Companies Test AI Systems That Click, Type and Complete Tasks

Companies are testing AI systems that can click, type, scroll and complete tasks across desktop, mobile and web interfaces. These tools are being positioned as operators that move through software rather than as chatbots that only answer questions.

2025 Work Trend Index reported that 82% of leaders expect to use digital labor to expand workforce capacity over the next 12 to 18 months. The same index found 80% of the global workforce reporting insufficient time or energy to complete their jobs.

McKinsey’s 2025 State of AI stated that 62% of organizations are experimenting with AI agents and 23% are scaling such systems in at least one part of their operations.

Anthropic said companies including Asana, Canva, DoorDash, Replit and The Browser Company have begun exploring Claude’s computer-use capability for tasks involving dozens or hundreds of steps. Replit used the system to evaluate apps during development.

5 Computer Use, a model designed for agents that interact with user interfaces by clicking, typing and scrolling. Baidu launched a family of similar agents in March 2026 for desktop, mobile, cloud and smart-home workflows.

The article notes that these systems are managed through access controls and observation rather than through employment policies. It identifies risks including over-permissioning, unclear accountability when actions go wrong, and the tendency to project human qualities onto software.

A recommended approach is to assign narrower roles, separate reversible from irreversible actions, and measure performance by friction reduction and auditability rather than by how human the system appears.