Automating Windows Workflows with CUA
Client: An enterprise level software company.
The situation
Testing Windows applications and workflows is slow and expensive. Many teams rely on large manual QA efforts that take weeks, while traditional RPA is too brittle to handle real-world UI variation and branching paths.
At the same time, today’s consumer-facing computer-use models are often high-touch: a user has to babysit the agent through a chat interface, step-by-step, which limits scalability and repeatability.
What Delphi Intelligence built
Delphi Intelligence delivered a cloud-native automation service that runs Windows desktop workflow tests end-to-end using computer-use–enabled LLMs, without requiring users to “drive” the agent.
Users simply provide a workflow document or video. The system then:
Ingests and parses the workflow into structured steps
Uses AI subagents to plan, supervise, and validate execution
Orchestrates a fully agentic execution loop (see screen → decide → act → verify → continue) through the CUA agent
Provides a simple interface to run workflows on demand and review outcomes
This turns computer-use AI from a guided demo into an operational capability: the user describes what to test, and the system executes autonomously.
Business impact
Cycle time collapsed: representative app/workflow testing reduced from weeks to minutes for many scenarios
Broader practical coverage than rigid RPA, thanks to visual, context-aware execution across dynamic UI states
Repeatable runs on demand, enabling faster iteration for IT and product teams as environments change
Deliverables
Deployed cloud-native services and workflow-run interface
Runbooks and documentation for operational handoff
A functioning, end-to-end workflow automation system suitable for ongoing use
