{
  "id": 20838470,
  "title": "Does your AI agent need an LLM to decide what to do next?",
  "url": "https://urgent.news/2026/11/20/does-your-ai-agent-need-an-llm-to-decide-what-to-do-next",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-20T01:37:32.000Z",
  "source": {
    "name": "The New Stack",
    "slug": "the-new-stack",
    "url": "https://thenewstack.io/aws-decider-models-agents/"
  },
  "original_language": "en",
  "account": "In a conversation at AGNTCon + MCPCon North America 2026 in San Jose, California, Mike Chambers, a senior AI DevEx engineer at AWS, discussed the potential of decider models in AI agents. These models, also known as decision models or System One models, help make small judgments that accumulate within an agent. Decider models return calibrated probabilities and confidence scores without generating free-form text, making them useful for quick, low-latency decisions. Unlike traditional classifiers, these models can be applied to new tasks without prior training using general-purpose pretraining and natural language input. This allows developers to test the model's output before incorporating it into the agent's workflow. Chambers demonstrated the concept using the open-source Strands framework and Amazon Bedrock AgentCore Memory. He placed the Strands Decider in front of the memory lookup to determine if an incoming message required fetching data from past conversations. If the greeting did not need context from previous conversations, it could proceed directly. However, if the request relied on prior information, the decider would trigger a search. Similarly, after retrieval, the decider could score each result for relevance before including it in the agent's context. This process could save service calls and reduce the amount of extraneous information the agent had to process. While the decider's output is a predictable JSON field, testing is necessary to ensure its reliability. Chambers acknowledges that these models are still experimental and that the stakes could be higher in certain scenarios, such as when an agent is making financial decisions. In such cases, he recommends having additional checks outside the agent's loop and requiring human approval for irreversible decisions.",
  "summary": "Say “good morning” to an AI agent, and it might search its long-term memory before saying hello back. Somewhere in The post Does your AI agent need an LLM to decide what to do next? appeared first on The New Stack .",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}