Lesson 2 of 6
Requirement generation in SAP Cloud ALM
8 min
Listen to this lesson
Fit-to-standard workshops produce hours of discussion and, traditionally, days of writing. SAP Cloud ALM's AI-assisted requirement generation shortens the writing: it reads workshop transcripts and related documents and drafts a requirement for you to review. SAP's product page describes it as filling a predefined customer requirements template and suggesting solution proposals from SAP Activate and SAP Best Practices content. SAP also claims it saves 50% of the time on requirement creation and 20% on user story creation. Treat those figures as SAP's marketing claims until you measure your own (lesson 6).
Before you start
- Licensing: SAP's product page says AI Units are currently required.
- Activation: request the feature in SAP for Me (Portfolio & Products > Business AI, Request Activation); SAP Support notifies you when it is active. SAP's toolchain guide also lists the setting under SAP Cloud ALM Administration > AI Configuration.
- Region: not available for tenants in the China (North 3) region, cn20.
- Input: the workshop transcript must already exist as a document in the Documents app.
The procedure
- Create > Requirement > GenerateStart from the Requirements app.
- TitleWrite a clear title; it becomes part of the prompt.
- Template, scope, solution processPick the requirement template if you have one, the scope and the related solution process.
- DocumentsSelect the transcript plus process documents and minutes; remove irrelevant or inappropriate text first.
- SAP Help (optional)Include SAP Help documentation for standard behavior.
- Generate, review, saveEdit or delete what's wrong, or discard the draft; the text stays editable after saving.
Notice what the procedure assumes: one requirement per run, with a human-written title and a human-selected set of documents. That is a feature, not a limitation. The title and the documents steer the model, so a vague title such as "Sales stuff" produces a vague requirement. Contoso's rule for Wave 2 is one requirement per business rule, with a title in the form process, object, rule, for example "O2C 1910: payment terms for migrated customers (CR30 to 0004)."
The ALM Requirement and User Story Agent
Alongside the generate button, SAP Cloud ALM lists a Joule agent called the ALM Requirement and User Story Agent. SAP describes it as helping key users, functional experts and process owners generate requirements and user stories from workshop transcripts, enrich attributes, and find, track, update and analyze requirements in conversation, for example "show me all requirements created in the last two weeks that are still waiting for a reviewer decision." Useful as that is, the same review rules apply to anything it writes.
Reviewing a generated requirement
| Check | What to look for | Wave 2 example |
|---|---|---|
| Traceable to the transcript | Every statement can be found in the selected documents | The CR30 to 0004 mapping appears at minute 42 of the O2C workshop |
| No invented values | Thresholds, codes, org units and dates match decisions | Sales area 1910 / 10 / 00, not 1710 / 10 / 00 |
| Standard vs gap stated | Says whether standard SAP covers it or a gap exists | Core return agreements need the custom table carried over from ECC |
| Testable | Contains acceptance criteria someone can verify | "All 1,946 in-scope customers carry payment term 0004 after load" |
| Right owner and scope | Solution process and scope match the workshop | Owned by the Monterrey O2C key user; scope Wave 2 |
Five checks for every AI-generated requirement before it goes for approval.
In SAP Cloud ALM, setting Ready for Approval moves the requirement to In Approval and locks its description. If a project lead has assigned the Requirement Approver role, only those users can approve, and the approval is logged in the history. Approved and Rejected can't be reversed; once approved, the requirement moves to In Planning and can be broken into user stories and tasks. That lock and log are exactly what you need to prove a person, not a model, accepted the requirement.
What would you do?
Lucía, the Monterrey order-to-cash key user, generates 30 requirements from the first workshop in one afternoon. Six of them mention sales area 1710 / 10 / 00 (Detroit) instead of 1910 / 10 / 00. What should Priya ask the team to do?
- Requirement generation drafts one requirement per run from transcripts, process documents and optional SAP Help content.
- It needs AI Units, activation via SAP for Me, a supported region (not cn20) and the transcript in the Documents app.
- A clear title and clean, relevant documents steer the output; review every statement against the transcript.
- Use Ready for Approval and the Requirement Approver role so approval is locked and logged.
Check your understanding · not graded
Lesson check
1. Where must the workshop transcript be before you can generate a requirement?
2. Why does the requirement title matter for generation?
3. What happens to a requirement's description when it is set to Ready for Approval?
4. SAP claims a 50% time saving on requirement creation. How should a PMO lead treat it?