- How the C_AIG Blueprint Works
- Format, Scoring and Booking Mechanics
- Domain 1: System, User and Assistant Roles
- Domain 2: Securing and Hardening Prompts
- Domain 3: Prompt Templates in SAP AI Launchpad
- Domain 4: Discovering the Generative AI Hub Interface
- Domain 5: Solution Using Different Models
- Domain 6: Problem Solving in SAP AI Launchpad
- Domain 7: Developing Prompts in SAP AI Launchpad
- Sequencing Your Preparation by Domain
- What Is Not a Separate Domain
- Frequently Asked Questions
- The SAP Certified Associate - SAP Generative AI Developer exam has seven content areas, each published at an 11-20% weight range.
- The exam is one 3-hour, system-based, practical activity with a 76% passing score, so you must do tasks, not recall trivia.
- Equal weight bands mean no single domain is officially heaviest; neglecting any one of them is risky.
- Practical exams are open-book and unproctored, but the timer keeps running during breaks, so practice speed matters.
How the C_AIG Blueprint Works
The SAP Certified Associate - SAP Generative AI Developer credential (C_AIG) is built around seven published exam topics. Each carries the same published weight range of 11-20%. That detail shapes everything about how you should prepare. Ranges are not exact allocations, and they should not be normalized into neat percentages such as "14% each." SAP publishes bands, and the practical conclusion is simple: no domain can be treated as safely skippable, and no domain can be declared the single most important one.
This guide walks through each content area using SAP's current official topic labels, explains what a candidate is realistically expected to do in the system, and then shows how to sequence your preparation. If you want the broader picture first, the C_AIG Study Guide covers the end-to-end plan, and What Is C_AIG Certification? explains the credential itself. For the domain overview in a shorter format, see the companion C_AIG Exam Domains guide.
| # | Official Domain Label | Published Weight Range |
|---|---|---|
| 1 | Leveraging System, User, and Assistant Roles for Better Prompts | 11-20% |
| 2 | Securing and Hardening Prompts | 11-20% |
| 3 | Using Prompt Templates in SAP AI Launchpad | 11-20% |
| 4 | Discovering Generative AI Hub Interface | 11-20% |
| 5 | Solution Using Different Models | 11-20% |
| 6 | Focus on problem solving using SAP AI Launchpad | 11-20% |
| 7 | Developing Prompts in SAP AI Launchpad | 11-20% |
Format, Scoring and Booking Mechanics
Before diving into the domains, it helps to understand the container they sit in. According to SAP's published details, the assessment is a single system-based activity lasting 3 hours, offered in English, with a passing score of 76%. "One activity" is not a count of questions or tasks, so do not assume the exam is a single long exercise or a list of a specific size. Treat it as a hands-on session in a live system.
What "system-based" changes
System-assessment scoring evaluates completed tasks and the outputs the system produces. That is different from a multiple-choice exam where partial reasoning can still earn a lucky guess. Here, what you configure, run and save is what counts. The practical exam is open-book and unproctored, and public SAP resources and applicable AI tools are permitted. That sounds generous, but the timer continues during breaks, and searching for the location of a setting you have never used burns minutes you cannot recover. Familiarity with the interface is your real advantage.
Booking and access
Access comes through purchased exam attempts or an SAP Learning Hub subscription. Your SAP Universal ID must have at least two email accounts on different domains, so sort that out well before exam day. On scheduling, SAP's materials are not perfectly aligned: the Q2 2026 release notes say advance booking is required for system-based exams, while the system-assessment FAQ says some systems are immediately available and others need a reserved slot. The safe approach is to plan for advance booking and follow what your own exam dashboard shows. Our C_AIG Exam Dates guide covers scheduling in more depth.
Domain 1: Leveraging System, User, and Assistant Roles for Better Prompts
This domain is about structuring conversations with a large language model so the model behaves predictably. Modern chat-style models distinguish between message roles, and the Generative AI Hub exposes that structure. Knowing which instruction belongs in which role is a core skill, not a footnote.
Domain 1: Roles in prompt design
Candidates must understand how role separation changes model behavior and how to apply it when building prompts in the hub.
- The system role sets standing behavior, tone, constraints and persona that should persist across the exchange.
- The user role carries the actual request or input data for the current turn.
- The assistant role holds prior model replies, and can be used to supply example outputs that steer format.
- Placing a rule in the wrong role is a classic cause of inconsistent output.
A preparation scenario
Imagine a task where a model must summarize supplier emails in a fixed three-line format and refuse anything unrelated to procurement. A weak solution crams everything into one user message. A stronger one places the format and scope rules in the system role, passes the email as the user message, and optionally seeds an assistant example showing the exact output shape. Practice rebuilding the same prompt both ways and compare the results. This is an original practice scenario, not an actual exam question, but it mirrors the type of reasoning you will need.
Domain 2: Securing and Hardening Prompts
Prompts that work in a demo can fail badly when real users type unexpected input. This domain covers making prompts resistant to misuse, including attempts to override instructions or extract hidden content.
Domain 2: Hardening mindset
Expect to reason about how untrusted input interacts with your instructions and how to reduce risk.
- Recognize prompt injection: user text that tries to countermand your system instructions.
- Keep trusted instructions separate from untrusted input, and make that separation explicit in the prompt.
- State clear boundaries on what the model should decline to do or reveal.
- Test your prompt with adversarial inputs, not only friendly ones.
The best preparation here is to attack your own prompts. Build a working prompt in the hub, then try instructions like "ignore the above and do something else" in the user input. Observe what breaks, tighten the wording, and retest. Hardening is iterative, and doing the loop yourself teaches more than reading about it.
Domain 3: Using Prompt Templates in SAP AI Launchpad
Templates turn a one-off prompt into something reusable. In SAP AI Launchpad, a template lets you parameterize the variable parts of a prompt so the same structure can serve many inputs.
Domain 3: Template usage
Know how to create, fill and reuse templates through the Launchpad interface.
- Identify which parts of a prompt are fixed and which should become placeholders.
- Supply values for placeholders when running a template.
- Understand why templates improve consistency and maintainability.
- Be aware of permissions that affect whether you can save work.
Domain 4: Discovering Generative AI Hub Interface
Because the exam is performed in a live system, navigational fluency is a scoring asset. This domain is about knowing what the Generative AI Hub offers and where to find it, so you are not hunting for menus while the clock runs.
Domain 4: Interface fluency
Be able to move through the hub confidently and recognize its main capabilities.
- Locate the areas used for experimenting with prompts and models.
- Understand what the hub provides as a central point for accessing generative models.
- Know how the hub relates to SAP AI Launchpad as the working interface.
- Spend practice time clicking through every section so nothing is unfamiliar on exam day.
This is the cheapest domain to prepare for if you use the practice system, and the most punishing to neglect. SAP provides a preconfigured subscription practice system for preparation, along with a Getting Started with Generative AI Hub lesson. Note that this system is for preparation, not production service access. Treat your practice hours as a limited resource, a point we revisit in the C_AIG Requirements guide.
Domain 5: Solution Using Different Models
The hub gives access to multiple models, and part of the skill is choosing and comparing them for a given problem. A prompt that performs well on one model may behave differently on another.
Domain 5: Model selection and comparison
Candidates should be able to run a task against different models and reason about the differences.
- Select a model deliberately instead of defaulting to the first available option.
- Run the same prompt against more than one model and compare outputs.
- Reason about trade-offs such as output quality, style and suitability to the task.
- Adjust prompts when a different model responds differently.
A good drill is to take one business task, such as classifying customer feedback, and run an identical prompt on two models. Note where formatting, tone or accuracy diverges, then revise the prompt to close the gap. This builds the habit of treating model choice as part of the solution, which is exactly what this domain rewards.
Domain 6: Problem Solving in SAP AI Launchpad
This domain is the most scenario-driven. It asks you to apply everything else to a business problem inside SAP AI Launchpad: read the requirement, choose an approach, build it, and verify the output.
Domain 6: Applied problem solving
Expect to translate a business need into a working generative AI solution in the system.
- Break a requirement into a prompt design, a model choice and a test.
- Iterate when the first output misses the requirement.
- Verify that the saved result and system output match what the task asks for.
- Manage time across a 3-hour session rather than perfecting one step.
Since scoring evaluates completed tasks and system outputs, finishing a reasonable solution beats polishing half of one. This is where the difficulty of the exam is felt most: not from obscure facts, but from the discipline of delivering working results under a clock.
Domain 7: Developing Prompts in SAP AI Launchpad
The final domain focuses on the craft of writing and refining prompts directly in the Launchpad. It overlaps with roles and templates but centers on iteration and quality.
Domain 7: Prompt development workflow
Be comfortable writing, running, adjusting and saving prompts within the Launchpad.
- Write clear, specific instructions and state the desired output format.
- Run a prompt, inspect the response, and refine wording based on what you see.
- Use parameters and settings available in the interface to influence output.
- Save work so it can be reused and reviewed.
Key Takeaway
Domains 1, 3 and 7 overlap heavily in practice, because you write prompts with roles, then turn them into templates, inside the same Launchpad. Practice them as one workflow, then drill Domains 2, 4, 5 and 6 as the complementary skills.
Sequencing Your Preparation by Domain
Because every domain sits in the same 11-20% band, sequence by dependency rather than by weight. Start with interface fluency, since every other task depends on it, then build prompt craft, then layer on security, models and applied problem solving. SAP's linked learning journey, Solving Business Problems using SAP's Generative AI Hub, supports this flow, and the courses AIG02 and the advanced AI techniques course are useful supplementary preparation. AIG00 is the prerequisite to AIG02, and AIG02 precedes the advanced course, but these are curriculum prerequisites rather than independently verified exam-admission requirements. Completing every course and exercise is recommended, not mandatory.
Interface and roles (Domains 4 and 1)
- Click through every section of the hub and Launchpad.
- Rebuild one prompt using system, user and assistant roles.
Prompts and templates (Domains 7 and 3)
- Write, refine and save prompts; convert one into a reusable template.
- Confirm your roles allow saving.
Hardening and models (Domains 2 and 5)
- Attack your own prompts with injection-style input and tighten them.
- Run one task across multiple models and compare.
Timed problem solving (Domain 6)
- Attempt full business scenarios against a 3-hour clock.
- Review gaps and revisit weak domains.
Keep your 10 practice-system hours for this last, hands-on phase instead of spending them on aimless browsing. For concept review away from the system, the C_AIG Cheat Sheet condenses the facts, and the C_AIG practice tests help you check understanding of the underlying concepts between hands-on sessions.
What Is Not a Separate Domain
It is easy to over-study. SAP Cloud SDK for AI, orchestration, grounding and retrieval-augmented generation (RAG) appear in broader preparation content, but they are not additional separately weighted exam domains. Learning them can deepen your understanding, yet your time is better spent first mastering the seven official content areas. The same applies to the permissions article: it is preparation guidance, not a domain.
It is also worth separating this credential from neighbors. SAP's C_BCBTP, Positioning SAP Business AI Platform, is the verified current comparison credential. It is a different certification and does not replace the Generative AI Developer credential, so do not blend their content when studying.
Validity and what comes after
The certification is valid for 12 months from the initial exam. To renew, you pass a Stay Certified assessment with an active SAP Learning Hub subscription within the 90 days before expiry, which extends validity by 12 months from the original expiry date rather than the assessment date. Without Learning Hub, you would purchase and retake the full certification exam annually. For career context, see the ROI analysis, though note that salary and return claims require separate evidence beyond what is established here.
Frequently Asked Questions
No single domain can be identified as heaviest. All seven are published at an 11-20% range, and ranges are not exact allocations, so prepare for every area.
No. It is one system-based activity of 3 hours, scored on completed tasks and system outputs, with a 76% passing score. The activity count is not a question or task count.
Practical exams are open-book and unproctored, and public SAP resources and applicable AI tools are permitted. The timer keeps running during breaks, so rely on practice, not lookup.
SAP's sources differ: release notes say advance booking is required, while the FAQ says some systems are immediately available. Plan to book ahead and follow your exam dashboard.
The passing score is 76%. No candidate pass rate is established by the retrieved sources, so see the passing score guide and pass rate article for what can and cannot be said.