Generative SAP Business AI creates new valuable content to optimize business processes on cloud platforms like SAP Business Technology Platform SAP BTP, Amazon AWS or Microsoft Azure.
SAP Generative AI harnesses Foundation Models (FM) for general purpose tasks or Large Language Models (LLM) to process natural language (NLP) and transform multi modal inputs to new output content.
Foundation models are pre-trained, but not specialized for specific tasks, and can be integrated into versatile business AI scenarios. Generative AI platforms like AWS Bedrock, Azure AI Foundry or SAP Generative AI Hub, which is built on top of SAP AI Core, offer foundation models together with tools to build intelligent Business AI applications.
Foundation models on the SAP Generative Business AI Hub are integrated into the SAP Business Technology Platform architecture to empower business processes with downstream tasks or domain specific business agents. SAP Business AI process steps implement downstream tasks to process multi-modal prompts with business data or documents as input and generate new business content with various output formats like text, audio, speech, code or image.
Based on Generative AI Transformer Architectures, SAP Generative AI Hub foundation models are trained with transfer or self-supervised learning to optimize costly machine learning processes. Transfer learning uses existing knowledge for new SAP Business AI scenarios without retraining and self-supervised training creates autonomously labels for new inputs which can be applied to new downstream tasks.
Customization and orchestration on cloud platforms like SAP Generative AI Hub, Azure AI Foundry or AWS Bedrock empowers GenAI models with AI capabilities like domain specific context, prompt management or enhanced security features.
Large Language Models (LLMs) are foundation models trained on large text datasets to perform natural language processing (NLP) as downstream tasks.The reusability of pre-trained foundation models for different SAP Business AI tasks reduces costly training effort and enables use-cases with small training datasets or machine learning teams.
NLP Business AI scenarios implemented with multi-modal Large Language Models (LLMs) can process language input and generate text, code or image output. Typical SAP Generative AI Hub NLP use-cases are classification, sentiment determination, summarization, comparison or text generation for application areas like autonomous AI assistants or different kind of document processing.
The comparison of SAP Business AI foundation models is one step in the development process of AI-powered solutions to find the best model for a specific Business AI scenario. This decision can be based on GenAI foundation model criteria like parameter count, model size, benchmarks, costs, context window size or latency.
The comparison tool below visualizes information about models available on the SAP Generative AI Hub.
The drop-down above the diagram offers the selection of model families like OpenAI GPT hosted on Azure or Anthropic Claude on AWS Bedrock. With the colored elements on the right, the visibility of the models in the grouped bars can be toggled.
Initially, the foundation models are sorted by a general ranking factor. This sorting can be changed for single selected metrics. On mobile, a double click on one element sets all bars invisible and sorts the first selected element afterwards.
Some popular GenAI model benchmarks, which are evaluated with prepared datasets and questions, are listed below.
| Name | Description |
|---|---|
| MMLU | general mostly knowledge-driven evaluation of various tasks |
| MMLU-PRO | enhanced, more reasoning-focused multi-choice options |
| GPQA | evaluates expert-level reasoning |
| HumanEval | measures the functional correctness of generated code and programs |
| MATH / MATH-500 / GSM-8K | measures arithmetic reasoning from basic arithmetic to advanced mathematics |
| BFCL / NFCL | evaluate the ability to generate arguments to call functions of external tools |
| MTEB | evaluates the performance of embedding models |
The SAP Generative AI Hub is based on SAP AI Core to integrate GenAI deep learning models into SAP Business applications. LLM foundation models are offered with toolsets to engineer LLM prompts and integrate GenAI into business processes.
SAP manages Open Source from MistralAI or Meta and GenAI models hosted on other cloud platforms like AWS Bedrock or Azure OpenAI as SAP AI Core scenarios.
SAP Generative AI Hub offers foundation-models and orchestration scenarios with predefined capabilities to implement GenAI Business AI scenarios. Foundation-models scenarios like azure-openai, aws-bedrock, gcp-vertexai or aicore-opensource group pre-built AI models with versions which can be parameterized in configurations. GenAI configurations can be deployed as instances of serving templates and executables to proxy models hosted on cloud platforms like AWS, Azure or Google.
SAP Generative AI Hub Embedding models:
| Model | Context | Dim | Costs | MTEB |
|---|---|---|---|---|
| AWS Bedrock | ||||
| Titan Embed Text | 8k | 256-1024 | 0.0004 | 66 |
| Azure OpenAI | ||||
| Embedding 3 Small | 8191 tokens | 1536 | 0.00006 | 62.3 |
| Embedding 3 Large | 8191 tokens | 3072 | 0.0003 | 64.6 |
State of the art (SOTA) GenAI models mostly implement transformer architectures with neural networks to remember long-range dependencies and to process whole sentences with positional embeddings and self attention. Based on transformer architectures, foundation models learn context with meaning to generate new business context with reasoning and actions.
Technically, transformer architectures can be separated into encoder and decoder parts, both with attention and feed forward components.
Encoders transform language tokens into coordinates within multidimensional vector spaces of semantic language models. The distance of tokens within these embedding models represent their semantic relationship. Embeddings are used for NLP analysis tasks like summarization, key phrase extraction, sentiment analysis with confidence scores or translation.
Decoder are able to generate new text sequences and enable decoder only Conversational or Generative AI solutions.
Attention layer weights of encoder or decoder components control the choice of prediction results. Encoder attention layer weights try to quantify the meaning of words within text sequences. Decoder attention layers predict the most probable output token in a sequence.
Interaction with Generative AI assistants like Microsoft Copilot or SAP Joule improve digital user experience based on data-driven decisions supported by foundation models or LLMs. Microsoft Copilot combines ChatGPT LLMs with data provided by Microsoft Graph and offers the option to build custom copilots for various business-specific tasks.
The Open Source community offers Large Language Models (LLMs), inference servers and GenAI tools to implement intelligent AI solutions.
Advantages of open-source GenAI are transparency, cost reduction, local solutions to fulfill advanced data security requirements or flexible model customization opens like fine-tuning. Examples of open source GenAI model families are LLama, Mistral or Falcon.
GenAI platforms like SAP Generative AI Hub or AWS Bedrock support training and serving Bring Your Own Models (BYOM) as containerized web applications. GenAI model inference server definitions use serving templates with parameters like number of replicas. Generative AI models can be called via inference request endpoints to return predictions for consuming customer services.