Agents in the Wild

AWS Sales put 20+ sales agents behind one assistant, Field Advisor, with 120K+ prompts

AWS built Field Advisor, one chat assistant that routes sales reps to more than 20 specialized agents. Reps can save up to 2 hours a week, the team says. It asks approval before CRM edits.

Organization
Amazon Web Services (Sales)
Industry
Cloud computing
When
May 2026
Evidence
Company reported
Agents and tools
Amazon Bedrock AgentCoreMCPSlackCRM

Reported outcome

More than 120K prompts submitted since launch; up to 2 hours a week saved per rep in the human-in-the-loop workflow; 41% lower latency; seven AWS accounts consolidated into one runtime.

  • 120K+ (company)
    Prompts since launch
  • up to 2 hours/week (company)
    Time saved per rep
  • -41% vs. previous setup (company)
    Latency
  • 20+ (company)
    Specialized agents behind it

Caveats

Stated by the company that uses the agent (blog, earnings call, interview). Not independently audited.

  • AWS wrote this about its own internal tool and sells the platform it runs on. No outside check.
  • 'Up to 2 hours' is an upper bound and covers only the human-in-the-loop part.
  • The 41% latency gain has no absolute numbers. The 15-minute and 450-account examples are single anecdotes.
  • The post says quality metrics are tracked but publishes no scores.

What the agent does

AWS Sales had deployed more than 20 domain-specific agents. Reps had to know which one to ask. Field Advisor is one interface that routes each request to the right agent. It sits in the CRM and Slack, remembers past chats and preferences, pulls from internal knowledge bases and sends proactive Slack alerts.

What happened

The team moved the agents from seven separate AWS accounts into one AgentCore runtime and onboarded more than 20 MCP tools. Before any CRM change, the assistant asks the rep to approve it.

What to copy

  • When agents multiply, add a router in front instead of asking people to pick.
  • Put the assistant where reps already work.
  • Require approval before writes to the CRM.

Starter prompt

Written by us, not by AWS.

You are a router for a sales team. Agents available: {agent_list_with_descriptions}
Rep request: {request}

Pick the one best agent. If two are needed, give the order.
If the request would change CRM data, say so and ask the rep to approve first.
If no agent fits, say so. Do not guess.

Sources

  1. Powering agentic AI sales strategy with Amazon Bedrock AgentCore — AWS Machine Learning Blog, May 27, 2026

Added Oct 26, 2026. Spotted an error? Send a correction.

Get new use cases weekly

One short email. Sources included.

Related use cases