/Logo%202023%20final%20dunkelgrau.png?width=221&height=97&name=Logo%202023%20final%20dunkelgrau.png)
After SAP Analytics Cloud (SAC), we have now taken a closer look at the AI premium features in SAP Datasphere. Here, SAP is primarily targeting tedious mandatory tasks that take up a lot of time in everyday work. The aim is to speed up searching for the right objects, maintaining metadata and modeling semantics. We put the features to a practical test and explain in this article which features are already convincing today and where SAP should make improvements.
Prerequisites for Activating SAP Datasphere Premium AI Features
The AI premium features are not included in the standard scope of SAP Datasphere but must be activated via the SAP Support Portal. The prerequisite for this is that your account has AI Units and that the tenant is located in a landscape supported by SAP AI Core. The complete process is described in SAP Note 3522010. Once the features have been enabled, you can turn the individual features on and off via the menu path System → Configuration → AI Services. In addition, users must be assigned the Data Warehouse AI Consumer role.
How Does AI-Assisted Search Work in SAP Datasphere?
Everyone is talking about self-service and data mesh concepts again, but in practice, users run into a problem. The sheer number of objects can be overwhelming. Finding the right object requires selecting multiple filters, which takes quite a bit of time.
This is exactly where AI-powered search comes in. Users can phrase their search query in natural language, and AI interprets the request. Multiple conditions are also possible. Instead of setting filters, users simply describe what they are looking for, such as "all views with revenue data at customer level that are based on the S/4HANA connection."
The feature is available in the Repository Explorer, Data Builder, Catalog, and Data Marketplace. With this feature, SAP promises significant time savings of 90% when searching for data artifacts. The search is intended to become not only faster but also more precise. This allows users to focus more on actual data modeling instead of spending time searching for existing objects.
In our tests, the search query was always reliably converted into filter conditions, even for queries with multiple conditions. We can very well imagine this being used in practice and the resulting time savings.
Watch the recording of our webinar:
"SAP Datasphere and the Databricks Lakehouse Approach"
Automating Catalog Metadata Generation in SAP Datasphere
While AI-powered search helps users find objects, metadata ensures that they can also understand them. Meaningful descriptions, tags and clean documentation are the foundation for self-service in the catalog. Only then can users determine whether a given object is suitable for their question. At the same time, well-maintained metadata also improves search results. In practice, however, maintaining this metadata is often neglected because it is time-consuming and requires detailed knowledge of the underlying data structures.
The next premium feature is aimed at data stewards and automates the creation of business metadata in the catalog. At the push of a button, the AI generates descriptions for an asset while taking the business context into account.
Classification can also be automated. The AI assigns appropriate tags to assets, selecting them from the hierarchical tag lists stored in the catalog.
With this feature, SAP promises significant relief in catalog maintenance, estimating time savings of 87 percent. The documentation and classification of assets is becoming possible even without in-depth knowledge of complex SAP data structures and their semantics.
In our tests, the generation of the description and documentation yielded good results. When assigning tags, we had mixed results. Sometimes the AI simply assigned all available tags. Other times, no tags were assigned at all. And even if the description functionality suggests otherwise, the AI assignment of KPIs and terms is not yet available. We see the usefulness of this feature and would like to see improvements in this regard.
Generating Business Semantics for Non-SAP Data in SAP Datasphere
The next premium AI feature simplifies modeling. When data is imported from SAP source systems into SAP Datasphere, it already includes the semantics. However, for custom-modeled objects, this semantic information is missing and must be assigned manually. Here, AI can assist by analyzing the entire object and suggesting the missing semantics.
AI suggests whether the table is a fact, text or dimension table and assigns currency or date fields.
Instead of checking each column individually and setting the semantic properties by hand, you receive a suggestion that only needs to be reviewed and adjusted if necessary.
In our tests, the AI reliably recognized fact, dimension and text tables. Semantic types such as currency, date, language and text, as well as the text association, were also set correctly.
Cost Structure & AI Units Pricing for SAP Datasphere AI Features
As the name suggests, premium AI features are billed separately. For this, you need to purchase so-called AI Units. The list price for one AI Unit is EUR 7. Consumption varies depending on the feature. Semantic generation is charged at 0.2 AI Units per request, AI-powered search at 0.7 Units per 100 requests and metadata generation at 0.29 Units per request. Converted into euros, semantic generation costs EUR 1.40, AI-powered search EUR 0.0049, and metadata generation EUR 2.03 per request.
Consumption can be viewed in SAP for Me under Finance & Legal → Business AI. For users themselves, consumption is not visible at the time of use. We therefore recommend monitoring usage regularly to keep costs under control.
SAP Datasphere premium AI features: Our Conclusion
The premium AI features in SAP Datasphere are not a revolution but targeted optimizations for everyday work. They take over the tedious legwork that costs a lot of time and brings little joy.
In our practical test, the results were largely convincing. We see room for improvement only in the assignment of tags, which still works unreliably and in the announced but not yet available suggestions for KPIs and business terms.
Nothing earth-shattering, but useful helpers that can quickly pay for themselves. We look forward to further features that will simplify our everyday work and are excited to see which functions SAP will deliver next.
Ready to Accelerate Your Data Architecture with SAP Datasphere? Implementing and optimizing SAP Datasphere Premium AI Features requires the right setup and governance. Whether you need help configuring AI Units, setting up SAP AI Core integration, or optimizing your non-SAP data modeling, our team of experts is here to guide you.
