All right, AI stories are nice and interesting; however, as SAP posted this week, “Today’s geopolitical environment is defined by uncertainty.” This is moving us toward the necessity of controlling innovative AI solutions in disconnected-from-Internet cases.
Can an Enterprise-Level AI Solution Run In Your Datacenter?
Of course, no problem at all! Let’s think a bit—there is an awesome SAP HANA Enterprise Edition available that has an embedded applications layer, and, in fact, this is a Cloud Foundry solution with all necessary enterprise security mechanisms, all delivered by SAP.
Fig. 1 Skybuffer AI Solution Architecture
Great! So, we have the following on-premise ingredients:
These ingredients allow us to develop SAP Fiori applications using XSA technology directly on the SAP HANA EE application layer.
What is missing?
In a general sense, a few things are missing in SAP HANA EE on-premise to act as your enterprise AI on-premise platform:
Is the on-premise gap closed?
Yes! Starting in September 2024, our Skybuffer AI partner solution, developed on SAP technologies, successfully passed the pilot stage and is now available in the SAP Store in both versions:
We have brought the following missing parts to SAP HANA EE on-premise:
Fig. 2 Skybuffer AI Modeling Dashboard
As a result, you get a ready-to-use platform that integrates Generative AI, RAG, and Conversational AI, allowing businesses to automate tasks quickly and efficiently. It saves costs, reduces development complexity, and accelerates time to value. Everything is on-premise on SAP HANA EE (which you can host even under the table 🙂 ).
Cherry on the cake!
Skybuffer AI is out-of-the-box connected to SAP ECC and SAP S4HANA on-premise systems. We have already developed approximately 400 different integrated business scenarios that allow you to use secure AI assistants connected to your SAP ECC and SAP S4HANA via single sign-on. Yes, another magic—no need to adjust roles and profiles for your users; they get in chat exactly the same information they are authorized to see in the SAP system.
Fig. 3 SAP ECC or SAP S4HANA in Microsoft Teams Channel
What channels are covered?
Communication can be done through many channels, the most popular are:
Everything is just a configuration on the Skybuffer AI platform (on your premises)—no need for Python, Node.js, or any other type of development.
What about AI Agents?
Sure, those nice 24/7 employees are also here! Any AI model from Skybuffer AI can be scheduled to execute monitoring, document processing, or any other task. Each time an AI Agent runs, you see the execution results in the conversation log because the AI Agent uses generative AI to explain what has been done and tag the log.
Fig. 4 AI Agent Generative Log
You can chat with the AI Agent anytime to check the heartbeat of the process. Is it not magic?
Fig. 5 AI Agents Status Check
What about RAG document processing?
Here, we not only supply an RAG solution that is ready to consume without any single line of code. Skybuffer AI also has an RAG Document Viewer that allows your employees to see the documents currently accessible by RAG, check document content, and upload new documents directly from the desktop.
Fig. 6 Skybuffer AI - RAG Document Management
All documents have metadata parameters, so you can classify documents using 48 out-of-the-box string, numbering, and date-type parameters. What you need to do is define your aliases for parameters, so the RAG Fiori application will speak your company’s terms.
What about Joule?
Let’s wait for Q1 2025. Once SAP makes Joule Build available for SAP Partners, we will connect the Skybuffer AI platform to be available via the Joule communication channel as well. In other words, we complement Joule by bringing Skybuffer AI skills directly into Joule’s embedded AI skills.
Interesting and exciting? Agree!
What do you need to run enterprise-level AI on-premise?
Only SAP HANA, which you most probably already have!
Fig. 7 Skybuffer AI - Action Server for no-code AI Modeling
Performance?
Hmmm, that is something really cool. Old, nice on-premise performance, finally 🙂
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
| User | Count |
|---|---|
| 29 | |
| 25 | |
| 24 | |
| 19 | |
| 14 | |
| 13 | |
| 12 | |
| 11 | |
| 11 | |
| 11 |