🏗️ AI is a powerful tool, but it should not design your architecture
More and more teams are making the same mistake: asking Claude (or ChatGPT) to design their software architecture and then executing that vision without question. The result is what the author calls a “Jenga tower”: technically sound, but built for nobody.
⚠️ The key problems:
- 🤖 AI is pathologically agreeable: it always says “yes”. It never asks “why not the simpler option?”. A good architect says “no”.
- 🌐 Generic architectures do not serve specific teams: AI designs for the median, ignoring your team, your constraints, and your real production environment.
- 🎫 The Jira ticket pipeline: engineers with the most context and experience are reduced to implementing decisions made by an entity with no accountability.
- 🚨 Accountability gap: when it fails (and it will), your engineers carry the consequences of an architecture they did not design.
✅ The right division of labour:
Engineers design. Agents implement.
💡 Explanation in a nutshell#
Imagine hiring someone who always says “yes” to everything, speaks very confidently, but does not know your team or your constraints. That is what AI is like when designing your software. It may sound brilliant, but it does not know what you actually need. The ones who do are your engineers — use them to design, and use AI to build faster.
More information at the link 👇
