Business AI Services on SAP BTP and Hyperscaler environments like AWS Bedrock or Azure AI Foundry empower SAP Business Suite with AI-driven insights.
Business AI Services offer intelligent capabilities in areas like Computer Vision, Natural Language Processing (NLP) or Generative AI, based on specialized or general purpose machine learning models.
Pre-trained Generative AI Services available on hubs like AWS Bedrock or Azure AI Foundry can be customized to perform domain-specific tasks in intelligent SAP Business Processes.
Cloud AI services can be orchestrated as pipelines with multiple steps to implement AI-powered end-to-end SAP business processes from data extraction to model inferencing.
Intelligent Document Processing (IDP) workflows are widely used in various SAP Business processes like Central Invoice Management (CIM) or S/4HANA Cloud Sales Order Automation.
Intelligent Document Processing automates data processing of text or image based business documents and integrates the extracted information into digital business processes. IDP combines processing steps into pipelines which can be implemented with Business AI services on multi-cloud platforms.
OCR capabilities for IDP can be implemented for instance with Azure AI Foundry Document Intelligence or Vision AI services.
Document Information Extraction (DOX) is the document intelligence service on the SAP Business Technology Platform (SAP BTP) which can be integrated into SAP Business AI processes.
DOX pipelines forward outputs, word boxes with text and spatial information, of OCR document processing as input to SAP Charmer transformer models, where neural networks detect lines in this OCR output, similar to convolutional encoders of the Chargrid algorithm.
Attention layers of the SAP charmer transformer model focus on information of interest and groups of similar information to extract line items. SAP charmer annotation functions assign labels to text data of the input document and transformer decoders extract the text from each line for further AI-driven downstream processing with automated workflows or intelligent apps.
AWS offers various managed AI services to realize intelligent solutions which can be integrated into SAP S/4HANA Business scenarios. AWS Bedrock and AWS SageMaker AI are managed services of the AWS Generative AI stack. Compared with AWS Bedrock, AWS SageMaker AI offers enhanced flexibility with the option to implement custom models which results in more development complexity and effort.
AWS Bedrock is a fully managed Business AI service which makes generative AI foundation models of various providers available through APIs. SAP S/4HANA Cloud processes can implement serverless Business AI cloud solutions with AWS Bedrock, without the need of infrastructure management, with advanced networking and highest security requirements.
AWS Bedrock offers fine-tuning capabilities to customize pre-trained foundation models, with custom labeled training datasets, to perform domain specific tasks. Additionally, Amazon Bedrock knowledge bases implement Retrieval Augmented Generation (RAG) to enhance Generative Business AI with contextual information. RAG can be used to augment chat or instruction prompts and to empower orchestration plans of AWS Bedrock Agents with context, to perform tasks in Business AI processes.
To optimize AWS Bedrock Business AI solutions, invocation logging can be switched on for model input and output to monitor and audit model usage.
AWS SageMaker AI is a fully managed service to manage complete machine learning lifecycles from data preparation, model building, training, fine-tuning to model deployment for inferencing. AWS SageMaker AI offers different inferencing options for Business AI models with real-time or near real-time latency, batch processing or serverless inference without managing the underlying infrastructure.
AWS SageMaker AI JumpStart empowers the implementation of Business AI solutions with a machine learning hub of fully customizable, pre-trained models which can easily be deployed into production. AWS SageMaker Canvas provides a no-code environment for business AI implementations.
AWS offers different options to integrate vector database into Business AI solutions like the managed OpenSearch service which offers scalable index management and nearest neighbor search capabilities.
Responsible AI refers to standards of responsible practices to mitigate potential risks and negative outcomes of Business AI apps with fairness, explainability, security and transparency as core dimensions
Amazon AWS Bedrock Guardrails implements safeguards with contextual grounding and automated reasoning checks for GenAI foundation models (FMs) to reduce hallucinations, block undesirable topics and harmful content.
Amazon AWS SageMaker Clarify analysis inputs and outputs of generative AI foundation models in business AI scenarios. AWS Clarify helps to detect bias and offers reports with metrics to improve the transparency and explainability of GenAI models to meet regulatory and responsible AI requirements.
Interpretability of Business AI models enables humans to explain the output and depends on the complexity of the model.
AWS Business AI transparency features are AWS AI Service Cards with information about use cases and best practices for AWS services and AWS SageMaker Model Cards with documented and cataloged information about Business AI models.
Business AI monitoring can be implemented with Amazon AWS SageMaker Model Monitor to detects inaccurate predictions or as human review with Amazon AWS Augmented AI (A2I).
Amazon AWS managed security services implement defense in depth strategies on different layers for generative and narrow business AI apps.
AWS SageMaker AI offers advanced privacy and security options for Business AI solutions with custom managed virtual networks and limited internet access to ensure that all data is encrypted and does not leave the AWS virtual private cloud (VPC).
To identify and protect sensitive data like personally identifiable information (PII) or financial data, Amazon AWS Macie scans data in AWS S3 buckets.
AWS Cloud Trail logs API calls to protect against unauthorized access to Business AI solutions and Amazon CloudWatch tracks usage metrics to implement alarms for thresholds. AWS Artifact can be used to access compliance reports of idependent software providers (ISV).
Amazon AWS Inspector is a vulnerability management system which performs automated security checks based on best practices and common vulnerabilities.
Azure AI Foundry provides a platform to organize projects for Business AI solutions from experimentations to productive deployments. Generative AI models from multiple vendors are available within the Azure AI Foundry model catalog and can be integrated with the Azure AI Agent service into Business AI apps.
Azure AI Foundry supports three model types Azure OpenAI models, models as a service and open or custom models with different deployment targets from fully to custom managed compute environments. Independent operating agentic Business AI apps can be built in AI Foundry projects with GenAI model selection, tool access and knowledge bases.
As part of the Azure AI Foundry platform, the Azure AI service suite provides prebuilt APIs for common AI scenarios. AI Seevices are offered as single or multi-service resource with RESTful APIs to be integrated into Business AI. Bundled Azure AI Services require only one key and url for combined language, vision or search AI solutions.
Azure AI Foundry support the complete machine learning lifecycle from data preparation, training, deployment, interferencing and model evaluation with integrated machine learning frameworks such as MLflow.
Azure AI Search offers cognitive search capabilities as combination of AI with indices which define schemas defined as JSON structures. Knowledge mining creates a searchable knowledge store from huge amounts of structured or unstructured data.
Indexer import data from external data sources to create indices as searchable content. Shared private links for are premium features to enable secure outbound calls to Azure PaaS resources.
Document cracking is the first stage in the index creation process which includes the opening of files and extracting content.
Enrichment pipelines integrate built-in skills like OCR, translation or AI Language capabilities into skillsets to provide insights which can be stored in knowledge stores.
Customer-Managed Keys (CMK) require Azure Key Vaults and increase index size with query times.
Increasing replica or partition size can help to resolve performance issues like throttling errors with HTTP errors 503 on service or 207 on index side.
Azure AI Vision offers services with pre-trained models to realize OCR, Image Analysis, Face and Video Analysis AI solutions.
Optical Character Recognition (OCR) offers text recognition or text extraction capabilities for images. Image analysis returns phrases as image description with confidence scores.
The Microsoft Florence foundation model is a pre-trained general model on which you can build multiple adaptive models for specialized tasks like image classification, object detection, captioning or tagging.
Azure AI Custom Vision is an image recognition service that allows you to build and deploy your own image models to predict labels or detect options with supervised machine learning. Models converted to compact domains can be retrained and exported for offline usage.
Safety system layer includes options for content filters to suppress prompts and responses based on security levels.
Azure AI Language offers language understanding and analyzing features to realize cloud AI solutions with Natural Language Processing (NLP) capabilities. Some of these preconfigured, customizable NLP features are Named Entity Recognition (NER), Personally Identifiable Information (PII) detection, Language detection, Summarization, Key Phrase Detection and Question Answering.
Conversational Language Understanding (CLU) enables the implementation of bot capabilities like intent prediction and important information extraction from incoming utterances.
Workflow orchestration models connect bots with Conversational Language Understanding (CLU) or Question & Answering projects.
Azure AutoML identifies best algorithms and parameters for specific use-cases automatically to create new or customize existing machine learning models with no-code tools.
AutoML automates iterative machine learning development runs with scoring and ranking by specified metrics. The explain best model feature ensures that the model meets the transparency principle.
Azure AI Services provide two subscription keys to enable regeneration without service interruptions which can be securely stored in Azure Key Vault. These subscription keys can be accessed by clients to initiate token based authentication.
Azure AI services support Microsoft Entra ID authentication with managed identities or service principals, with the difference that managed identities can only be assigned to Azure resources. System-assigned managed identity are coupled to the lifecycle of their linked resource, in contrast to user-assigned managed identity which exist independently of any single resource.
Role Based Access Control (RBAC) enables least privilege security like restricted rotation of subscription keys with contributor roles.
Network access to Azure AI services can be restricted for selected Azure networks, using private endpoints or with Firewall settings for internet or on-premise access. Private Link connections to private endpoints ensure that traffic remains in the Azure backbone.
Service Tags represent groups of IP address prefixes of Azure services to create security rules and routes in network security groups.