Amazon Web Services has announced that its Amazon WorkSpaces service can now function as managed virtual desktops for artificial intelligence agents, a capability currently available in public preview. The development allows AI agents to authenticate through AWS Identity and Access Management and operate legacy desktop applications using computer vision and input simulation, bypassing the need for application programming interfaces.
What the New Feature Does
Amazon WorkSpaces, traditionally used to provision cloud-based virtual desktops for human users, has been extended to support AI agents as desktop operators. These agents log in via IAM roles, then interact with legacy software by visually analyzing on-screen elements and simulating mouse clicks, keystrokes, and other inputs. This approach enables automation of tasks in applications that lack modern API integration, such as older enterprise resource planning systems or custom internal tools.
The preview release does not require any modifications to the underlying desktop applications, making it potentially suitable for organizations that rely on software no longer supported by vendors or that cannot be easily updated.
Performance Benchmarks
Independent benchmarks from Reflex, a provider of AI agent testing tools, indicate that vision based agents consume significantly more computational tokens compared to API based agents. According to the benchmarks, vision agents use approximately 45 times more tokens than agents operating through APIs. Token consumption directly affects operating costs and latency in AI systems, as tokens represent the smallest units of data processed by large language models and vision models.
The higher token usage stems from the need to process and interpret visual data from desktop screens, whereas API agents rely on structured data exchanges. Reflex noted that while vision agents offer wider compatibility with legacy systems, the increased token usage may require organizations to plan for higher infrastructure spending.
Authentication and Security
Authentication for these AI agents is managed through AWS IAM, allowing administrators to apply the same access controls used for human users. This includes role based permissions, multi factor authentication policies, and audit logging. AWS has stated that no additional identity management systems are required, and agents do not store credentials locally.
The architecture also supports session isolation, meaning each agent operates within a dedicated virtual desktop environment, reducing the risk of cross tenant data exposure.
Industry observers note that the feature could help organizations automate repetitive tasks in legacy environments without rewriting software or deploying robotic process automation tools that require separate licensing and infrastructure.
Amazon WorkSpaces is available in multiple AWS regions globally. The AI agent capability is currently in public preview, meaning it is not yet recommended for production workloads. Pricing remains aligned with standard WorkSpaces billing based on compute and storage resources consumed, with additional costs incurred for any associated AI model usage.
AWS has not announced a general availability date for this feature. The company typically evaluates public preview feedback over several months before setting a production release timeline. Organizations interested in testing the capability can enable it through the AWS Management Console, with documentation available on the official AWS documentation site.
