AI in SAP BTP: Q3 2023 Highlights – SAP AI Business Services, SAP AI Core and SAP AI Launchpad
Dear SAP Community,
As stated by SAP CTO’s Juergen Mueller, SAP Business Technology Platform acts as the enabler for SAP and partner applications to leverage AI capabilities, including generative AI, in an SAP context.
With SAP Business AI, your company can benefit from relevant, reliable, and responsible built-in AI for your core business processes.
SAP Business AI also enables companies to infuse AI capabilities available on SAP Business Technology Platform (SAP BTP) into their apps with scalability and responsibility in mind, so they can run business faster and smarter:
- Build AI into your SAP applications with a library of pre-trained models using SAP AI Business Services
- Deploy and run AI models at scale without compromising data privacy using SAP AI Core
- Manage the AI model lifecycle in one central place using SAP AI Launchpad
Ready to ride the AI wave with us?
This quarter, our team introduced new product innovations for the AI portfolio of SAP Business Technology Platform, helping your IT teams to make your business applications and processes more intelligent.
Navigate to the corresponding sections by clicking on the product name.
- SAP AI Core
- Data Attribute Recommendation, part of SAP AI Business Services
- Document Information Extraction, part of SAP AI Business Services
- Personalized Recommendation, part of SAP AI Business Services
Confidently deploy and integrate AI models designed for SAP applications, cost-efficiently at scale while preserving privacy and compliance.
Visual Studio Code extension for AI-driven development with SAP solutions
Streamline your AI development process by connecting your SAP AI Core instance directly with VS Code, eliminating the need for multiple tools. You can leverage your existing Docker, Git, and object storage setups for training data and trained models, speeding up your work by directly leveraging your content.
Main benefits include:
- Connect to your instance of SAP AI Core directly from Visual Studio Code (VS Code)
- Register your docker registry, synchronize your AI content from your Git repository, and register your object store for training data and trained models
- Manage your AI scenarios by:
- Creating and registering artifacts in SAP AI Core from your object store connector
- Creating configurations and running and managing executions as well as deployments
New SAP AI Core Learning Journey
Designed for intermediate users, this brand-new SAP AI Core learning journey is now available on SAP Learning for free. In just 4 hours, learn step-by-step how to:
- Embed AI into SAP solutions
- Configure SAP AI Core and SAP AI Launchpad on SAP BTP
- Train an AI model
- Serve an AI model
Perform classification and regression tasks.
Enablement of transfer learning*
We introduced a new model lifecycle where it is no longer necessary to upload the whole dataset for model retraining. It enables users to upload only a delta dataset to the service to decrease the training time to update an existing trained model.
This feature helps to:
- Increase usability by only uploading new training data for model retraining
- Decrease training time
Explainable AI: Exposing feature contribution after model training*
Get feedback on how features contribute to a model, such as not increasing the model performance and indicating how well individual features correlate to defined labels of a model.
This feature helps to:
- Expose feature contribution after model training gives insights on trained features
- Enable transparency featuring a high and low correlation for specific labels
- Decrease debugging time after training a model as the feature contribution for more insights
Updated batch file upload for inference*
Users can now expand the current maximum batch file upload size of 50 for an inference call, and send more inference requests in a single call.
This feature helps to:
- Increase inference batch file upload size
- Decrease the need to split inference calls into multiple batch files
Automate and scale your business document processing for a wide range of document types using machine learning.
Availability of API end points for creating schemas
Users can now connect directly to the corresponding dedicated API end points for schema creation and create schemas from custom applications.
Schema versioning – edit, update, and create new schemas using an existing schema name
Users can now able to create a new schema under an existing schema name, edit and update an existing schema through versioning, leverage an optional parameter as a version in all schema API calls and add version parameters when specifying a template.
This feature helps to:
- Obtain a better overview of what you are doing, by eliminating the need to create multiple copies of the same schema
- Use the latest version of the schema to process documents, which is possible through one schema with multiple versions.
Improved SAP managed global model for processing supplier invoices
The machine learning model for the extraction of invoice documents has been improved and yields better extraction accuracies for many fields.
In addition to the better handling of new layouts, the processing of the following fields has been optimized:
- Better cleansing of bank account numbers simplifies and improves the business partner matching in downstream applications or when using the enrichment feature in Document Information Extraction
- Better processing of amounts with non-standard formats
- Better parsing of dates, in particular numerical dates with whitespaces
Improved SAP managed global model for processing payment advice notes
A new machine learning model for the extraction of payment advice notes has been rolled out providing a higher accuracy for all fields on a variety of layouts.
The results are more consistent within tables and across pages of multi-page documents, especially for amount fields. Smaller fixes for tax IDs and bank account numbers improve the business partner matching in downstream applications or when using the enrichment feature in Document Information Extraction.
New schema property to define user friendly names for your schema fields
Users can select the “label” property in Schema API payload to set the property for the fields.
If the label is available for your schema fields, it will be used for displaying extraction results in the Document UI.
Embedded support in the UI
Users can submit “Service Now” tickets directly through the Document Information Extraction UI.
They have to submit a short description of the ticket and have the UI code automatically collect the information required for the ticket.
Get accurate recommendations based on users’ browsing history and/or item description using machine learning.
Complementary recommendations for end users
We introduced a new API endpoint allowing administrators to provide additional options to their end users based on items often selected together to:
- Improve retention and engagement with end users and meet KPIs more quickly
- Recommend extra items on top of other items already recommended, through the upselling channel
Watch this video to learn more 👇
Support for consideration of the types of user interactions in boosting items
This new feature enables administrators to define a weight for interaction events during recommendation inferences, take the end-user’s reactions to browsed items into account, and then change the prioritization of specific items in the recommendation results.
- Be able to manage recommendations at a finer level of granularity
- Tailor items being pushed to end users to better fit the business strategy
Stay tuned for AI at SAP TechEd – November 2-3, 2023
As every year, SAP TechEd brings together developers, practitioners, enterprise architects, and global IT leaders.
Joined by SAP experts and partners, they’ll unite to explore the latest innovations in AI, app development tools, clean core for cloud ERP – and much more.
Register now for the in-person event in Bangalore, India, or the interactive virtual event not to miss anything about latest SAP Business AI innovations.
In the meantime, follow our AI tags on SAP Community and start interacting with your peers today.