AWS AI Updates: November 20, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. AWS Marketplace Adds Agentic Payments So AI Agents Can Buy Paid APIs Per Request Through AgentCore
AWS Marketplace now supports agentic payments over the x402 protocol, letting AI agents discover and buy select products and pay per request in US dollars through AgentCore payments, part of Amazon Bedrock AgentCore. Customers enable the AWS Marketplace payment connector, and agents find listings through the Marketplace connector in AgentCore Gateway. Charges settle on the existing AWS bill with no crypto wallet, spending limits and approved sellers can be set before agents transact, and transactions appear in AWS CloudTrail and AWS Cost Explorer. Sellers can charge agents per request by integrating the AWS Marketplace x402 SDK. It is available in all Regions where AgentCore payments is available. Source
2. Amazon Connect Launches Resolutions, an Agentic Solution for Refunds, Disputes, Fraud Investigations and Claims
Amazon Connect Customer announced general availability of Resolutions, an agentic operations solution that handles high-volume work end to end, including refund requests, transaction disputes, fraud investigations and insurance claims. Managers describe a process in natural language or upload standard operating procedures, and Resolutions generates a Work plan in minutes while asking clarifying questions about exceptions, decision rules and approval thresholds. It opens a Resolutions Case for each request, and each Step can run on prebuilt AI agents, agents configured in Amazon Connect Customer, or customers’ own agents, with human specialists handling approvals and exceptions. Resolutions verifies each outcome and records every action and decision for an auditable view of progress and service-level risk. Source
3. AWS Marketplace Agent Mode Now Assesses Seller Security, Buys and Deploys Supported Products
AWS Marketplace expanded agent mode, its conversational AI experience for procuring cloud solutions, so buyers can assess a seller’s security, buy, and deploy supported products in one conversation. Agent mode can request documents such as SOC 2 reports and return a criterion-by-criterion assessment against the Consensus Assessments Initiative Questionnaire. At purchase it explains pricing and terms, shows the organization’s purchase requirements, and completes the order after the buyer confirms. For supported products it reviews the AWS environment, proposes a deployment plan with an estimated monthly cost, and deploys after approval. It is available in all Regions where AWS Marketplace is available. Source
4. Bedrock AgentCore Runtime Instances Add Nested Virtualization for Agents That Run Emulators and VMs
Amazon Bedrock AgentCore runtime instances now support nested virtualization with Kernel-based Virtual Machine (KVM), so agents can run their own virtual machines and emulators inside a session. It serves agents that depend on tools such as mobile emulators or virtual machine sandboxes, for example running Android emulators to build and test mobile apps or starting lightweight VMs to isolate code. Users enable it when creating a Linux capacity provider, and it works on supported Intel-based instance types including the C7i, C8i, M7i, M8i, R7i and R8i families. It is available in all Regions where AgentCore runtime instances are available. Source
5. SageMaker AI Now Flags Slowing Responses, Idle GPUs and Missed Optimizations on Generative AI Endpoints
Amazon SageMaker AI now monitors response time and GPU use of generative AI inference endpoints by default and sends an AWS Health notification per endpoint when responses slow, GPUs sit mostly idle, or the configuration falls behind, such as an older serving container, missing speculative decoding or missing GPU kernel tuning. AWS states monitoring does not read customer requests or responses. Each notification links to the SageMaker AI Agentic Experience in SageMaker Studio, where users can run Generative AI Inference Recommendations that compare response time, traffic handled and cost per million tokens without touching the production endpoint. Monitoring can be turned off at any time. Source
6. AgentCore Identity Adds Private Certificate Authority Support for Custom OAuth 2.0 Credential Providers
Amazon Bedrock AgentCore Identity now supports private certificate authorities for custom OAuth 2.0 credential providers, so agents can connect to identity providers hosted in a VPC that use certificates issued by AWS Private CA, enterprise CAs or self-signed root CAs. Previously the identity provider still had to present a publicly trusted TLS certificate over the private connection through Amazon VPC Lattice. The CA certificate is stored in AWS Secrets Manager or Amazon S3 and referenced from the provider configuration, and it can be rotated without downtime by combining current and new certificates in one PEM file. It is available in all Regions where AgentCore Identity supports private identity provider connectivity. Source
7. AWS AppConfig Adds Configuration Profiles for AI Model Invocation Settings and Prompts
AWS AppConfig introduced two configuration profile types for generative AI applications: AI model invocation configurations and AI prompt configurations. The first stores the model identifier, inference parameters such as temperature and maximum tokens, and messages such as system prompts, with provider-specific options in custom parameters. The second stores prompts as separately versioned artifacts that can be updated on their own and reused across applications. Changes use AppConfig’s gradual rollouts and can roll back automatically when an alarm in CloudWatch, Datadog, New Relic or another tool fires. It is available at no additional cost in all Regions where AppConfig is available. Source
8. Amazon Polly Voice Cloning Reaches General Availability as a Gated Release
Amazon Polly announced general availability of voice cloning as a gated release. From a few minutes of recordings by a consenting speaker, users can create a cloned voice that runs on the Polly Generative engine, in English, Spanish, German, French, Italian, Portuguese and Hindi, with each cloned voice speaking the language it was recorded in. Cloned voices are encrypted and accessible only to their creator, and Polly does not keep the recordings after the voice is created or use them to train its models. Building a cloned voice is free, with pay-as-you-go pricing for synthesized speech, and access is requested through an AWS account team or an application form. Source
9. Amazon Connect Customer Adds Generative AI Post-Contact Summaries for Chat in the Agent Workspace
Amazon Connect Customer now shows agents a generative AI post-contact summary of each chat in the Contact Control Panel within seconds after the chat ends, covering what the customer needed, the actions taken and any open items. Summaries already existed for voice and email contacts through conversational analytics, and this launch adds chat. They follow the flow’s redaction settings and the same security profile permissions as voice summaries. Administrators who enabled voice summaries can turn on chat analytics with post-contact summary in the Set recording and analytics behavior block. It is available in 10 Regions including US East (N. Virginia), Europe (Frankfurt) and Asia Pacific (Tokyo). Source