AI Struggles

I received this message the other day, and I thought rather than replying directly, I would share my thoughts here because I’m seeing a lot of these questions.

The question:

“I’m currently seeing a lot of teams struggle keeping up with AI, building AI into their legacy stack, and being an AI-powered engineering team.

Curious to learn what others are doing and respective successes or challenges?”

Many engineering teams think they’re struggling with AI. What they are really struggling with is the system AI exposes.

AI makes every architectural shortcut, every unclear data contract, every brittle integration path instantly visible. It doesn’t create chaos. It amplifies whatever is already there.

The teams that are keeping up aren’t “more AI‑savvy.” They’ve rebuilt the system around AI‑level velocity:

• Event‑driven patterns instead of request/response bottlenecks
• Data contracts built for model stability
• CI/CD extended to include model evaluation and drift checks
• Developer workflows augmented with AI, not bolted on after the fact

AI isn’t a feature. It rewards teams who think in systems, not tools.

Curious where others see the real friction: architecture, data, or operating model?