This week I've been getting feedback on an AI adoption framework we're building. What keeps coming up in every conversation is how much of the challenge isn't technical. It's organizational. Who decides what to pursue. Who's accountable when something goes wrong. Who's measuring whether any of it worked.
Also in this issue:
- This Week — CEOs grabbed the AI wheel but need a co-pilot (BCG: half believe job stability depends on getting AI right). 80% of agentic AI work is data engineering, not the AI (MIT Sloan). Microsoft warns ungoverned agents could become corporate "double agents." AI readiness is becoming a financing requirement, not just an operational one.
- Adoption at Machine Speed — The main essay. Define what AI means for your company first (employee tools, internal builds, vendor platforms are three different things). Build a decision process, not a strategy deck. Start light, earn complexity. The three walls that block scaling: technical, legal, organizational. BCG's CEO archetypes: 70% are Pragmatists who want to move but won't move without proof.
- The Wire — Only 18% of professional services firms measure AI ROI (Thomson Reuters). Microsoft Copilot ignored sensitivity labels twice in eight months. OpenAI acquired Promptfoo, the red-teaming tool used by 25%+ of Fortune 500. Shadow AI agents are already on your employees' laptops.
- Quanta Lab — The "harness" matters more than the model. Why AutoGPT failed and today's agents work: planning tools, sub-agents, file system access, and system prompts that are specifications, not prompts.
- After Hours — 2001: A Space Odyssey re-read. HAL doesn't go evil. He's given contradictory instructions and his breakdown is the logical result. That felt like science fiction in 1968. It feels like a Tuesday in 2026.