Skip to content
Free since 2015. Built for the engineers who keep things running.

cat /articles/does-your-ai-agent-need-an-llm-to-decide-what-to-do-next.md

AI & MLOps

Does your AI agent need an LLM to decide what to do next?

Small decisions inside AI agents, like whether to search memory or how much a retrieved result matters, may not need a full language model. AWS engineer Mike Chambers describes decider models as a lighter alternative.

The New Stack20 Nov 2026 · 4 min read · summary

article --summary

The summary

This New Stack piece looks at decider models, specialized AI models that pick from a fixed set of options or score inputs and return calibrated probabilities instead of generating free-form text. Readers will learn why these models can be faster and cheaper than a general LLM for routine judgments that happen on every request. The article draws on an interview with Mike Chambers, a senior AI developer experience engineer at AWS, who compares a decider to an agent-powered if statement. He describes two experiments with the open-source Strands framework and Amazon Bedrock AgentCore Memory: using a decider to skip memory lookups for simple greetings, and using one to score retrieved memories for relevance before they reach the model. The piece also covers the limits. A numeric score still has to be tested for reliability, and Chambers argues that higher-stakes actions, such as spending money, need checks outside the agent's loop and human approval for decisions that are hard to undo.

Editorial note: This is a summary of an article first published by The New Stack. The full piece is on their site.

Reading now

Topic
AI & MLOps
Source
The New Stack
Published
20 Nov 2026
Format
Curated summary

Have a take on this?

Write it up and submit it. It goes on the list once an editor approves it.

Submit an article

articles --latest

Latest articles

New from the blogs we follow and the community.

All articles

newsletter --weekly

DevOps Weekly, minus the noise.

The best reads, new jobs and upcoming events, in one email a week.