AI changed my work in a way I’m still trying to understand.

Six months ago, agentic coding crossed a threshold for me. Since then, I’ve used AI to design landing pages, automate SEO reporting, build SEO testing workflows, and ship internal apps I would never have built before.

That shift has three consequences, as I outline in a Growth Memo episode:

First, software is not becoming worthless. Enterprise software is more protected than people think because it is not just code. It is integrations, uptime, permissions, procurement, legal review, change management, support, and accountability. Self-serve software has a harder road because simple workflows are getting easier to rebuild.

Second, distribution gets harder. Google keeps more users inside its own products, AI chatbots send little traffic out, social is flooded, and paid keeps getting more expensive. In that world, velocity matters more. Product matters more. Marketing and product development melt together.

Third, people become the limiting factor. AI can now produce more than we can review. The cost to automate keeps falling, but the cost to verify does not fall nearly as fast. That means tasks with clear, measurable outputs will automate first, while work that requires taste, strategy, market context, or domain judgment compounds more slowly.

The scarce skill is knowing what to build, what to kill, what to trust, and what the agent missed.