# Regression testing for SAP, Oracle, IFS and PeopleSoft: UiPath's dark testing factory in practice > Regression testing for SAP, Oracle, IFS and PeopleSoft on every vendor update. How the dark testing factory in UiPath Test Cloud works and where people decide. - Source: https://snok.ai/en/news/blog/erp-regression-testing-sap-oracle-dark-testing-factory-uipath/ - Author: Jacek Bugajski - Published: 2026-10-08 --- On 6 October 2026 UiPath Test Cloud went live on Oracle Marketplace, with a stated scope of Oracle Fusion Cloud Applications, Oracle E-Business Suite, PeopleSoft, Siebel and JD Edwards. A fortnight earlier, at FUSION, UiPath said that from October Autopilot would be able to read a manual test case and run it against an application with no automation built in advance. The company has a name for where the product is heading: the dark testing factory. For anyone who owns an ERP estate, regression testing keeps presenting the same problem. You are checking functions nobody touched, against a deadline nobody in the building chose. This piece looks at what UiPath means by a testing factory, how a single change travels through its components, how regression differs across SAP, Oracle, IFS and PeopleSoft, and where agent autonomy stops. ## The vendor now owns your release calendar The days when IT decided when to take a new version are over for a growing share of the application portfolio. Vendors have moved their customers onto a cadence they set themselves. **Oracle Fusion Cloud Applications.** Oracle ships new features four times a year in quarterly updates, and its own documentation describes them as mandatory for all customers. Non-production environments receive the update about two weeks before production, with Oracle tying the dates to the first and third Fridays of the update month. That is the regression window. **SAP S/4HANA Cloud Public Edition.** Two releases a year, each with a four-week testing period before production, plus monthly Hotfix Collections with a one-week preview. On-premises and in Private Edition, SAP has been on a two-year release cycle since 2023, with feature packs every six months during the first two years of each release. **IFS Cloud.** Two releases a year, which in 2025 meant 25R1 at the end of May and 25R2 at the end of November, with monthly service updates in between. **Oracle E-Business Suite and PeopleSoft.** Here the pressure is of a different kind. Oracle has committed to Premier Support for E-Business Suite 12.2 and for PeopleSoft through at least 2037. These systems may well be around for more than another decade, and every PeopleSoft Update Manager image lands on an environment that is often layered with years of customisation and integration. The rest of a typical estate follows the same pattern. Workday releases twice a year, in March and September, with weekly fixes. Salesforce releases three times a year. Dynamics 365 Finance and Operations requires at least two service updates a year and gives you seven days to test before production is updated. A company running only Oracle Fusion, Salesforce and Workday absorbs nine major vendor releases a year, each with its own short test window. ![Major releases per year and pre-production test windows across ERP and SaaS platforms](https://snok.ai/images/blog/fabryka-testow-kalendarz-wydan-en.webp) Then there is the change on your own side. On 7 October 2026 SAP confirmed that mainstream maintenance for SAP Business Suite 7 ends on 31 December 2027 and optional extended maintenance on 31 December 2030. In many organisations S/4HANA conversions are therefore landing on top of quarterly cloud cycles, often in the same teams. ## Why test teams cannot keep pace Regression effort does not scale with the size of a change. It scales with everything the change might have touched: integrations, configuration variants, master data, roles and authorisations, reports. The window stays the same length, so testing often stops when the deadline arrives, some way short of the end of the case list. Meanwhile, change is being produced faster. In its report “The Acceleration Whiplash”, Faros AI compared data from 22,000 developers and 4,000 teams across periods of low and high AI adoption. Task throughput per developer rose by 33.7 per cent. Bugs per developer rose by 54 per cent, incidents per pull request by 242.7 per cent, and pull requests merged with no review at all by 31.3 per cent. Review is not keeping pace with change. In the process model we use in client workshops, a test team's work comes to twenty activities across four phases: test design, test automation, execution and management. People carry out thirteen of them. In that model, conventional automation mainly takes over test execution. Writing test cases, repairing scripts after a screen changes, reviewing results and telling a genuine defect from a flaky test still wait for a person, and they set the length of the cycle. ## What UiPath means by a dark testing factory The term comes from manufacturing. A lights-out factory is one where robots work the floor unattended and nobody needs the lights on. UiPath applies the idea to software quality and describes the testing factory as a self-operating quality system that plans, designs, executes and maintains tests end to end. Two lines on the product page deserve a careful read. The first says the dark testing factory is where Test Cloud is heading, explicitly not a separate product. The second says it does not replace testers but elevates their role. The distinction matters, because “dark factory” sounds like a place nobody is watching. In practice it means moving people out of execution and into decisions about quality and risk. UiPath lays out the route in five steps, with a different role for people at each. ![Five levels of testing maturity, from manual testing to the dark testing factory](https://snok.ai/images/blog/fabryka-testow-poziomy-dojrzalosci-en.webp) A typical team with an automation library sits somewhere between levels two and three: a library of automations that someone has to maintain, and early use of AI assistants for writing test cases. Testing with agents starts at level four, where an agent works independently within a narrow, supervised flow. Level five adds one thing. The system runs continuously, in step with change, instead of waiting for someone to start it. ## How one change moves through the factory In our model, one factory covers one application. A portfolio is a set of factories at different stages of readiness, with a single supervisory layer across all of them. The animation below follows a single change through four systems, each with a different outcome. ![A change moving through the regression test factory: signal, risk assessment, test selection, execution, result analysis and a human release decision](https://snok.ai/images/blog/fabryka-testow-obieg-en.gif) **1. Change signal.** The factory is triggered by an event: an SAP transport, a Jira ticket, a commit to GitHub, a CI/CD pipeline run or a manual start. In our model every signal is logged with the reason it was accepted, and a signal with no impact on the application creates no work, only a record of why. **2. Risk assessment.** Conventional automation has no equivalent step. In our model several independent agent roles share the assessment. One identifies the risks the change itself introduces and is deliberately kept unaware of existing tests, so it does not narrow the list. Others map those risks onto the current risk picture, rank them against the quality owner's tolerance and rate coverage as strong, partial, missing or uncertain. A final role checks the whole assessment independently. The output is a recommendation on depth: repeat existing tests, extend the scope or write new ones. For SAP, UiPath ships a ready-made tool here, Change Impact Analysis, which lists the transactions a transport touched and flags those with no test assigned. We covered it in detail in our piece on [SAP S/4HANA testing during conversion and before go-live](https://snok.ai/en/news/blog/sap-s4hana-testing-conversion-performance-uipath-test-cloud/). **3. Test selection and generation.** UiPath Test Manager is the control plane: planning, linking test cases to requirements, and results. UiPath cites sync with more than fifty ALM tools, and its documentation covers integration with SAP Solution Manager and SAP Cloud ALM. Autopilot generates manual test cases straight from requirements, up to fifty at a time, from text documents, spreadsheets, BPMN files and even screenshots. Where an automation is missing, a new reusable test is built. **4. Regression run.** UiPath Test Cloud robots run tests in parallel across web, desktop and mobile applications, SAP GUI and Fiori, and APIs. The trigger can be a CI/CD pipeline in Jenkins, GitLab, GitHub Actions or Azure DevOps. UiPath has also announced a Playwright integration in public preview from October 2026, intended to let teams with existing Playwright suites run them from Test Manager without rewriting them. **5. Result analysis.** This is where teams lose most of their time. In our model an analysis agent separates genuine application defects from flaky tests. UiPath Healing Agent repairs, among other things, broken selectors and timing issues during execution. When the cause is a screen the vendor changed in an update, the test is healed and goes back for a re-run. When the cause is a real defect, our loop raises a ticket and sends the case back to risk assessment. **6. Release decision.** The loop ends with a readiness report: a verdict, the evidence behind it and a list of unresolved risks and limitations. UiPath Maestro runs the whole loop as a process with explicit gates, in which people, agents and automations work in a single flow. A person makes the release call. ## UiPath Test Cloud components at a glance ![Test factory architecture in UiPath Test Cloud: control layer, agents, execution and governance foundation](https://snok.ai/images/blog/fabryka-testow-komponenty-en.webp) **UiPath Test Manager** is the control layer: plans, test cases, requirement links, results and ALM integrations. **UiPath Maestro** coordinates agents, automations and people in one flow. Decision gates appear in it as named, visible process steps. We wrote about this in more depth in our piece on [HITL gates in UiPath Maestro and AI Trust Layer](https://snok.ai/en/news/blog/hitl-gates-uipath-maestro-ai-trust-layer/). **UiPath Autopilot** is the AI assistant across test design, automation, execution and management. **UiPath Healing Agent** repairs automations while they run. **UiPath Studio and Studio Web** cover low-code and pro-code test authoring. **UiPath Agent Builder** lets you build agents for processes specific to your own application, such as a risk assessment role for a regulated area. **AI Trust Layer** is the governance layer over how AI models are used in the platform. Test Cloud is available as a cloud service (Public Test Cloud) and as a deployment in your own environment (Private Test Cloud, on Automation Suite). The second option matters wherever test data cannot leave the organisation. UiPath notes that capabilities vary by deployment option, release and platform tier, so the exact configuration is worth confirming during solution design. ## Regression testing in SAP, Oracle, IFS and PeopleSoft: how they differ **SAP.** Changes reach SAP systems through transports, and UiPath has a dedicated set of SAP capabilities: process-level testing across SAP S/4HANA and ECC, in SAP GUI, WebGUI, Fiori and Business Client, a patented Heatmap for SAP, and Change Impact Analysis across the user interface, APIs (BAPI, RFC, IDoc) and authorisations. The SAP Cloud ALM integration works both ways, so test cases, execution and results are visible in both systems, within what each release supports. In a conversion from ECC, regression carries an extra burden, because the old test suite checks a data model that is about to disappear. Our [S/4HANA conversion](https://snok.ai/en/offer/sap-security/s4hana-conversion/) page sets out what such a project covers. **Oracle Fusion Cloud Applications.** The difficulty is timing. Four times a year the vendor changes screens, flows and fields on a date you cannot move, with roughly two weeks between the test environment and production. This is exactly the scenario where test self-healing stops being a novelty: a test broken by a changed screen, with no defect in the application, should not eat half of someone's test window. **Oracle E-Business Suite and PeopleSoft.** Systems with years of support ahead of them and, in many installations, heavy customisation. Every PeopleSoft update image and every E-Business Suite patch lands on custom changes the vendor never tested. Assessing the risk of a change carries more weight here than execution, and the test library is often the only place where knowledge of the system is written down. Since 6 October 2026 UiPath Test Cloud has been available for both through Oracle Marketplace. **IFS Cloud.** UiPath has published no dedicated material on testing IFS, and we will not fill that gap for them. IFS Cloud runs in the browser, so we test it as a web application: automations built in Studio, selectors maintained by Healing Agent, test cases generated from requirements by Autopilot. Two releases a year plus monthly service updates give it a rhythm close to Oracle Fusion, with the same need for regression on every update. **Workday, Salesforce and Dynamics 365.** UiPath covers Workday and Salesforce on its enterprise applications page, including regression around Workday releases. We test Dynamics 365 the same way as IFS, as a web application. ## Where agent autonomy stops A testing factory is not an unsupervised system. The boundary sits in three places. **What the system restores on its own:** the test run and its connections to the application, the test environment and test data for each run, and a retry after a failed attempt. **What stays a human decision:** any change to application code, approval of fixes before they are applied, and sign-off on a release that still carries open risk. Restoring a test run is not the same as fixing the application. Self-healing applies to the test and never touches the code of your system. **What is agreed before the factory starts:** its goal, connected sources and operating rules. In our model tests run only in a test environment, on masked data, with no copies of production data, and production is read-only at most. Any risk an agent cannot assess goes to a person. Autonomy is set by operating mode, stage by stage. In manual mode a person starts every item. In semi-autonomous mode results wait for sign-off. In autonomous mode the agent writes results without asking, and policies still apply. Budget works at two levels, a per-run limit and a monthly limit, and you choose in the settings what happens when the pool runs out. Wherever change accountability matters, the audit trail of a test result matters as much as the result: where it came from, who approved it and under which rule. ## The same model in our own product: SNOK MDM Before offering this model to clients, we put it to work on our own software. Testing of our [SNOK MDM](https://snok.ai/en/products/snok-mdm/) application is automated in UiPath: every change triggers a regression run over the data quality rules and reference lists, and data quality tests run continuously. A new release no longer waits for someone to click through the screens by hand. For SNOK MDM clients the split of work is simple. What stays on their side is defining the test scenarios, meaning the processes and data that must still work after every change. We take on building the automations, running them and keeping them working as the application changes, using either the client's own UiPath licences or licences we supply together with the SNOK MDM licence. ## Piloting a test factory: one application, one process We start a pilot with one application and one process, such as month-end close in SAP, or the procure-to-pay process in Oracle Fusion ahead of the next quarterly update. Four inputs are needed: the chosen process, the sources of change signals, access to a test environment and a quality owner on your side. You come away with a factory for that process, a library of test cases and test data, a decision trail with artefacts, and a readiness report listing the gaps. After the pilot there are three options: extend the model to further processes, keep the pilot scope, or stop with no further commitment. The first step, a scoping workshop, needs no changes to your environment and no purchasing decision. UiPath is positioned as a Leader in two analyst reports on testing: the 2025 Gartner Magic Quadrant for AI-Augmented Software Testing Tools and The Forrester Wave: Autonomous Testing Platforms, Q4 2025. SNOK is a UiPath Platinum partner. Our services are described on the [agentic testing](https://snok.ai/en/offer/ai-automation/agentic-testing/) page, and process orchestration on the [UiPath Maestro](https://snok.ai/en/offer/ai-automation/uipath-maestro/) page. An earlier piece on [agentic testing and tests that hunt for bugs](https://snok.ai/en/news/blog/agentic-testing-uipath-test-cloud/) shows where this shift began. If your next ERP update already has a date and your regression plan does not, [write to us](https://snok.ai/en/contact/). We will start by choosing the process for a pilot. ## Sources - UiPath, announcement of UiPath Test Cloud availability on Oracle Marketplace, 6 October 2026, https://www.uipath.com/newsroom/uipath-oracle-marketplace-test-cloud (accessed 7 October 2026) - UiPath, announcement of UiPath Test Cloud enhancements (FUSION 2026), 23 September 2026, https://www.uipath.com/newsroom/uipath-enhances-test-cloud (accessed 7 October 2026) - UiPath, Test Cloud product page, https://www.uipath.com/product/test-cloud (accessed 7 October 2026) - UiPath, Agentic Testing, https://www.uipath.com/platform/agentic-testing (accessed 7 October 2026) - UiPath, Agentic software testing for SAP, https://www.uipath.com/platform/agentic-testing/sap-testing (accessed 7 October 2026) - UiPath, enterprise application testing, https://www.uipath.com/platform/agentic-testing/enterprise-applications (accessed 7 October 2026) - UiPath, white paper on the dark testing factory, https://www.uipath.com/resources/automation-whitepapers/dark-testing-factory-accelerated-software-delivery (accessed 7 October 2026) - UiPath Documentation, Autopilot manual test generation, https://docs.uipath.com/autopilot/other/latest/user-guide/manual-test-generation (accessed 7 October 2026) - UiPath Documentation, Test Manager and SAP Cloud ALM, https://docs.uipath.com/test-manager/automation-cloud/latest/user-guide/sap-cloud-alm (accessed 7 October 2026) - Oracle, Fusion Applications documentation on environment families, https://docs.oracle.com/en-us/iaas/Content/fusion-applications/plan-environment-family.htm (accessed 7 October 2026) - Oracle, Premier Support for E-Business Suite 12.2 through at least 2037, 18 March 2026, https://www.oracle.com/a/ocom/docs/applications/ebusiness/ebs-122-premier-support-extended-through-at-least-2037.pdf (accessed 7 October 2026) - SAP, SAP S/4HANA Cloud Public Edition maintenance schedule, 24 September 2026, https://d.dam.sap.com/x/qgSFemF/S4HANA_3SL_Maintenance_Schedule_for_Customer_Communication_2026Q34_2027Q12_2026_09_24.pdf (accessed 7 October 2026) - SAP News, SAP S/4HANA release and maintenance strategy, 15 September 2022, https://news.sap.com/2022/09/new-sap-s4hana-release-maintenance-strategy/ (accessed 7 October 2026) - SAP News, SAP Business Suite 7 maintenance beyond 2030, 7 October 2026, https://news.sap.com/2026/10/beyond-2030-sap-business-suite-7-sap-netweaver-customers-third-party-databases-java/ (accessed 7 October 2026) - IFS, what's new in IFS Cloud 25R2, 24 October 2025, https://www.ifs.com/en/assets/cloud/what-new-in-ifs-cloud-25r2 (accessed 7 October 2026) - Workday, release schedule, 11 July 2025, https://doc.workday.com/admin-guide/en-us/about-workday-documentation/workday-documentation---in-depth/workday-release-schedule.html (accessed 7 October 2026) - Salesforce Help, testing and preparing for major releases, https://help.salesforce.com/s/articleView?id=sf.availability_test_and_prepare_for_major_releases.htm (accessed 7 October 2026) - Microsoft Learn, Dynamics 365 Finance and Operations service updates, 1 September 2026, https://learn.microsoft.com/en-us/dynamics365/fin-ops-core/dev-itpro/get-started/one-version (accessed 7 October 2026) - Faros AI, The Acceleration Whiplash, AI Engineering Report 2026, https://pages.faros.ai/hubfs/AI_Engineering_Report_2026_The_Acceleration_Whiplash_Faros.pdf (accessed 7 October 2026) - UiPath, Leader in the Gartner Magic Quadrant for AI-Augmented Software Testing Tools, 9 October 2025, https://www.uipath.com/blog/product-and-updates/uipath-named-leader-gartner-mq-for-ai-augmented-software-testing-tools (accessed 7 October 2026) - UiPath, Leader in The Forrester Wave: Autonomous Testing Platforms, Q4 2025, 17 December 2025, https://www.uipath.com/blog/product-and-updates/uipath-named-leader-forrester-wave-autonomous-testing (accessed 7 October 2026)