By Sagar Shankaran, Founder of CallSphere
Amazon expands Bedrock Agents with native multi-agent orchestration, an extensive tool marketplace, and enterprise-grade governance for building production AI agent systems on AWS.
Key takeaways
Amazon Web Services has announced a sweeping expansion of its Bedrock Agents platform, introducing native multi-agent orchestration, a marketplace of over 40 pre-built tool integrations, and enterprise-grade governance controls. The updates, announced at AWS AI Summit on March 10, position Bedrock Agents as a comprehensive platform for building, deploying, and managing production AI agent systems at enterprise scale.
Bedrock Agents, originally launched in 2024 as a managed service for building AI agents on AWS, has evolved from a relatively simple retrieval-augmented generation (RAG) tool into a full-featured agent orchestration platform. The March 2026 update represents the largest single feature release in the platform's history and reflects AWS's recognition that AI agents — not standalone model inference — are becoming the primary way enterprises consume generative AI.
The headline feature is native support for multi-agent workflows, where multiple specialized agents collaborate to handle complex tasks. AWS has implemented three orchestration patterns:
sequenceDiagram
autonumber
participant Caller as Caller
participant Agent as CallSphere Agent
participant API as CRM API
participant DB as CRM Database
participant Webhook as Webhook Listener
Caller->>Agent: Inbound call begins
Agent->>Agent: STT plus intent detection
Agent->>API: Lookup contact by phone
API->>DB: Read contact record
DB-->>API: Contact and history
API-->>Agent: Personalized context
Agent->>API: Create call activity
Agent->>API: Update deal stage
API->>Webhook: Outbound webhook fires
Webhook-->>Agent: Confirmed
Agent->>Caller: Spoken confirmation
A supervisor agent receives incoming requests, decomposes them into sub-tasks, delegates each sub-task to an appropriate worker agent, monitors execution, and synthesizes results. The supervisor agent maintains a task graph that tracks dependencies, handles failures, and ensures all sub-tasks complete successfully before returning a final result.
This pattern is designed for complex workflows like customer onboarding (where different agents handle identity verification, account creation, regulatory compliance checks, and welcome communications) or procurement processes (where agents manage vendor evaluation, price negotiation, contract review, and purchase order creation).
Agents are arranged in a pipeline where the output of one agent becomes the input of the next. Each agent in the chain can transform, enrich, or act on the data before passing it forward. This pattern is effective for data processing workflows like content moderation (classification agent, policy evaluation agent, action agent) or lead qualification (data enrichment agent, scoring agent, routing agent).
Hear it before you finish reading
Talk to a live CallSphere AI voice agent in your browser — 60 seconds, no signup.
A coordinator agent distributes independent sub-tasks to multiple worker agents that execute concurrently. Results are collected and merged when all workers complete. This pattern is useful for scenarios like competitive analysis (where agents simultaneously research different competitors) or multi-channel notification (where agents handle email, SMS, push, and in-app notifications in parallel).
AWS has implemented these patterns as managed infrastructure, handling the complexities of state management, error recovery, timeout handling, and retry logic that would otherwise require significant custom engineering.
Perhaps more practically significant than multi-agent orchestration is the new Bedrock Agent Tool Marketplace. AWS has partnered with over 40 technology providers to offer pre-built, managed tool integrations that agents can use out of the box:
Business Applications: Salesforce, HubSpot, ServiceNow, Zendesk, Jira, Confluence, Slack, Microsoft Teams Data and Analytics: Snowflake, Databricks, Tableau, Looker, BigQuery (via cross-cloud connector) Developer Tools: GitHub, GitLab, PagerDuty, Datadog, Splunk Communication: Twilio (voice, SMS, WhatsApp), SendGrid, Mailchimp Commerce: Shopify, Stripe, Square, PayPal Document Processing: Adobe Document Cloud, DocuSign, Notarize
Each tool integration includes:
"Before the marketplace, connecting an agent to Salesforce required writing custom integration code, managing OAuth tokens, handling rate limits, and dealing with API versioning," explained Swami Sivasubramanian, VP of AI and Data at AWS. "Now it is a single configuration step. Select Salesforce, authorize access, and your agent can immediately read and write Salesforce data."
The governance features in this release reflect feedback from AWS's largest enterprise customers, who need fine-grained control over what AI agents can do in production environments:
A new policy engine allows administrators to define rules that govern agent behavior at the organizational level. Policies can specify:
Still reading? Stop comparing — try CallSphere live.
CallSphere ships complete AI voice agents per industry — 14 tools for healthcare, 10 agents for real estate, 4 specialists for salons. See how it actually handles a call before you book a demo.
Bedrock Agents now integrates directly with Amazon Bedrock Guardrails, providing content filtering, topic avoidance, and personally identifiable information (PII) detection and redaction across all agent interactions. Guardrails apply to both the agent's inputs and outputs, ensuring that sensitive information is never exposed regardless of the conversation path.
Every agent interaction generates a detailed audit record that includes the full conversation history, all tool invocations with parameters and results, reasoning traces showing the agent's decision process, cost breakdown by model inference, tool usage, and orchestration overhead, and performance metrics including latency, token usage, and error rates.
Audit records are automatically published to Amazon CloudWatch and can be exported to Amazon S3 for long-term retention, compliance analysis, or integration with security information and event management (SIEM) systems.
AWS has introduced a simplified pricing model for Bedrock Agents that consolidates the previously separate charges for model inference, knowledge base queries, and action group executions:
AWS estimates that a typical enterprise agent handling customer service inquiries costs between $0.05 and $0.15 per conversation, depending on complexity and the number of tool invocations required.
AWS reports that over 15,000 organizations are now using Bedrock Agents in production, up from 3,000 at the time of the last AWS re:Invent conference. Key deployments include:
"The multi-agent capabilities were the tipping point for us," said the CTO of a major insurance company. "Our claims processing workflow involves seven distinct steps, each requiring different data sources and decision criteria. With multi-agent orchestration, we decompose that into specialized agents that each handle one step, coordinated by a supervisor. It is far more maintainable and reliable than trying to build a single monolithic agent."
The Bedrock Agents expansion intensifies competition with Google Cloud's Vertex AI Agent Builder and Microsoft's Azure AI Agent Service. AWS's primary differentiator is the breadth of its tool marketplace and the depth of integration with the broader AWS ecosystem. Organizations already invested in AWS infrastructure can deploy agents that natively access DynamoDB, SQS, Lambda, Step Functions, and dozens of other AWS services without additional integration work.
Written by
Sagar Shankaran· Founder, CallSphere
LinkedInSagar Shankaran is the founder of CallSphere, where he builds production AI voice and chat agents deployed across healthcare, hospitality, real estate, and home services. He writes about agentic AI, LLM engineering, and shipping voice agents that handle real calls in production.
See how AI voice agents work for your industry. Live demo available -- no signup required.
How to design a multi-agent system using MCP for tools and A2A for cross-vendor coordination, with a CallSphere voice agent as a participating node.
A2A is the open standard for agent-to-agent coordination. Here is how the Agent Card JSON works, how discovery happens, and what to publish.
A2A unlocks cross-vendor agent coordination, but most enterprise voice/chat workloads still ship faster on a single-vendor stack. Here is how to choose.
AWS Multi-Agent Orchestrator ships supervisor routing, classifier, and shared memory. How to compose a customer-support agent team on Bedrock that scales cleanly.
AWS HealthScribe became the open scribe layer EHR vendors built on top of in 2026. Here's the API surface, the per-encounter pricing, the BAA terms.
AWS Trainium 2 supply caught up with demand in April 2026, prompting a re-set of EC2 Trn2 instance pricing and a fresh push into mid-market AI workloads.
© 2026 CallSphere Inc. All rights reserved.
Made within San Francisco
Watch how CallSphere handles real customer calls, schedules appointments, and processes payments — live.
Try Live DemoBook a DemoCalculate Your ROI