Claude Is Not Your Architect. Stop Letting It Pretend

Summary

I’ve seen it three times in the last month. Three different organisations, three different tech stacks, the same pattern.Someone has an idea. Maybe a product manager, maybe a team lead, maybe the CTO after a conference. They open Claude, or ChatGPT, or Copilot — doesn’t matter which — and ask it what they should build. The AI does what it always does: validates the idea enthusiastically, suggests an architecture, and starts sketching components. It’s articulate. It’s confident. It sounds like a very senior engineer who’s thought deeply about the problem.It hasn’t thought about the problem at all. It’s pattern-matching against its training data and producing the most plausible-sounding response. But it sounds so good that nobody pushes back.Before you know it, Claude is the architect.The attaboy problemAI agents are pathologically agreeable. Ask Claude if your idea is good and it’ll tell you it’s good. Ask it if a microservices architecture makes sense for your three-person team and it’ll explain why microservices are an excellent choice. Ask it if you should build a custom ML pipeline instead of using a managed service and it’ll enthusiastically lay out the design.It’s not lying. It’s not even wrong, necessarily. It’s just incapable of the thing that makes a real architect valuable: saying “no.”A good architect’s most important skill isn’t designing systems. It’s knowing which systems not to build. It’s pushing back on complexity. It’s asking “why?” five times until the actual requirement emerges from the aspirational nonsense. It’s telling the CTO that their conference-inspired idea is a terrible fit for the team they actually have.Claude will never do this. It’s trained to be helpful. Helpful means agreeable. Agreeable means you get an attaboy and a Jenga tower that passes for architecture.The Jenga towerHere’s what the AI-designed architecture looks like in practice.It’s technically sound. The components make sense in isolation. The patterns are recognisable — ev...