Agentic testing evolves again: the next chapter for UiPath Test Cloud

Software development is accelerating. AI is helping teams create more code, make more changes, and move faster than ever before. But every change creates more risk to evaluate, putting software testing teams under increasing pressure to keep pace.

At UiPath FUSION in Las Vegas, we're showing how UiPath Test Cloud is evolving again to meet that challenge.

Across sessions spanning AI-powered testing, performance testing, SAP®, agent orchestration, enterprise transformation, and customer stories, two themes stand out: enterprise scalability and agentic testing.

Enterprise scalability is about helping organizations mature their testing operations: standardizing how teams work, expanding automation across complex technology landscapes, and providing the governance and infrastructure needed to operate software testing at scale.

Agentic testing changes how the testing work itself gets done. At FUSION, we're showing how it’s evolving again.

From agentic testing toward autonomous testing at enterprise scale

Agentic testing augments testers with AI agents across the software testing lifecycle, helping teams design, automate, execute, and manage testing with greater speed and intelligence. UiPath Test Cloud brings this to life through built-in agents, the ability to build custom agents, and orchestration across agents, automations, and humans.

At FUSION, we're showing how agents are moving beyond assisting with individual tasks and beginning to take on more responsibility for the testing work itself: interpreting intent, executing tests, exploring applications, analyzing results, and determining what should happen next.

That shift points toward autonomous testing at enterprise scale.

The goal isn’t autonomy for its own sake. It’s to increase testing capacity and speed without increasing manual effort at the same rate, while maintaining the governance, traceability, and human oversight enterprises require.

That last part isn’t a footnote. As agents take on more of the operational loop, the governance question moves from “can we automate this?” to “can we account for what the automation did?” Role-based access control, centralized credential management, and audit trails are what make increasing autonomy something an enterprise can actually put into production rather than pilot indefinitely.

Several innovations we're highlighting at FUSION show how this transition is taking shape.

Moving from instructions to autonomous action

Traditional test automation requires someone to build the automation first. A tester defines the steps, an automation engineer implements them, and the machine executes those predefined instructions.

With the autonomous test execution capability in Test Cloud, you can give UiPath Autopilot™ a manual test, and it can interpret the instructions and execute the test against the application without requiring prebuilt automation.

And not all valuable testing begins with a predefined test case.

Human testers explore. They investigate unexpected behavior, challenge assumptions, and try paths nobody explicitly documented beforehand. With autonomous exploratory testing, a tester can provide a mission and let an agent autonomously explore the application, uncover potential defects, and report what it discovers.

The operational work around testing can evolve too. UiPath Delegate is the productivity agent built for the enterprise. It acts on the user's behalf and completes administrative and role-specific tasks, including the operational work around software testing. Delegate helps testers with various tasks, from gathering context and moving information between systems to updating records, attaching evidence, and preparing summaries. It can turn repeatable workflows from tester actions and recorded sessions into reusable routines and execute them on demand, on a schedule, or from defined triggers, while testers remain in control of judgment, strategy, and quality decisions.

Together, these capabilities mark a progression: from automating predefined steps toward giving agents greater autonomy in how the work gets carried out.

Automation remains essential for predictable, repeatable testing at scale. But increasing autonomy gives teams another way to expand coverage and shift human expertise toward risk, strategy, and judgment.

Testing AI…with AI

The software under test is also changing.

AI-infused applications and AI agents can behave differently from traditional deterministic software. The same interaction may not always produce exactly the same response, so quality can’t always be evaluated by checking whether one fixed output equals another.

Testing teams need to answer different questions. Did the agent accomplish the intended goal? Was its response appropriate? Did it remain within expected boundaries? Can the application behave reliably even when the exact output varies?

Test Cloud helps teams test AI where it surfaces in applications, including through user interface (UI) and API interactions, and evaluate non-deterministic outcomes using semantic and AI-based verification approaches.

This isn’t about replacing specialized model evaluation performed by AI engineers or data scientists. It’s about equipping software testing teams to evaluate the applications and agents that use AI, where those systems interact with real users and business processes.

There’s a second-order version of this problem worth naming: the testing system itself now runs on AI. Teams evaluating AI-infused applications with AI-powered testing need visibility into their own agents’ behavior, too. In Test Cloud, that’s the job of the AI Trust Layer—auditing AI activity, spend controls, and personally identifiable information (PII) masking, with support for bringing your own model. It’s the difference between adopting AI in testing and being able to explain, months later, what it did and what it cost.

As AI moves deeper into applications and systems, assuring the quality of AI-infused applications and agents becomes part of software testing itself.

Staying open as enterprises scale

Enterprise testing rarely starts with a blank slate. Organizations already have frameworks, pipelines, tools, and test assets they rely on.

That’s why the Playwright integration is an important part of the Test Cloud story at FUSION. We’re working toward enabling Playwright tests created outside UiPath to run through Test Manager, connecting existing coded testing investments with broader management and orchestration in Test Cloud, and enriching them with agentic AI, including self-healing that repairs shifted selectors mid-run rather than leaving a failed build for someone to triage.

This reflects a larger principle: enterprise-scale testing needs to be open enough to embrace how development and testing teams already work rather than forcing an all-or-nothing tooling decision.

The next horizon: the dark testing factory

Viewed together, these innovations point beyond individual agentic capabilities toward a broader operating model for autonomous testing at enterprise scale: the dark testing factory.

A dark testing factory is a largely self-operating quality system in which agents and automations increasingly handle the continuous operational loop of testing: understanding incoming changes, assessing risk, designing and executing tests, analyzing results, maintaining test assets, managing environments, and routing exceptions.

A system like that only works if it’s governable. The more of the loop agents run, the more the surrounding controls matter: who can approve what, which credentials an agent can use, what record exists of every decision, and where the data lives. That last point is why deployment choice matters as much as capability. Test Cloud is available as Public Test Cloud, managed by UiPath, or Private Test Cloud, deployed on-premises where organizations need full control and data residency.

No matter the deployment model, one thing stays true: people are essential to the dark testing factory, and their role changes because of it.

Instead of spending their time inside the routine operating loop, testers increasingly define intent, provide business context, establish guardrails, resolve ambiguity, and apply judgment where it matters most.

The dark testing factory isn’t about removing people from quality. It’s about removing people from repetitive operations so human expertise can be focused on the decisions machines can’t make alone. The dark testing factory is our longer-term vision for where this evolution can lead, and some of the building blocks we showed at FUSION are available today.

Building quality for the next era of software

That’s the bigger Test Cloud story coming out of FUSION. Enterprise scalability gives organizations the foundation to operate testing consistently across teams and applications; agentic testing changes the operating model built on top of it.

And now agentic testing is evolving again, toward increasingly autonomous testing, coordinated and governed at enterprise scale. Autonomous test execution, autonomous exploratory testing, testing AI-infused software, openness to frameworks like Playwright, and the dark testing factory all point the same direction.

Software development is becoming more autonomous; the quality system around it needs to evolve with it.

Want to go deeper? Explore UiPath Test Cloud and see how agentic testing works in practice.