Starting today, customers can create nested environments within virtualized Amazon EC2 instances. Previously, customers could only create and manage virtual machines inside bare metal EC2 instances. With this launch, customers can create nested virtual machines by running KVM or Hyper-V on virtual EC2 instances. Customers can leverage this capability for use cases such as running emulators for mobile applications, simulating in-vehicle hardware for automobiles, and running Windows Subsystem for Linux on Windows workstations.
Starting today, customers can create nested environments within virtualized Amazon EC2 instances. Previously, customers could only create and manage virtual machines inside bare metal EC2 instances. With this launch, customers can create nested virtual machines by running KVM or Hyper-V on virtual EC2 instances. Customers can leverage this capability for use cases such as running emulators for mobile applications, simulating in-vehicle hardware for automobiles, and running Windows Subsystem for Linux on Windows workstations.
AWS Batch now provides Queue and Share Utilization Visibility, giving you insights into how your workloads are distributed across compute resources. This feature introduces queue utilization data in job queue snapshots, revealing compute capacity used by your first-in-first-out (FIFO) and fair share job queues, along with capacity consumption by individual fair share allocations. Additionally, the ListServiceJobs API now includes a scheduledAt timestamp for AWS Batch service jobs, allowing you to track when jobs are scheduled for execution.
Queue and Share Utilization Visibility helps you understand which fair-share allocations consume the most capacity and pinpoint the specific jobs driving resource consumption. You can monitor overall queue utilization and drill down into active shares to optimize resource distribution, or filter jobs by share identifier to analyze consumption patterns and scheduling behavior across your workloads.
You can access this feature using the GetJobQueueSnapshot, ListJobs, and ListServiceJobs APIs, or through the AWS Batch Management Console by navigating to your job queue details page and selecting the new Share Utilization tab. This feature is available today in all AWS Regions where AWS Batch is available. To learn more, visit the Job Queue Snapshot, List Jobs, and List Service Jobs pages of the AWS Batch API Reference Guide.
AWS Batch now provides Queue and Share Utilization Visibility, giving you insights into how your workloads are distributed across compute resources. This feature introduces queue utilization data in job queue snapshots, revealing compute capacity used by your first-in-first-out (FIFO) and fair share job queues, along with capacity consumption by individual fair share allocations. Additionally, the ListServiceJobs API now includes a scheduledAt timestamp for AWS Batch service jobs, allowing you to track when jobs are scheduled for execution. Queue and Share Utilization Visibility helps you understand which fair-share allocations consume the most capacity and pinpoint the specific jobs driving resource consumption. You can monitor overall queue utilization and drill down into active shares to optimize resource distribution, or filter jobs by share identifier to analyze consumption patterns and scheduling behavior across your workloads. You can access this feature using the GetJobQueueSnapshot, ListJobs, and ListServiceJobs APIs, or through the AWS Batch Management Console by navigating to your job queue details page and selecting the new Share Utilization tab. This feature is available today in all AWS Regions where AWS Batch is available. To learn more, visit the Job Queue Snapshot, List Jobs, and List Service Jobs pages of the AWS Batch API Reference Guide.
Amazon Connect now supports in-app notifications in the workspace header, visible from any page, so your team can stay informed without interrupting their workflow— whether configuring, analyzing data, or servicing customers. A notification icon appears in the header of every workspace page, with a badge indicating unread messages. Click the icon to view messages, access relevant resources through embedded links, and manage read/unread status—all without navigating away from your current task. For example, if all supervisors need to complete a certain training by end of week, a notification can be published to non-compliant users to remind them.
The new notification APIs enable you to programmatically send targeted messages to specific audiences within your organization, ensuring teams stay aware of urgent updates, policy changes, and action items requiring immediate attention. Amazon Connect will also leverage this capability to deliver system updates and important announcements.
In-app notifications are available in all AWS regions where Amazon Connect is available and offer public API and AWS CloudFormation support. To learn more about in-app notifications, see the Amazon Connect Administrator Guide. To learn more about Amazon Connect, please visit the Amazon Connect website.
Amazon Connect now supports in-app notifications in the workspace header, visible from any page, so your team can stay informed without interrupting their workflow— whether configuring, analyzing data, or servicing customers. A notification icon appears in the header of every workspace page, with a badge indicating unread messages. Click the icon to view messages, access relevant resources through embedded links, and manage read/unread status—all without navigating away from your current task. For example, if all supervisors need to complete a certain training by end of week, a notification can be published to non-compliant users to remind them. The new notification APIs enable you to programmatically send targeted messages to specific audiences within your organization, ensuring teams stay aware of urgent updates, policy changes, and action items requiring immediate attention. Amazon Connect will also leverage this capability to deliver system updates and important announcements. In-app notifications are available in all AWS regions where Amazon Connect is available and offer public API and AWS CloudFormation support. To learn more about in-app notifications, see the Amazon Connect Administrator Guide. To learn more about Amazon Connect, please visit the Amazon Connect website.
Amazon Connect now provides AI-powered Task overviews with suggested next actions so agents can understand work items faster and resolve them more quickly. For example, when an agent receives a Task to process a refund request submitted through an online form, Amazon Connect summarizes earlier activities such as verifying order details, checking return eligibility, and confirming the payment method, and then presents recommended next steps to complete the refund.
To enable this feature, add the Connect assistant flow block to your flows before a Task contact is assigned to your agent. You can guide the recommendations of your generative AI-powered Tasks assistant by adding knowledge bases.
This new feature is available in all AWS regions where Amazon Connect real time agent assistance is available. To learn more and get started, refer to the help documentation, pricing page, or visit the Amazon Connect website.
Amazon Connect now provides AI-powered Task overviews with suggested next actions so agents can understand work items faster and resolve them more quickly. For example, when an agent receives a Task to process a refund request submitted through an online form, Amazon Connect summarizes earlier activities such as verifying order details, checking return eligibility, and confirming the payment method, and then presents recommended next steps to complete the refund. To enable this feature, add the Connect assistant flow block to your flows before a Task contact is assigned to your agent. You can guide the recommendations of your generative AI-powered Tasks assistant by adding knowledge bases. This new feature is available in all AWS regions where Amazon Connect real time agent assistance is available. To learn more and get started, refer to the help documentation, pricing page, or visit the Amazon Connect website.
Amazon Relational Database Service (RDS) and Amazon Aurora now offer greater flexibility for restore operations to view and modify backup retention period and preferred backup window prior to and upon restoring database snapshots. The backup retention period lets you specify how many days backups are retained, while the preferred backup window allows you to set your desired backup schedule.
Previously, restored database instances and clusters inherited backup parameter values from snapshot metadata and could only be modified after restore was complete. This launch introduces two enhancements – you can now view the backup retention period and preferred backup window settings as part of automated backups and snapshots, providing visibility into backup configurations before initiating restore operation. Additionally, you can now specify or modify the backup retention period and preferred backup window when restoring database instances and clusters, eliminating the need to modify the instance or cluster after restoration.
These enhancements are available for all Amazon RDS database engines (MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, and DB2) and Amazon Aurora (MySQL-Compatible and PostgreSQL-Compatible editions) in all AWS commercial regions and AWS GovCloud (US) regions where RDS and Aurora are supported and respective database engines are available. You can use these features through the AWS Management Console, AWS Command Line Interface (CLI), and AWS SDKs at no additional cost. For more information, see Amazon RDS and Amazon Aurora User Guide.
Amazon Relational Database Service (RDS) and Amazon Aurora now offer greater flexibility for restore operations to view and modify backup retention period and preferred backup window prior to and upon restoring database snapshots. The backup retention period lets you specify how many days backups are retained, while the preferred backup window allows you to set your desired backup schedule. Previously, restored database instances and clusters inherited backup parameter values from snapshot metadata and could only be modified after restore was complete. This launch introduces two enhancements – you can now view the backup retention period and preferred backup window settings as part of automated backups and snapshots, providing visibility into backup configurations before initiating restore operation. Additionally, you can now specify or modify the backup retention period and preferred backup window when restoring database instances and clusters, eliminating the need to modify the instance or cluster after restoration. These enhancements are available for all Amazon RDS database engines (MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, and DB2) and Amazon Aurora (MySQL-Compatible and PostgreSQL-Compatible editions) in all AWS commercial regions and AWS GovCloud (US) regions where RDS and Aurora are supported and respective database engines are available. You can use these features through the AWS Management Console, AWS Command Line Interface (CLI), and AWS SDKs at no additional cost. For more information, see Amazon RDS and Amazon Aurora User Guide.
Amazon Aurora DSQL now supports identity columns and sequence objects enabling developers to generate auto-incrementing, integer-based IDs directly in the database using familiar SQL patterns.
This launch simplifies migrations of existing PostgreSQL applications and supports development of new workloads that rely on database-managed integer identifiers. Developers can create compact, human-readable IDs, such as order numbers, account IDs, or operational references without custom ID generation logic in application code or middleware.
Amazon Aurora DSQL now supports identity columns and sequence objects enabling developers to generate auto-incrementing, integer-based IDs directly in the database using familiar SQL patterns. This launch simplifies migrations of existing PostgreSQL applications and supports development of new workloads that rely on database-managed integer identifiers. Developers can create compact, human-readable IDs, such as order numbers, account IDs, or operational references without custom ID generation logic in application code or middleware.
AWS is announcing local support for the latest generation of x86-powered Amazon EC2 instances on second-generation AWS Outposts racks, including C8i compute-optimized instances, M8i general-purpose instances, and R8i memory-optimized instances. These new instances deliver 20% better performance and 2.5x more memory bandwidth compared to the C7i, M7i, and R7i instances on second-generation Outposts racks. In addition, C8i, M8i, and R8i instances on second-generation Outposts racks deliver 20% more compute capacity than C7i, M7i, and R7i instances within the same rack space and power draw, enabling better space and energy efficiency for your on-premises workloads.
C8i, M8i, and R8i instances on second-generation Outposts racks are powered by custom Intel Xeon 6 processors available only on AWS and are ideal for a broad range of on-premises workloads requiring enhanced performance, such as larger databases, more memory-intensive applications, advanced real-time big data analytics, high-performance video encoding and streaming, and CPU-based edge inference with more sophisticated machine learning (ML) models.
To learn more about second-generation Outposts racks, refer to the Outposts racks product pageand the user guide. For the most updated list of countries and territories and the AWS Regions where second-generation Outposts racks are supported, check out the Outposts rack FAQs page.
AWS is announcing local support for the latest generation of x86-powered Amazon EC2 instances on second-generation AWS Outposts racks, including C8i compute-optimized instances, M8i general-purpose instances, and R8i memory-optimized instances. These new instances deliver 20% better performance and 2.5x more memory bandwidth compared to the C7i, M7i, and R7i instances on second-generation Outposts racks. In addition, C8i, M8i, and R8i instances on second-generation Outposts racks deliver 20% more compute capacity than C7i, M7i, and R7i instances within the same rack space and power draw, enabling better space and energy efficiency for your on-premises workloads. C8i, M8i, and R8i instances on second-generation Outposts racks are powered by custom Intel Xeon 6 processors available only on AWS and are ideal for a broad range of on-premises workloads requiring enhanced performance, such as larger databases, more memory-intensive applications, advanced real-time big data analytics, high-performance video encoding and streaming, and CPU-based edge inference with more sophisticated machine learning (ML) models. To learn more about second-generation Outposts racks, refer to the Outposts racks product page and the user guide. For the most updated list of countries and territories and the AWS Regions where second-generation Outposts racks are supported, check out the Outposts rack FAQs page.
Amazon Bedrock is a fully managed service that provides secure, enterprise-grade access to high-performing foundation models from leading AI companies, enabling you to build and scale generative AI applications. Today, Amazon Bedrock announced support for the latest open-weight models in Asia Pacific (Sydney) using the bedrock-mantle endpoint. These include models from industry-leading providers, including DeepSeek, Google, MiniMax, Mistral, Moonshot AI, MiniMax, Nvidia, and OpenAI. The bedrock-mantle endpoint is powered by Project Mantle, a new distributed inference engine for large-scale machine learning model serving on Amazon Bedrock. Project Mantle simplifies and expedites onboarding of new models onto Amazon Bedrock. It provides highly performant and reliable serverless inference with sophisticated quality of service controls, unlocks higher default customer quotas with automated capacity management and unified pools, and delivers out-of-the-box compatibility with OpenAI API specifications.
To learn more and get started, visit the Amazon Bedrock console or the Amazon Bedrock service documentation. To get started with Amazon Bedrock OpenAI API-compatible service endpoints, visit the OpenAI API compatibility documentation.
Amazon Bedrock is a fully managed service that provides secure, enterprise-grade access to high-performing foundation models from leading AI companies, enabling you to build and scale generative AI applications. Today, Amazon Bedrock announced support for the latest open-weight models in Asia Pacific (Sydney) using the bedrock-mantle endpoint. These include models from industry-leading providers, including DeepSeek, Google, MiniMax, Mistral, Moonshot AI, MiniMax, Nvidia, and OpenAI. The bedrock-mantle endpoint is powered by Project Mantle, a new distributed inference engine for large-scale machine learning model serving on Amazon Bedrock. Project Mantle simplifies and expedites onboarding of new models onto Amazon Bedrock. It provides highly performant and reliable serverless inference with sophisticated quality of service controls, unlocks higher default customer quotas with automated capacity management and unified pools, and delivers out-of-the-box compatibility with OpenAI API specifications.
To learn more and get started, visit the Amazon Bedrock console or the Amazon Bedrock service documentation. To get started with Amazon Bedrock OpenAI API-compatible service endpoints, visit the OpenAI API compatibility documentation.
You can now create Amazon S3 Access Grants in the AWS Asia Pacific (Taipei) Region.
Amazon S3 Access Grants map identities in directories such as Microsoft Entra ID, or AWS Identity and Access Management (IAM) principals, to datasets in S3. This helps you manage data permissions at scale by automatically granting S3 access to end users based on their corporate identity.
Visit the AWS Region Table for complete regional availability information. To learn more about Amazon S3 Access Grants, visit our product page.
You can now create Amazon S3 Access Grants in the AWS Asia Pacific (Taipei) Region.
Amazon S3 Access Grants map identities in directories such as Microsoft Entra ID, or AWS Identity and Access Management (IAM) principals, to datasets in S3. This helps you manage data permissions at scale by automatically granting S3 access to end users based on their corporate identity.
Visit the AWS Region Table for complete regional availability information. To learn more about Amazon S3 Access Grants, visit our product page.
IA en el trabajo: ¿El futuro del comercio? Experiencia al alcance de su mano
La IA cambia la toma de decisiones, de las opciones de navegación a la aprobación de resultados, y rediseña la manera en que consumidores y productores se conectan.
Por: Jared Spataro, CMO de IA en el Trabajo de Microsoft.
Durante la mayor parte de mi carrera, la promesa definitoria de la tecnología era la información al alcance de la mano. El ordenador personal hacía que el conocimiento fuera accesible, internet lo hacía buscable, y el móvil lo metía en el bolsillo. Ese cambio transformó el comercio al facilitar que los consumidores descubrieran, compararan y realizaran transacciones.
También creó un nuevo problema. A medida que se amplía el acceso a la información, la toma de decisiones se convierte en el cuello de botella. Cualquiera que haya recurrido al Dr. Google para un diagnóstico conoce esa sensación: no se van con claridad, se van con una larga lista de posibilidades y muy poca orientación sobre cuáles son importantes.
La IA cambia el paradigma de la información al alcance de la mano a la experiencia al alcance de la mano: orientación, interpretación y contexto que les ayudan a decidir qué hacer a continuación.
Esto no se limitará a la sanidad o las compras. Se reflejará en finanzas, viajes, educación, atención al cliente, compras empresariales—en cualquier lugar donde haya un punto de decisión entre consumidores y productores.
El sistema de consumidor-productor está a punto de cambiar
Al más alto nivel, el comercio es un intercambio entre las entidades que producen y las que consumen. A veces es una persona que compra un producto. Otras veces es una empresa que compra servicios o un paciente que busca atención. El «qué» varía, pero la estructura es la misma: se necesita algo, se ofrece algo, y alguien tiene que decidir y actuar.
A nivel histórico, el puente entre ambos bandos ha sido la información. Los motores de búsqueda, los marketplaces, las herramientas comparativas y los feeds de recomendaciones ayudaron a los compradores a ver sus opciones. Pero son intermediarios de información. Presentan elecciones a la superficie; no llevan la carga del razonamiento—esa tarea queda en manos del humano.
En una economía mediada por agentes, ese puente se convierte en experiencia, con agentes en cada lado. Un agente del lado del consumidor representa la intención, el contexto y las restricciones. Un agente del lado del proveedor representa ofertas, políticas, capacidad y rendimiento. Esos agentes coordinan para reducir el conjunto de opciones viables y ayudar al consumidor a tomar la decisión final.
La sanidad hace que la diferencia sea evidente. Hoy en día, buscar síntomas en internet nos inunda con información, no con experiencia. Pronto, un agente de salud al consumidor hará preguntas de seguimiento, recopilará el historial médico relevante (con su permiso) y razonará el patrón como haría un clínico. Un agente del lado del proveedor comparará ese contexto con casos comparables, vías disponibles y restricciones como la cobertura y el momento.
Los humanos todavía tienen el poder de decisión, pero los agentes se encargan de la admisión y el razonamiento que determina hacia dónde va el paciente a continuación—en lugar de acudir a su médico de cabecera con una queja vaga, el paciente llega con una hipótesis más clara y una ruta sugerida hacia el nivel adecuado de atención.
Por qué la experiencia y no la personalización, es el verdadero avance
Es tentador describir este cambio como personalización, porque los agentes entenderán sus preferencias, su historial y sus limitaciones de una manera que se siente adaptada.
Pero la personalización es un resultado. El avance es la experiencia escalable. Estos sistemas pueden interpretar el contexto, plantear preguntas aclaratorias y aplicar el conocimiento del dominio para guiar las decisiones. Esa es la diferencia entre un sistema que los ayuda a navegar y un sistema que los ayuda a elegir.
También cambia el trabajo del lado del productor. Cuando la experiencia pasa a la capa de interacción, los productores deben empezar a diseñar para la interpretación de agentes. Eso significa hacer que lo que se ofrece sea fácil de entender, verificar y confiar para los agentes.
Una vez que los agentes hacen la clasificación, comparación y razonamiento a primera pasada, el centro de gravedad cambia.
Los compradores dedican menos tiempo a navegar por opciones y más tiempo a aprobar los resultados. Los productores dedican menos tiempo a optimizar la atención y más tiempo en optimizar para una ejecución fiable. Y los sistemas subyacentes al comercio evolucionan de estar diseñados para la navegación humana a estar diseñados para la coordinación de agentes.
Una nueva tensión: la brecha principal-agente
En la era de la información, el principal y el agente eran la misma entidad: el consumidor. Buscabas, interpretabas y elegías. No había distancia entre la intención y la acción. En la era del agente, eso cambia. El principal, ya sea una persona que compra o un líder que toma decisiones empresariales, establece el objetivo y las limitaciones. Después, un agente de IA gestiona el bucle de evaluación—para reducir opciones, sopesar compensaciones y recomendar un curso de acción.
Ese cambio genera velocidad y claridad, pero también introduce una nueva brecha. Una vez que un agente realiza el trabajo de decisión inicial, el riesgo pasa de «¿encontró las opciones correctas?» a «¿eligió los sacrificios correctos?» ¿Cómo saber que entendió lo que más importaba—y cómo corregir al agente cuando optimiza para un resultado equivocado?
No tenemos que resolver esa tensión de inmediato. Pero influirá en la confianza, el comportamiento del mercado y cómo los productores diseñan las ofertas en un mundo donde el comprador está cada vez más representado por un agente.
Lo que todo esto significa para los líderes empresariales
Si la experiencia se convierte en la nueva interfaz, el objetivo es sencillo para los productores: ser un negocio que los agentes puedan entender, confiar y recomendar.
Eso empieza por hacer que sus ofertas sean legibles: datos limpios, políticas claras y una estructura legible para una máquina. También requiere estrechar la conexión entre lo que se promete y lo que se entrega, porque los agentes evaluarán esa brecha más rápido que los humanos.
Por otro lado, cuando su negocio es el consumidor, asegúrense de que su agente de compras tenga acceso a todo el contexto posible sobre cómo se ve lo que es «bueno» para los parámetros que importan—precio, velocidad, fiabilidad, resultados—para que pueda hacer la comparación y su equipo pueda centrarse en revisar los sacrificios.
Las empresas que se adelantan no lo tomarán como un turno de marketing o compras. Lo tratarán como un turno operativo: diseñar para la evaluación de agentes y luego usar agentes para avanzar en las decisiones más rápido, con mejor juicio en el centro.
Para más información sobre la IA y el futuro del trabajo, suscríbanse a este boletín.
