• 100 School
  • Posts
  • đź’Ż Why your team isn't seeing results from AI

đź’Ż Why your team isn't seeing results from AI

Plus: Anthropic's scariest model yet, 93% of teams are running faster on the same treadmill, and the 4 stories from this week that explain both.

Everyone is lying about AI right now and I'm tired of it.

Not maliciously but optimistically. The LinkedIn posts (AI transformed our workflow). The vendor decks (10x productivity). The all-hands slides (we're an AI-first company now). I read this stuff every day and then I have conversations with ops leads, L&D heads, and founders; and something doesn’t add up.

This week, I finally figured out why.

Starting with the fact that on Wednesday, Anthropic published a post about a new model called Claude Mythos that is so capable at finding security holes they decided not to release it, and built a whole safety initiative around it called Project Glasswing.

That's where we are in April 2026. Anthropic is chaining up models for being too smart, and the rest of us are still trying to convince Claude to format a table without arguing about it.

And four stories from this week explain exactly why we’re seeing this distance👇

Window into the Future đź”® 

The first one came from Every. Their COO, Brandon Gell, used to spend the first two hours of every Monday cross-referencing strategy docs against launch calendars to figure out what the team should work on. Three years of Mondays, gone the same way. Now he asks an agent named Anton, sitting in their Slack, that reads the same docs in two seconds and posts a prioritised summary to the team. Brandon's Mondays start at 9am instead of 7.

Anton's daily Monday broadcast inside Every's Slack.
Screenshot from Katie Parrott's full story in Every's Source Code

Brandon's job didn't get automated. It got redesigned when he stopped asking "how do I do this faster?" and started asking "why am I doing this at all?"

That same question is what Dr. Philippa Hardman wrote about this week in a totally different industry. About 95% of corporate L&D teams right now are using AI to build courses faster. The other 5% have stopped and asked whether the course was the right answer in the first place. Rolls-Royce ran a controlled pilot replacing engine training courses with an in-workflow AI coach, and formal training hours dropped 35% while issue resolution sped up 12%. The course wasn't improved but it was replaced by something it was always a poor substitute for.

Then Unframe surveyed 255 companies and put a number on it. AI is now turning the 5-day workweek into 6 days of output. People are saving 2 to 6 hours a week. By every measure of adoption, it's working. And yet only 7% of those companies are seeing it land on the income statement. Most of us are running faster on the same treadmill.

The fourth story is mine. On Tuesday I had to make a call about rebuilding the 100 School website, and I was torn between Webflow (safe, predictable) and Claude (riskier, weirder, no muscle memory for it). I picked Claude because if a company is teaching other companies how to work with AI, you should be able to tell from the outside.

None of us did anything technically impressive. We just stopped using AI as a faster version of the old work and started letting it change what the work is.

This is what I was getting at last week when I wrote about doubling your team's AI output without buying anything new. I think I undersold it. The real unlock is noticing that the tools you already have are waiting for you to ask a different question.

Three things keep coming up in every honest conversation I have with teams right now. The skills people need are different. The experience of doing the work is different. And the way you measure whether any of it is working is completely different. That's the work we're building 100 School around because every team I talk to has the tools, the budget, and the buy-in. What they don't have is the new shape of the work.

That's the thing worth figuring out right now. And it's not on LinkedIn yet.

Before you go ✌️

A real question for you this week, because I'm trying to get smarter on it myself:

You've been using AI for months now. What's one thing on your calendar this week that should look different by now and somehow doesn't?

Hit reply. I read every one.

See you next Sunday!

Max 

P.S. Want to make your team & company AI-first? Let us help here.