown, or sale.

I think this is maybe a time for some of the larger incumbents to be able to get back some of the high-end talent that kind of left all the companies kind of in the last couple of years.

— Andrew Chen

The regulatory environment blocks mega-acquisitions, but there's an opening for acquihires—especially in AI. Anyone with a resume in AI started a company instead of taking a job. When the first generation flames out, expect a scramble for talent.

We're Living in AI's Pre-Google Ads Era

AI models haven't been commercialized yet. Chen compares the moment to when you could search AltaVista or Hotbot and see zero ads—just the search engine trying its best to give you results. That phase didn't last.

The inevitable question: How will AI responses get commercialized? Right now, models are focused on accuracy and capability. But as open-source alternatives like Mistral and Stable Diffusion chase closed-source leaders, competitive dynamics will shift. Maybe what matters most won't be the best answer, but the smallest model that runs locally on your phone.

Enterprise adoption remains experimental. Customers use AI for small parts of workflows—concepting in 3D asset generation, for example—but haven't moved their full process over. The next phase is tools and products that integrate AI as a feature, not a product. Just like chat and messaging in mobile apps went from "we're a chat-based travel agent" to table stakes.

I think everyone's experimenting with AI right now, but I think we'll see it more like a feature. The way that we talk about chat and messaging in mobile apps as an example.

— Andrew Chen

Video generation is the most exciting near-term breakthrough. Today's models produce 2-3 second idle animations. Soon, you'll be able to create compelling 5-minute YouTube Shorts based on dialogue, scenes, and panning shots. The feedback loop is powerful: AI outputs are visual and creative, making them perfect for TikTok, Reels, and short-form content streams.

The Product Manager Role Is About to Compress

The ratio of engineers and designers to product managers is going to shift dramatically. Some companies reached 3:1 ratios of engineers to PMs. Chen expects a return to 10:1, 12:1, or even 20:1 as AI tools create leverage and organizations flatten.

The shift is already happening. CEOs have realized that a flat org of five teams is far more efficient than five-deep hierarchies of managers managing managers. The future isn't fewer PMs in absolute terms—it's fewer PMs per engineer, and a rise in very senior IC principal PM roles for critical projects.

Instead of hiring four product managers and giving someone a director title to manage them, companies will trust a director-level person with a team of five engineers to build directly. Taste and execution will matter more than managing personalities. One theory: the PM title has always been tech's catch-all for business things engineers don't want to deal with. In the early years, it meant Gantt charts and project management. Then it evolved into UX and feature prioritization. Then growth metrics became non-negotiable—now every PM needs to understand D30, DAU/MAU, and retention curves.

What happens when no-code tools let PMs build signup funnels, notification systems, and conversion flows without engineers? The CMO who once needed an army of people to execute a marketing website can now direct creative, analyze business impact, and build assets solo. The most important work becomes creative direction and strategic analysis—not coordination.

The PM role isn't disappearing. It's compressing into higher-leverage, more senior positions where one person can do what used to take a team.