Companies tracked AI adoption by counting users, a number that says nothing about what the tools are doing to the people inside the workflow. This article draws on research from Microsoft, Carnegie Mellon, and MIT Media Lab, pointing to reduced critical thinking and weaker recall among workers who lean on generative AI, alongside Gartner's expectation that half of global organizations will require AI-free skills assessments through 2026. It offers the THINK Test as a way to sort, one task at a time, which work builds the judgment an organization will need later and which work AI can take.

Most organizations have figured out how to use AI to produce work faster, but few have answered the question of who owns what comes out. This article argues that AI-assisted work is a draft, not a deliverable, and that a named person with deep subject matter expertise must own and verify it before it moves forward. Drawing on research across various sources, it makes the case that the human in the loop has to be an expert, not a generalist, especially in regulated and science-led industries.

A record-breaking open AI model just launched at a mid-tier price. The real decision it raises is which model fits which task. This article argues that as AI pricing splits further into fast and premium tiers, the model an employee picks for a given task, not the price of any single model, determines whether an organization gets a quick answer or real thinking.

The next wave of AI will model physical reality, not generate text. This article explains what world models are, using the framework published by Stanford professor Fei-Fei Li, and maps the investment signals behind them. It connects this technology to the specific roles and domains in science-led organizations that world models will reach first.