Test scoping and test case generation with Joule

The newest AI features in SAP Cloud ALM target test design. With Joule, two agents work in sequence: the Test Scoping Agent decides what to test, and the Test Preparation Agent decides how, proposing a test plan and then generating manual test cases. A third, the Test Data Management Agent, reads a manual test case and lists the test data it needs. As of October 2026, SAP's help pages mark all three as part of an Early Adopter Care program, so they are not available outside it. Learn how they work now; plan production use only once they're GA for you.

Prerequisites

Joule activated with SAP Cloud ALM, plus the agent features requested in SAP for Me under Portfolio & Products > Business AI (SAP lists the ALM Project Setup Agent and Test Data Management Agent features).
A custom solution process with a custom Solution Process Flow Diagram containing BPMN content.
At least one requirement or user story linked directly to that diagram or its parent solution process.
Display access to Test Management, Process Management, Project Management and Requirements Management, and Manage authorization for Test Management to generate.

Two stages, two confirmations

Nothing is generated until a person confirms the scope and then the test plan.

The Sign-Off Table · thesignofftable.com/academy/sap-ai-206

In stage 1 you ask Joule, for example, "I want to test process Sell from Stock 1910. Prepare the test scope for review." The agent can't see your screen, so name the process, requirement, user story or project. It returns a Scope Summary: relevant requirements and user stories, main process paths, alternative and exception variants, and reference information. If the scope is wrong, fix the underlying requirements or process documentation and start again; the workflow doesn't continue until you confirm.

In stage 2, in the same conversation, the Test Preparation Agent proposes a Test Plan Summary: planned test cases, coverage categories such as happy path, error case and edge case, and the variants covered. Only after you confirm does it generate, asynchronously. The result is a new test plan with manual test cases (steps and expected results), traceability links to the source requirements, user stories and processes, and coverage information.

Limits that shape how you use it

Limit (per SAP's documentation)	Consequence for your test strategy
One Solution Process Flow Diagram per run	Plan runs per process; end-to-end scenarios across processes still need a person
Direct links only; indirect relationships are ignored	Link requirements and stories directly to the process, or they won't be in scope
Manual test cases only, no automated scripts in this release	Automation remains a separate step (see the lab)
Output reflects the quality of your requirements and process documentation	Weak requirements produce weak tests; fix the source, not the output

What the agents can and can't do in this release.

What would you do?

Contoso is in the Early Adopter Care program. Priya asks the Test Scoping Agent to prepare scope for the Monterrey process "Sell from Stock 1910". The Scope Summary lists only the happy path, although four requirements about credit blocks and CFDI cancellation exist for Wave 2. What is the most likely cause, and what should she do?

The Test Data Management Agent

Ask Joule, for example, "What test data do I need for test case 8-1-1578?" and the agent reads the instruction text of the case's in-scope activities and returns parameters grouped as Org (company code, plant, sales organization), Master (customer, material, supplier), Config (document type, posting period, Fiori app) and Transactional Data (quantity, delivery date, net price), each with a description, an example value and its reasoning. Two cautions from SAP: it doesn't create or provision data, and its example values are SAP reference defaults such as company code 1010 or plant 1710, not values from your system. For Wave 2, every "1710" it suggests must become 1910 before anyone runs the test.

Is the feature GA for your tenant, or are you in the EAC program?Neither → write tests the usual way and track the feature · Yes → continue.
Is there a custom process diagram with directly linked requirements?No → fix the process documentation first · Yes → run the Test Scoping Agent.
Does the Scope Summary match what the business must prove?No → correct requirements and rerun · Yes → confirm and review the Test Plan Summary.
Have key users reviewed each generated case and set it to Prepared?No → cases stay In Preparation and can't be executed · Yes → test.

A decision path for using Joule test generation on a project.

The Sign-Off Table · thesignofftable.com/academy/sap-ai-206

Hands-on lab · ACCELQ

Automate a reviewed AI-drafted test case with data-driven Test Cases

SAP Cloud ALM's agents produce manual test cases only. This lab practices the next step: turning one reviewed, AI-drafted manual case into an automated Scenario, then using ACCELQ's Autopilot option to propose Test Case rows that you review like any other AI output.

Practice on SAPUI5 Shopping Cart demo app (public, no login)

If you haven't set up ACCELQ yet, do the setup lab first.
Write a short manual test case as if an AI agent had drafted it: "Add an available product to the cart; the cart lists the product with its price." Mark one detail ASSUMED (for example, the price format) so you remember to check it.
Click + > Action, name it Add product to cart, and in the Logic Editor start the Web Recorder. In Fluid mode, open any category, pick any available product, add it to the cart and open the cart. Switch to OFF.
Switch to Intent mode, right-click the product line in the cart and add Verify Element Text. Replace the recorded category and product names with Action input parameters.
Build a Scenario with Initiate Browser Session and your Action. On the Test Cases tab, add two rows by hand with a category and product you saw in the app.
Use Generate > Autopilot to propose more Test Case rows. Review every proposed row against the app: delete rows whose category or product doesn't exist, and keep a note of how many you removed and why.
Click Run, then open the Test Report. For each failure, decide whether the test, the data or the app was wrong.

Done when The Scenario passes for every Test Case row you kept. Keep the Test Report link and a three-line review note: rows proposed, rows kept, and why the others were removed. That note is the same evidence you'd attach when promoting an AI-generated test case from In Preparation to Prepared.

Learn more in ACCELQ’s docs: Working with Test Cases in ACCELQ · Concept of Parameterization and Test Cases · Recording Fundamentals

Start the free ACCELQ trial Matt Angerer, who runs The Sign-Off Table, works at ACCELQ. Disclosure

Key points

The Test Scoping Agent proposes what to test; the Test Preparation Agent proposes a plan and generates manual test cases; both are Early Adopter Care as of October 2026.
You confirm the Scope Summary and the Test Plan Summary before anything is created.
Generated cases are flagged "Manually created with AI" and must be set to Prepared by a person before execution.
The Test Data Management Agent lists required data with SAP reference examples; it doesn't provision data or know your values.

Prerequisites

Two stages, two confirmations

Limits that shape how you use it

The Test Data Management Agent

Lesson check