SAP Community - SAP AI Launchpad

Oppt Prioritization score custom Agent in SAP CX AI Toolkit

2025-07-22T17:26:06.300000+02:00

Hi Experts,

I am trying to create a custom AI agent for opportunities called Oppt Prioritization score agent. For this i created an AI Tool by giving the following prompt:

You are an opportunity prioritization agent in the SAP CX AI Toolkit for SAP Sales Cloud V2. Analyze all opportunities with Opprtunity.Status with Open status.

Calculate the prioritization score based on the following weighted criteria:
1. Opportunity.Expected_Value (50% weight)
2. Opportunity.Probability (30% weight)
3. Opportunity.Close_Date within 30 days from the Opportunity.Current_Date (20% weight)
Assumptions
4.Opportunity.Probability is assumed to be a percentage (0-100).
5.The Opportunity.Close_Date condition is binary, based on whether it falls within 30 days from the Opportunity.Current_Date, interpreted as future dates for prioritization.

Priotitization Score = 0.5*{Opportunity.Expected_Value} + 0.3*{Opportunity.Probability} + 0.2*{cases}
1 {ifOpportunity.Close_Date is within 30 days from the Opportunity.Current_Date} else
0
end{cases}

Provide a one-sentence explanation per opportunity justifying its ranking and, sorted by Amount (descending). Use a professional tone and avoid technical jargon.
For each opportunity, provide a brief, actionable follow-up recommendation (e.g., “Schedule a discovery call to discuss budget constraints”). Use a concise, professional tone and leverage intelligent Q&A to ensure recommendations are contextually relevant.
Include a summary of recent interactions for each opportunity, generated using the toolkit’s AI summary capabilities. Use a professional tone and ensure recommendations support team collaboration.
Ensure Output Fits Within 5000 Chatacters
Compact Formatting--> Remove unnecessary spaces & line breaks.
Limit Answer Length--> Max 500 Characters per answer, truncate with "..."

I assigned this AI tool to the custom AI Agent.The details (PFA screenshot) which are used are

Expected value = 15000 USD

Probability = 80

Close Date = 25 Jul 2025

The calculation for close date is incorrect, it should be 1, but it is coming as 0. Any suggestions how can i achieve the condition for the close date?

Gemini API Integration Attempts and Current 404 Issue

2025-08-11T07:17:21.432000+02:00

Hi Developers,

We are encountering the following issue while integrating the Gemini model with SAP AI Core:

404 Not Found is returned when calling the Deployment ID at the root URL, even when using the correct orchestration payload.
The deployment URL has been verified to match exactly.
Gemini API is confirmed to be up and running, and tests using logs captured from network activity match the expected configuration.
Despite this, the API call consistently returns:
"The requested URL was not found on the server. If you entered the URL manually please check your spelling and try again."

Could you please confirm the correct deployment details or orchestration URL for the Gemini deployment, or advise on a possible workaround so we can proceed?

Best regards,
Aditya Srivastav

Gemini API Integration

2025-08-11T07:37:51.276000+02:00

Hi Developers,

We are encountering the following issue while integrating the Gemini model with SAP AI Core:

404 Not Found is returned when calling the Deployment ID at the root URL, even when using the correct orchestration payload.

The deployment URL has been verified to match exactly.

Gemini API is confirmed to be up and running, and tests using logs captured from network activity match the expected configuration.

Could you please confirm the correct deployment details or orchestration URL for the Gemini deployment, or advise on a possible workaround so we can proceed?

Best regards,
Aditya Srivastav

AI launchpad, Error Code: 04000001

2025-08-14T18:41:47.425000+02:00

Hi community,

It seems a mistake when I try to open the SAP AI launchpad, https://sap-learning-dev-venusaur.ai-launchpad.prod.eu-central-1.aws.ai-prod.cloud.sap/aic/index.html#/workspaces&/a/detail/TwoColumnsMidExpanded/?workspace=aicore-dev&resourceGroup=default

When I click 'view detail', it shows: it looks like something went wrong. Please check your details and try again.

Error Code: 04000001
Message: An API error is preventing access to the AI Core Administration app. Contact technical support.

When I used Python code locally to connect to the AI core and call the previously deployed AI Launchpad model, it was still possible to run.：

deployment_url = "https://api.ai.prod.eu-central-1.aws.ml.hana.ondemand.com/v2/inference/deployments/df48994be265dea9"

When I choose another api connection in API connection, it works. That means if I choose another api connection, I can see workspaces. generative ai hub, sap ai core administration and ml operations on the left side.

This seems to be a problem with the front end. Whenever I click on any resource group, a 500 error is reported. The backend of Launchpad failed to call AI Core. The error did not occur in the browser, but within SAP Launchpad itself.

What should I do?

Need guide

2025-08-18T10:23:23.417000+02:00

I am currently using Anthropic AI in my project, but I need to deploy it in SAP AI Core. Since Gen AI is not included in my plan, I want to understand the recommended approach to integrate and run my model using SAP AI Core instead of the open-source or Gen AI options

About: SAP Certified Associate - SAP Generative AI Developer

2025-08-21T07:59:54.645000+02:00

Hi,

I am learning the course [SAP Certified Associate - SAP Generative AI Developer].

My trial account is below:

I want to set up Generative AI Hub in SAP AI Core, But I can't find out the item [SAP AI Core] in the Entitlements.

Could you help me to resolve this problem?

Best Regards!

Unable to copy paste a value in SAP AI Launchpad

2025-08-21T20:47:54.318000+02:00

Hello There,

I am trying following tutorial:

https://developers.sap.com/tutorials/ai-core-orchestration-consumption.html

In step 4, when I try to paste the default value for the variable then paste action is not working. Also, if I try to add the default value manually, then also I get stuck because "Add" button is not shown.

Could you please suggest me next steps?

Thanks,

SAP AI Launchpad Grounding with Sharepoint: Failed to authenticate document grounding api

2025-08-25T08:10:47.294000+02:00

Hi,

created a Generic Secret to my Sharepointsite.
Than I want to create Data Repository under "Grounding Management".
I can choose my created generic secret and when I press the button "Create" I got the following error:

"AI API call failed. Check the error log for details.
Error Code: 401
Message: Failed to authenticate document grounding api. Make sure to have a valid token."

I'am shure, that my clientId and secret is working, because I can fetch a token with postman.

What can be wrong?

Greets,
Andreas

How to save a prompt template in SAP AI launchpad?

2025-09-08T12:55:49.357000+02:00

Hi,

Im not able to add the version for the prompt template in AI launchpad, since there is an issue with the version number?

Can someone please guide me on this

Thanks,

Rithika

RBAC: access denied #SAPAICORE

2025-09-24T08:09:07.252000+02:00

Hello Everyone, I am on a free tier and I am facing RBAC: access denied after deployment of a chatbot in sap ai core

POST

https://api.ai.prod.eu-central-1.aws.ml.hana.ondemand.com/v2/inference/deploymen/<dep_id>predictions

Headers -

Key: AI-Resource-Group

Value: default

Key: Authorization

Value: Bearer <token>

Body is set

Facing RBAC: access denied

SAP GENAI gemini Deployment URL is not able to connect in POSTMAN

2025-09-26T05:46:55.907000+02:00

Hi,

I have a deployment for Gemini in the AI launchpad .
I am getting below error while testing it in Postman

{

"error": "NotFound",

"message": "The model `` does not exist."

}

Below are the details I am providing as the input
url : https://api.ai.prod.us-east-1.aws.ml.hana.ondemand.com/v2/inference/deployments/<deployment_id>/chat/completions/?api-version=001

inside body

{

"messages": [

{

"role": "user",

"content": "Sample prompt"

}

]

}

model : gemini-2.5-pro

Can someone please let me know what I am missing or doing wrong.

Thanks in advance.

Vibe Coding with MCP Servers & SAP AI Core: Toward "Coding by Conversation"

2025-09-29T15:36:36.701000+02:00

Over the past few months, I’ve explored vibe coding — a workflow where you prompt the AI, it generates code, and you iterate. Fast and interactive, yes — but in enterprise projects, context and governance are critical. Without them, hallucinations and mismatched code structures quickly emerge.

This is where MCP servers (Model Context Protocol) and SAP AI Core become essential. Tools like fiori server and cap server allow AI assistants to understand your CDS models, Fiori metadata, and application structure — enabling contextual, reliable code generation.

In this blog, we’ll explore the architecture, practical usage, and enterprise best practices that make AI-assisted coding robust and production-ready.

What is Vibe Coding?

Vibe coding is a transformative approach to software development — essentially, coding by conversation.

You describe requirements in natural language.
The AI assistant interprets your intent and proposes code patches, scaffolding, or tests.
You review, refine, and approve the generated output.

The power of vibe coding lies in freeing developers from repetitive boilerplate tasks, allowing them to focus on high-value work like architecture, user experience, and business logic.

SAP’s reference architecture for vibe coding with Cline + SAP AI Core demonstrates how AI-assisted development can be safe, repeatable, and fully aligned with enterprise standards — turning AI from a helper into a structured development partner.

Core Principles from the SAP Architecture

From the reference architecture, five principles stand out:

Prompt-Driven Development – Express features in natural language, and let the system automatically generate the initial scaffolding.
Context Engineering – Supply LLMs with only the necessary structural context, such as APIs, CDS models, or MCP metadata.
AI-Assisted Test-First Approach – Begin by generating tests, then guide the AI to produce code that fulfills those requirements.
Human-in-the-Loop Governance – Developers retain responsibility for ensuring compliance, performance, and security.

Introducing MCP Servers

MCP (Model Context Protocol) acts as the bridge between your project and the AI assistant.

MCP (Model Context Protocol) is designed to act as a powerful bridge between your project and the AI assistant, ensuring that development is not just driven by prompts but also grounded in real project context. Instead of leaving the AI to guess or hallucinate about how your application is structured, MCP Servers provide it with a deep, contextual understanding of your project.

They expose key elements such as metadata, annotations, models, services, entities, relationships, and queries. This allows the AI to reason intelligently about your project’s architecture, whether it’s related to the user interface, the underlying data definitions, or the service layer.

With MCP servers running, your AI assistant doesn’t hallucinate blindly — it knows your project structure.

Architecture in Action: Cline + SAP AI Core + MCP

Here’s how the pieces fit:

Cline –Your coding agent inside the IDE, mediating prompts, code diffs, and project changes.
SAP AI Core – Hosts the foundation models (Claude, OpenAI, Gemini) in a governed, enterprise-ready environment.
MCP Servers – mcp server expose structured context:
- UI annotations, metadata, and manifest structure
- Style guides and constraints
- CDS entities and services

Workflow:

You prompt: “Create a ListReport for SalesOrder with a status field.”
AI (via MCP) inspects CDS models and annotations.
AI generates the CDS extension, Fiori annotations, and manifest changes.
You review, test, and deploy.

This architecture transforms AI coding from a “toy experiment” into a repeatable, enterprise-grade engineering practice.

How to Get SAP AI Core Credentials

Access SAP BTP Cockpit: Go to your SAP BTP subaccount
Navigate to AI Core: Services → Instances and Subscriptions → AI Core
Create Service Key:
- Go to your AI Core service instance
- Create a new service key
- Copy the credentials (clientid, clientsecret, url, serviceurls)
Configure Authentication URL: Usually follows pattern https://{subdomain}.authentication.{region}.hana.ondemand.com/oauth/token
Set Base URL: Typically https://api.ai.ml.hana.ondemand.com (check your service key for the exact URL)

SAP_AI_CORE_CLIENT_ID=your-client-id
SAP_AI_CORE_CLIENT_SECRET=your-client-secret
SAP_AI_CORE_AUTH_URL=https://your-subdomain.authentication.sap.hana.ondemand.com/oauth/token
SAP_AI_CORE_BASE_URL=https://base-ai-core-url
SAP_AI_CORE_RESOURCE_GROUP=your-resource-group

Refer to the screenshot below, which shows how we subscribed to the SAP AI Launchpad application to deploy our model according to the use case. We also created an SAP AI Core instance to obtain the necessary key details required for configuration in the Cline client.

Cline Configuration

Install the Cline extension in VS Code
Open Cline settings
Configure these fields:

API Provider: Select "SAP AI Core"
Client ID: .clientid
Client Secret: .clientsecret
Base URL: .serviceurls.AI_API_URL
Auth URL: .url
AI Core Resource Group: Extract UUID from .appname (first part before the !)

4. Test with a simple query to make sure it works

Add Server Configurations

To enable AI-assisted coding with full context, add the following server configurations to your cline_mcp_settings.json file. This configuration allows Cline to connect to the MCP servers that expose structured information about your project, including CDS models, Fiori metadata, manifest files, and coding style constraints. By providing these settings, you ensure that the AI assistant can:

Understand project structure: MCP servers give the AI visibility into your CDS entities, services, and UI annotations, allowing it to generate code that aligns with your existing models.
Follow enterprise standards: Style guides, naming conventions, and other constraints are passed through MCP servers, ensuring that generated code adheres to your organization’s best practices.
Generate context-aware code: Instead of generic snippets, the AI can create extensions, annotations, or manifests that fit seamlessly into your project.
Improve efficiency and reduce errors: By providing a structured context, the AI minimizes mismatches and hallucinations, letting developers focus on architecture, business logic, and user experience rather than repetitive boilerplate.

With the server configurations in place, Cline can interact with MCP servers in real time, making AI-assisted coding a safe, reliable, and enterprise-ready process.

{
  "mcpServers": {
    "cds-mcp": {
      "command": "cds-mcp",
      "args": [],
      "env": {},
    }
  }
}

Set Guardrails with Cline Rules (How-To)

Purpose: Provide persistent, project-level guidance — covering scope, coding style, documentation standards, and more — that’s always active in your workspace.

Cline Rules act as a continuous, automated advisor for your development team. By defining guardrails, you ensure that AI-assisted coding adheres to enterprise standards and project-specific conventions. This prevents inconsistent implementations, reduces technical debt, and maintains code quality across the team.

Create a Rule File
1. In Cline’s Rules tab, click “+” or use the command: /newrule.
2. Save as a single file: .clinerules at the repo root, or use a folder-based setup .clinerules/ for multiple rules.
Recommended Folder Structure

your-project/
├── .clinerules/
│   ├── 01-project.md      
│   ├── 02-coding.md
│   └── 03-documentation.md  
├── docs/

Each file addresses a specific aspect of project guidance: project scope, coding standards, and documentation expectations.

Example Prompt Flow

You: "Initialise an empty SAP CAP application in this folder."
AI: The empty CAP application has been successfully initialised in the current folder

You: "Generate an entities sales and sales order items"
AI: The entities is successfully created.

Result: you stay focused on implementing business logic and refining the user experience, while the AI assistant handles all the scaffolding, boilerplate, and repetitive code generation.

Open the Cline application and start the conversation with the above commands — this is where vibe coding truly comes alive, as your AI assistant (powered by MCP servers and SAP AI Core) begins generating context-aware code that you review, refine, and evolve into enterprise-ready solutions.

Once you select the Plan and submit the prompt, Cline (with MCP servers in the loop) interprets your intent, fetches the right CAP/Fiori context, and proposes structured code changes that you can review and refine into a production-ready solution.

You will now see a fully scaffolded CAP (Cloud Application Programming) application generated — with entities, services, and annotations created automatically based on your prompt, thanks to the MCP servers providing structured context to the AI assistant.

Benefits & Caution

Benefits

Faster Prototyping and Scaffolding – Quickly generate CAP entities, Fiori views, and annotations without manual boilerplate work.
Reduced Hallucinations – Structured context from MCP servers ensures AI proposals are relevant and accurate.
Freed Mental Bandwidth – Focus on UX, domain logic, and performance optimization, while AI handles repetitive scaffolding.
Aligned with SAP Standards & Governance – Generated code respects enterprise best practices and architectural guidelines.

Caution

MCP Servers Are Evolving – Always review AI-generated code before deployment.
Complex Business Logic Needs Oversight – Human expertise is essential for transactional integrity, security, and edge-case handling.
Prompt Precision Matters – Well-crafted prompts yield better, more reliable results.
Governance Remains Your Responsibility – Testing, security, and compliance cannot be fully delegated to AI.

Conclusion

Vibe coding isn’t just about writing code faster — it’s about reimagining how we build enterprise software. By combining AI assistants, SAP AI Core, and MCP servers, we provide AI the context it needs to act as a true development partner.

This approach lets developers move beyond boilerplate work and focus on what matters most: architecture, business logic, user experience, and innovation

Human expertise remains essential. Governance, compliance, performance tuning, and creativity stay firmly in the developer’s hands. But with vibe coding, tedious and repetitive tasks are no longer blockers — they’re delegated.

As these patterns mature, this may be remembered as the moment enterprise development evolved from “writing code” to “designing conversations.” The future of coding in SAP isn’t just AI-assisted — it’s context-aware, collaborative, and human-centered.

AI core unable to load resource groups

2025-10-13T17:47:06.653000+02:00

Hello Experts,

I have admin access to AI Core but getting below error after setting up workspace. Please suggest what could be wrong.

Error - Unable to retrieve the resource group.

SAP Generative ai certification

2025-10-16T16:37:54.114000+02:00

Please, anyone, guide me on old Sap generative AI certification.
How-to is helpful for SAP FI consultants.

SAP Access Management Use Cases can be enabled by SAP Gen AI

2025-10-17T16:39:58.752000+02:00

SAP Access Management Use Cases can be enabled by SAP Gen AI

1. Real-Time SoD Conflict Detection
Use Case: Instantly identify Segregation of Duties (SoD) conflicts for any user.
Example: “What conflicts exist for user Jacob?”
Benefit: Immediate, accurate conflict analysis.

2. Risk-Based Role Review
Use Case: Review roles based on actual risk exposure and usage.
Example: “Show me users with SOD risk conflicts”.
Benefit: Prioritize remediation based on real impact, not theoretical risk.

3. Usage-Based Role Optimization
Use Case: Identify unused roles or transactions to streamline access.
Example: “Who has Z_FB02 but never used FB02?”
Benefit: Reduce access footprint and license cost by removing unused permissions.

4. Impact Simulation for Role Changes
Use Case: Simulate the effect of removing a transaction or role.
Example: “If I remove FB02 from Z_FB02, who else is affected?”
Benefit: Avoid unintended disruptions by previewing changes.

5. Alternative Role Suggestions
Use Case: Recommend safer role alternatives.
Example: “What alternative roles could work instead of Z_FB02?”
Benefit: Accelerate remediation with validated, low-risk role options.

6. Executive Risk Reporting
Use Case: Explain access risks in business terms to leadership.
Example: “Explain our P2P risks to the CFO”
Benefit: Translate technical risks into business impact scenarios (e.g., fraud potential).

7. Visual Risk Mapping
Use Case: Generate visualizations of risk across business processes.
Example: “Create a visualization of the risk in the process flow”
Benefit: Improve stakeholder understanding and audit readiness.

8. Automated Audit Response
Use Case: Generate audit-ready reports and remediation plans.
Example: “Generate a slide deck for our SoD remediation progress”
Benefit: Save time and ensure completeness in compliance reporting.

9. Access Governance Across Hybrid Landscapes
Use Case: Manage access across on-premises and cloud SAP systems.
Example: “Show SoD violations across Cloud SAP S4 Hana systems and ECC systems”
Benefit: Unified view of access risks across environments.

10. Token-Efficient AI Conversations
Use Case: Enable deep AI reasoning without exceeding context limits.
Example: “Summarize top 5 risks by transaction volume”
Benefit: Fast, cost-effective AI interactions using pre-analyzed data.

🧭 Strategic Shift for Security Teams
Instead of being data gatherers, teams become strategic risk managers:
Focus on decision-making, not data extraction.
Use AI to triage, simulate, and remediate access risks.
Maintain audit readiness with minimal manual effort.

Generative AI Powered IAM Business Role Augmentation (GAIRA)

2025-10-17T16:43:15.834000+02:00

Generative AI Powered IAM Business Role Augmentation (GAIRA) is a transformative concept for modern enterprises, shifting Identity and Access Management (IAM) from manual, reactive processes to a proactive, intelligent, and efficient system.

Limitations of traditional IAM include:

- Manual Provisioning: Managing user accounts and privileges can be time-consuming and error-prone, particularly in large organizations.
- Inflexible Security: Rule-based systems often lack the context needed to detect subtle, behavioral-based anomalies indicative of modern threats, such as insider risks.
- Compliance Burden: The manual nature of access reviews and audit trail generation drains resources and struggles to keep pace with regulatory changes.
- Subpar User Experience: Slow access request workflows can frustrate users and hinder productivity.

GAIRA addresses these challenges by integrating generative AI into the four pillars of IAM:

1. Adaptive Authentication
- Risk-Based Access: GAIRA enables continuous, risk-based authentication by analyzing user behavior, location, and device in real-time. For instance, a login attempt from an unusual location could trigger additional verification steps.
- Frictionless Security: This dynamic security approach balances convenience and trust, enhancing security without burdening trusted users.

2. Intelligent Authorization
- Automated Role Management: Generative AI analyzes historical access data and user roles to recommend optimal role assignments, significantly reducing manual effort and errors.
- Contextual Access Recommendations: For access requests needing approval, GAIRA provides context-rich insights based on peer-group comparisons and historical outcomes, facilitating faster, informed decisions.

3. Streamlined Administration
- Lifecycle Automation: GAIRA automates the entire user lifecycle, from onboarding to offboarding, dynamically assigning and removing access rights based on predefined policies and roles.
- Enhanced User Experience: Users can interact with the IAM system using natural language via conversational AI interfaces, improving productivity and reducing the burden on IT help desks.

Please help to get prequisite for Joule studio for creating agents

2025-10-30T07:32:37.040000+01:00

Hi All,

Please help to understand what new license/system to be procured/provision to start using Joule studio for creating agents.

We have Datasphere and SAC tenant, as i understand we need to procure BTP, SAP builds and then provision Joule within SAP Build. And how we can use/integrate SAP BDC with Joule studio.

i have referred the help section and some blogs on this. Please help me to get expert advice on this.

Regards,

Prakash

Exploring AI in the ERP Ecosystem – Especially in the SAP FICO Space

2025-11-11T21:53:22.662000+01:00

Hi Everyone,

I wanted to start an open discussion on how Artificial Intelligence (AI) is shaping the ERP ecosystem, especially in the SAP FICO area.

We all can see that automation and intelligence are becoming core to financial operations — from predictive accounting to intelligent reconciliations, cash forecasting, anomaly detection, and even automated decision support.

But I’m curious to hear from everyone here:

How do you see AI blending with traditional FICO processes?
What are the most promising areas for AI-driven improvements in Finance and Controlling?
Are there real examples where AI has already made a visible impact in your projects?

No specific questions here — just an open space for sharing ideas, thoughts, experiences, and even challenges you’ve faced while exploring or implementing AI capabilities in ERP or FICO.

Looking forward to hearing your perspectives and insights!

Best regards,
Eswar Nekkalapu

Unable to save prompt templates in SAP AI Launchpad

2026-02-12T16:15:16.795000+01:00

Hi All,

I was trying to experiment with some prompt templating in SAP's AI Launchpad. I was following a course from the Learning hub, where the following step was there, which requires us to save a prompt template:

In my launchpad, I do not see this option to save or run a template at all:

I have all the necessary roles (prompt_manager, prompt_experimenter) required to save and manage prompts.

Has this features been permanently removed? What is the alternative method to save these templates? Is there any other step that I am missing, or are any additional configurations required to save templates?

Regards.

Guruprasad GL

SAP AI Launchpad add BTP Azur Object Store

2026-02-20T12:55:14.349000+01:00

Hello,
I'm working on the "Predictive AI with SAP AI Core Ingest" tutorial and in the session "Live Data into your House Price Predictor with SAP AI Core" you have to create a Cloud Storage. I'm a BTP user, so i created a Object Store in my Cloud Foundry . My Subaccount is hosted in eu20 so a got a Object Store in the Azure Cloud. I uploaded the files with azcopy.
When i create the Object Store Secret in der SAP AI Launchpad i get this message

but i just have this URL.

Is possible to add a BTP Object Store in the SAP AI Launchpad?