UiPath Delegate Replays Work; Testers Set the Quality Bar

UiPath’s FUSION platform announcement on September 23, 2026 detailed how Delegate can turn recorded tester actions into reusable routines for operational work around software tests. The agent can gather context, move information between systems, update records, attach evidence and prepare summaries. Testers retain judgment, strategy, quality decisions and final sign-off.

That division of work comes with a release distinction. An Investing.com account of the Test Cloud announcement reports a Delegate public preview planned for October 2026. The described testing workflow is therefore a planned preview, even though the general-purpose Delegate agent has reached general availability.

What a recording gives Delegate

A tester can record repeatable actions and save the resulting workflow as a routine. Delegate can run that routine on demand, on a schedule or when a defined trigger occurs. This is the basis for the headline’s replay: an observed sequence becomes work the agent can repeat, rather than a new quality judgment inferred from the recording.

The surrounding tasks are concrete. Delegate can assemble context for a test run, transfer information across systems, change records, attach supporting evidence and prepare a summary for review. Those actions can reduce repeated coordination, but completing them does not establish that a test covered the right risk or that an apparent failure is acceptable. The distinction depends on the output: a record update is observable; a release decision requires an assessment of what the evidence means.

A routine can also preserve an obsolete step. If the application, record format or testing requirement changes, a successful replay may still produce an unsuitable result. That is an inference from how a recorded routine works, not a reported product failure. It explains why the evidence and summary need a reviewer who can judge them against the current test objective.

The handoff between agent and tester

The announced capabilities support a responsibility split across the testing workflow. This matrix interprets the product description; it does not claim that Delegate independently validates every result it handles.

Captured routine: A tester provides the example actions and decides which procedure is suitable for repetition. Delegate saves the workflow for reuse; the recording does not set testing strategy.
Trigger and execution: Delegate can run the routine when requested, on a schedule or from a defined trigger. A person remains responsible for deciding whether the routine fits the task and whether its scope remains current.
Evidence and records: Delegate can collect context, transfer information, update records and attach evidence. The tester decides whether the material is complete and relevant to the result under review.
Policy boundaries: Administrators can configure the agent’s available modes, context and permissions. The organization sets those boundaries for its own systems; a configurable control does not establish that any particular routine complies with its policy.
Quality decision and approval: Delegate can prepare a summary. People resolve ambiguity, judge the test outcome and give final sign-off.

The separation matters most when a routine appears to complete cleanly. An attached file or updated status shows that an action happened, but it may not show that the evidence supports the stated conclusion. The reviewer’s task is to connect the material to the test objective and decide whether an exception, another run or approval is warranted.

General availability and testing preview have different scopes

The September Delegate release notes date general availability of Delegate and Cartographer to September 22, 2026. They describe centrally managed Automation Ops settings for approval mode, execution mode, screen context, blocked apps and sites, and tool permissions. They also describe Delegate for Testing as a QA/Testing work profile under Test Cloud.

The release notes’ description of a testing profile and the planned Test Cloud public preview are not interchangeable availability statements. The former describes the generally available base agent and its documented role-specific experience; the latter concerns the testing-focused rollout. The public material does not clearly establish which testing functions every eligible customer can already access before that preview begins.

Those controls define what an agent may do, but they do not make it the authority on software quality. For a controlled workplace use, the responsibility split implies a defined routine, permitted actions, an output that can be reviewed and a person able to reject the result. This is an editorial deployment model drawn from the available controls and the stated human sign-off requirement, rather than a claim that UiPath has prescribed one universal approval process.

What the preview still needs to show

Delegate’s documented capability is to reuse recorded work and handle tasks around a test run while people retain the quality call. The testing preview will provide a clearer view of how that arrangement works when applications change, evidence is incomplete or a tester must intervene. The announcements do not provide measured reliability results for those situations.

For now, the confirmed boundary is specific: the general-purpose agent is available, a Test Cloud public preview is planned, and final testing judgment remains human. Whether the testing rollout makes the handoff dependable in varied enterprise workflows remains an open question.