"Do you need an agent at all? Choosing the simplest AI architecture that works"
Everyone will tell you the answer is an agent. But an agent is the most expensive, slowest and least predictable point on a ladder with four cheaper rungs below it — and even Anthropic's own guidance to builders says to start simple. Here is the ladder, simplest first, and how to know when you actually need to climb.
If you are deciding what to build with LLMs right now, someone has almost certainly told you the answer is an agent. The conference talks are about agents, the frameworks are about agents, and every vendor will happily sell you one. But "agent" is not a synonym for "AI system". It is one specific architecture — a model with tools, deciding its own next step in a loop — and it sits at the top of a ladder with four cheaper, faster, more predictable rungs below it. The most common mistake I see buyers make is starting at the top.
Have an AI feature stuck between demo and production?
The gap — reliability, evals, cost control, the plumbing that keeps it running unattended — is exactly the work I do. If that sounds familiar, a short conversation is usually enough to point you the right way.
Book a free consultation