Transforming AI Tools into Real-World Solutions on SAP BTP

[Siarhei]

Capabilities do not always translate into adoption. This is a common oversight when companies consider Generative AI, RAG technologies, and their integration for daily use by employees, partners, or customers.

No Code

No-code solutions are a must for AI technology adoption and utilization in valuable business cases. The ability to create simple implementations without writing a single line of code is essential. By leveraging the integration of Generative AI and RAG, powered by an Action Server with an embedded Conversational AI layer, companies can encapsulate the action server skills of their AI model.

Action Server

The action server is the heart of any AI model, yet it is often overlooked or misplaced. The action server acts as the director, orchestrating interactions with ML models and the vector database. It forms the top layer of the AI model and should include the following components:

• Action Server
• Conversational AI
• Generative AI
• RAG

AI Model

A conversational AI model serves as the metadata layer for memory-enhanced conversational assistant applications. It is an abstract layer where you define:

• The skills of the AI assistant, controlled by logical triggers
• The action groups each skill contains, controlled by logical conditions
• The action types executed when a skill is triggered

No Code? It Means Application

The application layer of an AI solution must enable AI modeling without coding, allowing users to define groups of skills, triggers, and actions while benefiting from the integration of Generative AI and RAG. Conversational AI plays a key role in deriving the intent of dialogues, acting as a source for logical triggers, and serving as the memory of the conversational assistant.

SAP BTP Services

SAP BTP provides all the necessary components for transforming AI tools into real-world solutions. We recommend the following (inspired by the AI Cookbook):

• 1 x SAP BTP Cockpit
• 1 x SAP AI Core
• 1 x SAP AI Launchpad
• 1 x SAP HANA Cloud (including Vector Engine)
• 1 x SAP BTP, Cloud Foundry runtime
• 1 x SAP BTP, Object Store
• 1 x SAP Job Scheduling Service
• 1 x SAP S/4HANA or SAP ECC (as the source for SAP-integrated scenarios)

Using these, we developed Skybuffer AI on SAP BTP—a solution that integrates Generative AI and RAG functionality while providing an SAP Fiori interface for AI modeling. After a successful 6-month pilot phase with leading companies in the USA, Europe, and Asia, Skybuffer AI is now market-ready.

Skybuffer AI: No Code, Real-World Solutions on SAP BTP

Skybuffer AI is a set of SAP BTP partner-developed applications that form a ready-to-use platform integrating Generative AI, RAG, and Conversational AI. It enables businesses to automate tasks quickly and efficiently, saving costs, reducing development complexity, and accelerating time to value.

 Fig. 01 Skybuffer AI. Models Overview

Smooth Migration from SAP Conversational AI: Is It Possible?

It took us six years to develop Skybuffer AI into its current form for SAP customers. The journey began with Recast.AI, which was later renamed SAP Conversational AI. Given the development path of Skybuffer AI, we also offer a one-click migration from SAP Conversational AI, ensuring a smooth and efficient transition to the future-ready Skybuffer AI on SAP BTP.

How Does Skybuffer AI Complement SAP Joule?

While SAP Joule functions as an in-application assistant within SAP S/4HANA and SAP Cloud Applications, Skybuffer AI focuses on addressing custom tasks across both SAP and non-SAP systems. It provides flexible integration through APIs and multi-channel support, allowing organizations to model their AI processes. Together, Skybuffer AI and Joule offer a powerful combination, optimizing both standard workflows and custom use cases.

Conclusion

With over 400 SDK-based, backend-integrated Business AI scenarios, Skybuffer AI covers all lines of business in S/4HANA and S/4HANA Private Cloud Edition, leveraging SAP BTP AI Core and HANA Cloud services. Follow our article series for insights into Skybuffer AI, or contact us for a demonstration in the comments section.