The Last Moat
AI compressed every competitive advantage except one.
The CEO of Decagon — an AI customer service startup competing against Salesforce, Intercom, Sierra, and every enterprise’s own internal engineering team — wrote something unusually honest last week.
He said product moats are gone. AI tools have compressed the time to copy any feature to weeks. You ship something new; your competitor puts it on their roadmap. The advantage evaporates before it compounds.
“Ultimately, staying a few months ahead on product is an advantage, but not enough.”
So what’s left? When Decagon asks its customers why they chose them over the competition, the answer isn’t features. It’s: “We were most impressed with your team.”
The only thing AI can’t compress is the judgment of the person using it.
The same week, OpenAI launched its Partner Network — a structured program certifying systems integrators, consulting firms, and deployment specialists to co-sell and implement OpenAI products in enterprise environments. Tiers, specializations in Codex and agents and cybersecurity, a Forward Deployed Experts pilot for complex accounts.
Read that again. The most powerful AI company in the world just built a program to solve for the human layer.
Because OpenAI knows what Decagon knows: the model isn’t the bottleneck anymore. Enterprise AI fails on workflow redesign, change management, and the quality of the people deploying it. The capability is abundant. The judgment to apply it well is scarce.
A startup founder and the world’s most valuable AI company arrived at the same conclusion through opposite doors. One by losing deals until the pattern became obvious. The other by watching enterprise adoption stall despite having the best models on the market.
Everyone is racing to build AI. The winners are racing to build the teams that know how to use it.
This is the part of the AI story that doesn’t get written about much. The infrastructure war gets covered. The IPO race gets covered. The model benchmarks definitely get covered.
But the question of who actually deploys this stuff well — who sits across from a bank or an airline or a telecom and earns their trust over a multi-year relationship — that’s treated as an afterthought. A services problem. Not a strategy.
It is the strategy.
The moat isn’t the model. It never was. It’s the people who know what to do with it when the demo ends and the real work begins.
THE CHAOS TRANSLATOR • @currentlyted
Making complex systems legible. Subscribe for the full analysis.