SAP pushes generative AI front and center for developer productivity, custom apps

SAP is gearing up for the age of AI.

At its TechEd conference in Bengaluru, India, the German software major doubled down on its AI efforts and announced a bunch of tools and capabilities to not only boost the productivity of developers but also enable them to build custom AI-powered applications.

It unveiled AI-infused pro-code tools that accelerate application development in different ways and debuted much-needed vector database capabilities in the HANA Cloud, along with an all-encompassing AI foundation hub to help teams easily build LLM applications targeting different business use cases.

“When customers think about AI and SAP, we’re going to give three different ways to experience it,” said JG Chirapurath, chief solutions and marketing officer of the SAP, in an interview with VentureBeat. “First will be embedding AI inside the application experience. Second will be Joule, our assistive copilot to help with different tasks. And the third will be the ability for customers to create their own AI use cases – to meet needs bigger than what we can think of.”

He also shared that the company plans to launch its own foundation model, as the demand for generative AI-driven apps continues to grow from enterprises across sectors.

Business Technology Platform upgraded with AI

SAP’s Business Technology platform (BTP) has been critical to powering business applications in the cloud. It brings together application development, data and analytics, integration, automation and Al capabilities in one unified environment.

Last year at TechEd, SAP pushed BTP ahead with its Build solutions providing low-code, drag-and-drop features for web and mobile app development. Now, this is going to the next level with Build Code, an application development solution that embeds AI-based generation to produce code, data models, app logic and test scripts for Java and JavaScript applications.

“It provides one stack for Java and JavaScript development, which means that our SAP developers do not have to look outside the SAP stack to go find a piece of software developer toolkit to do that. The second thing it does is that you basically get to reuse code snippets from Build solutions as there are common architectural underpinnings. And the third thing we’re doing is offering interoperability with ABAP (advanced business application programming) cloud environment so you can do side-by-side development,” Chirapurath explained.

While the offering will become generally available in Q1 2024, SAP executives demonstrating the technology did confirm to VentureBeat that it can drive up developer productivity by 40-60%.

AI-driven productivity is just the beginning

Beyond Build Code, SAP is also moving to give developers the ability to build custom AI-powered applications, targeted at their own specific business scenarios and use cases — with vector database capabilities in HANA Cloud and AI Foundations.

With vector capabilities, contextual business data, like invoices in PDFs, images and docs, is stored in vector format in the HANA Cloud, allowing easy search and discovery of all relevant data (regardless of their original format) for training an AI model. This, Chirapurath said, saves the hassle and expense of setting up a second database solely for the purpose of cleaning and searching relevant data for training.

“Our promise with HANA Cloud has always been to add all the different features and modalities that you need for the price of one database. That’s one of the reasons why we call it multimodal,” Chirapurath added.

Enterprises using SAP BTP can tap the vector database capabilities of HANA Cloud through AI Foundation, the company’s new one-stop-shop to build and create AI and generative AI-powered applications. It provides everything one needs to start creating business-ready AI tools, right from access to all the open-source LLMs the company has partnered with to the ability to fine-tune, manage and monitor them.

This way, if an enterprise is not interested in using what’s already available from SAP, like an AI application for data analysis or for creating a job posting, they can start building their own set of AI use cases using contextual data and models of choice.

A proprietary foundation model in works

Currently, SAP is offering multiple general-purpose models as part of its AI hub, including those from OpenAI, Anthropic and Meta, as well as its own cross-platform multi LLM-powered solution, Joule, introduced last month and which is available throughout the SAP cloud and connected apps.

However, in order to make the outcomes even better and business-relevant, the company also plans to launch its own foundation model – which will be highly tailored to the unique context and data it holds.

“It will really understand the processes and the context, and it will give enterprises a fast start. Nothing can be 100%, but our aim is to get it to at least 80-85%,” Chirapurath added.

He didn’t give an exact timeline for the model’s launch but noted that the plan is to make it available “as soon as possible.”

That said, even after adding the option of its own model into the mix, the company has no plans to slow down its partner-first approach on AI. It will continue to rope in partner solutions and will consider an investment or acquisition only in cases where it’s important and strategic.

Compliance with AI executive order

As the company continues to build AI tooling, including the proprietary model, it will also work to abide by the new standards for AI safety and security announced in the executive order issued by President Biden.

“We have an ethics board to weigh it (AI developments) against ethical use and responsible use. Further, we’ve also got lawyers to weigh it against government-prescribed regulations to make sure we’re not releasing anything that alters the safety/security posture. More importantly, our customers also operate in in some of these areas. This is a duty of care. We cannot say I sit in Europe, you figure it out. We have got to figure it out so that the burden is much lower for them to be compliant,” Chirapurath said.