SAP lays out generative AI strategy

German software giant SAP has outlined its strategy for leveraging generative artificial intelligence (GenAI) to assist businesses and software developers in harnessing the capabilities of the emerging technology.

During the SAP TechEd conference in Bangalore this week, SAP announced that it had integrated GenAI capabilities into a new low-code tool, SAP Build Code, to help developers and business users build applications and automate processes.

SAP Build Code will use SAP’s new GenAI co-pilot called Joule to enhance developer productivity by incorporating code-generation capabilities for data models, application logic and test script creation.

In addition to expediting code development through GenAI, SAP is enhancing its Hana Cloud multimodal database with a new vector database engine. This will allow organisations to combine the capabilities of large language models (LLMs) with enterprise data to answer queries.

Juergen Mueller, chief technology officer of SAP, explained: “Large language models bring sparks of intelligence, but they also have severe limitations. They have no idea what happened in the last one or two years and they have no access to any business data, so it’s hard to deploy them in production.”

The vector database engine in Hana Cloud will enable users to search for suppliers based on contract language, analyse payment history and track individual orders, among other tasks, for example.

This feature is set to be generally available in the first quarter of 2024 and is considered a significant development in SAP’s database offerings since the launch of Hana Cloud itself, said Mueller.

Having worked with LLMs for over a year, SAP recognises the complexities of putting GenAI workloads into production. To address this, it is introducing AI Foundation, a new hub in SAP’s Business Technology Platform (BTP) that will empower developers to create AI- and GenAI-powered extensions and applications.

“It gives access to large language models and Hana Cloud’s vector database engine to enable retrieval augmented generation,” said Mueller, referring to the ability to retrieve data from external sources of knowledge to improve the quality of responses. “We also provide support with prompting, so you can ask a model the right questions, on top of the capabilities related to responsible AI.”

Julia White, SAP’s chief marketing and solutions officer, emphasised the company’s commitment to data privacy and AI ethics, noting that it engages its internal and external advisory councils to ensure its privacy and ethical principles are upheld for different AI use cases.

Mueller added that SAP customers were concerned about sharing data with LLMs and wanted to prevent misuse or retraining without consent.

“SAP, independent of the model we use, guarantees that the data shared – for example, in prompts or with retrieval augmented generation capabilities in Hana Cloud – is not used in any form or shape by the large language models, vendors and partners we have,” he said.

As for GenAI use cases SAP is targeting, White said while many organisations were looking at those that improve employee efficiency and productivity, the supplier was prioritising areas that drive cost savings and business growth.

“For example, one of our earliest use cases was applying generative AI to transportation and logistics, making sure that trucks coming in and out of warehouses are moving inventory to the right place at the right time.

“With a very simple use case, we were quickly able to see that a customer could save $1m a year per warehouse, and most of these companies have 20 to 30 warehouses. Those are hard dollar savings, not just employee productivity improvements,” she added.

Regarding pricing, White said SAP would base its strategy on an “AI unit”, a measure of value for different GenAI use cases with varying cost structures.

“We’re looking at the AI unit across use cases where we can [show] the value we’ve created and what a customer will pay for, and from a vendor perspective, making sure it’s covering our costs,” she added.

White said SAP would announce prices as it put more GenAI use cases into the market. In many cases, such as the Joule AI assistant that is built into SuccessFactors, GenAI capabilities are included in customer subscriptions, she added.

According to a study by Enterprise Strategy Group and TechTarget, 75% of Asia-Pacific respondents plan to adopt generative AI within the next 12 months, with nearly a third already running GenAI workloads in production or testing the technology.