The weekly drop. Sharp, short, accurate. Subscribe to get this in your inbox every Friday.
Every week the AI press produces approximately 4,000 articles that can be summarised as “AI is either going to save us or kill us, here are five quotes from people who haven’t shipped anything.” I read them so you don’t have to. (This is either an act of service or a character flaw. Jury’s out.)
Here’s what actually mattered.
The One Thing You Should Have Noticed
The real story this week wasn’t any single model release or funding round. It was the gap between what AI companies are announcing and what enterprise buyers are actually deploying.
Every major analyst firm is revising their enterprise AI adoption numbers downward for year one and upward for years three through five. Translation: companies bought the story in 2023–2024, hit the wall of actual implementation complexity in 2025, and are now figuring out what they actually need to do to make this work.
This is healthy. The hype cycle is compressing into something more useful. Companies that did the boring work of understanding their data infrastructure, their process documentation, and their change management requirements are starting to see real ROI. Companies that bought an enterprise AI platform because the board asked about AI strategy are extending their contracts while quietly not deploying anything.
The winners in the next eighteen months are not the companies with the biggest AI budgets. They’re the companies where someone in operations actually understood what needed to change before the software arrived.
Three Things the Coverage Got Wrong This Week
1. “[Model X] will replace [profession Y].”
It won’t. Or at least not in the way described. The pattern holds: volume work at low-to-medium complexity, yes. Judgment under uncertainty with real accountability, no. If you’re reading a headline that says AI will replace [profession], divide the claimed timeline by three and add “at the junior level” to the job title. Now it’s probably accurate.
2. “OpenAI’s valuation means AI is overvalued.”
Valuation of a private company in a winner-take-most market where the underlying capability is still improving exponentially tells you almost nothing about whether AI is overvalued as a technology. These are different questions. The press treats them as the same question constantly. (It’s like saying Nvidia’s P/E ratio means electricity is overpriced. No.)
3. “AI is hitting a wall.”
The “AI scaling laws are ending” narrative keeps cycling back. It keeps being premature. The honest answer is: we don’t know. Neither does anyone writing that headline with certainty in either direction.
The Number Worth Knowing
€847 billion. The estimated total addressable market for enterprise AI software by 2030 according to IDC’s latest forecast. For context: the entire global ERP software market — SAP, Oracle, the lot — is roughly €50 billion annually.
This is either the most important technology market in history or the most overhyped forecast in decades. Probably both, with the actual number landing somewhere between them and arriving three years later than the projections say.
What You Should Actually Do
If you run a business and you’re trying to figure out where to focus on AI: ignore the weekly headlines. They’re mostly noise. Focus on one question instead — which task in your operation do people spend the most time on that produces the least differentiated output?
That’s your entry point. Not “AI strategy.” Not “digital transformation.” One task. Automate it. Measure the result. Then do the next one.
The companies winning at AI are not the ones with the most sophisticated AI vision. They’re the ones with the most boring, systematic implementation practices.
Back next Friday with more things the internet got wrong.
Stay sharp.
The best AI analysis, straight to your inbox.