Here are nine things across AI, product and company-building that got my attention this month. Plus, my perspective on what they actually mean for product leaders and people who want to become one.
We’ll start with AI-native operating models, then get into siloed AI workflows, what AI should and shouldn’t change about product management, the economics of great AI companies, infinite software and finite attention, the Rule of 40, faster pricing decisions… and some slop grenades.
Here we go.
1/ AI-native is becoming an operating model.
OpenAI published an interesting look at how companies like Clay, Basis, and Exa Labs are using AI internally. The stat that got my attention: Companies in the top 10% of enterprise AI usage now generate 8.3x more output tokens per active user than typical companies, up from 2.6x in January.
The more interesting part is what they’re doing with it. These companies are connecting agents to company context and tools, and putting them to work across onboarding, account management, developer integrations, and other real workflows.
The term “AI-native” is getting watered down by overuse, but in practical terms it comes down to what the product does AND how the company operates. From researching, designing, and shipping products to replacing SaaS tools, visualizing data, and automating internal workflows… both the breadth & depth of AI usage are rising quickly.
The real difference won’t be who uses the most AI. For a product leader, more AI tokens, credits, or outputs should be seen as little more than a vanity metric. What matters is whether the company, and its customers, are actually getting meaningfully better outcomes from them.
The companies that stand out won’t stop at chat boxes and AI-assisted workflows. They’ll change how the work gets done, and let automation take on more of the repetitive and execution-heavy parts.
Because here’s the thing: More companies are cancelling software and they’re building internally. And their customers are increasingly capable of doing the very same. Tokenmaxxing alone won’t solve that.
