SAP's Joule is the company's generative AI assistant for its cloud portfolio, but the technology is still in the early stages and the ERP giant will need to prove its value.
Generative AI is coming to SAP cloud applications as SAP took the wraps off its AI assistant, Joule, but analysts said customers should hold off on their enthusiasm as the offering is more idea than reality right now.
Joule will be embedded across SAP's cloud product portfolio, including SuccessFactors later this year and the public edition of S/4HANA Cloud in 2024. SAP Customer Experience, Ariba and SAP Business Technology Platform (BTP) will follow, according to the company.
The AI assistant is designed to provide customers with contextualized information and assistance when performing tasks in SAP applications, said Julia White, SAP chief marketing and solutions officer, at a press conference Tuesday.
"AI is moving past the hype cycle and various pilots into more meaningful adoption of generative AI to deliver real value," White said. "Looking ahead, AI will power every aspect of our technology, and this will bring tremendous business value to SAP customers."
As a generative AI technology, Joule will act as an interface for users to ask a question in plain language and get answers that will draw on third-party large language models (LLMs) built on business data from SAP applications, according to White. In July, SAP unveiled that it was investing in AI startups Aleph Alpha, Anthropic and Cohere. This was in addition to third-party AI partnerships formed with Microsoft, Google Cloud and IBM in May.
Pricing information was unclear. SAP had previously stated that embedded AI capabilities would command a 30% premium.
It's expected that some Joule functionality will be available with no extra premium cost to customers, but other functionality will carry a premium increase based on the particular business use case, White said. More details about these use cases will come out as they're developed.
"The generative AI use cases of the applications that Joule calls upon will have unique business models associated with them based on the value provided," she said.
Other characteristics of Joule include a focus on data privacy and guardrails that are intended to prevent biases from infecting the LLMs that Joule uses, according to SAP executives.
Joule will not use customer data to train LLMs, said Bharat Sandhu, senior vice president for SAP's AI and application development platform.
SAP has the "highest ethical standards" and since 2018 has used an external AI council that goes through "every possible use case," said Thomas Saueressig, executive board member for SAP product engineering, at the press conference.
"There were a couple cases where we said we cannot do them for ethical reasons," Saueressig said.
Too early in the game
SAP customers should keep some of their enthusiasm for Joule in check, according to analysts.
The product announcement is early, and customers don't have access to the technology yet, said Jon Reed, co-founder of Diginomica, an enterprise computing industry analysis firm. Also, SAP is plowing ground that has already been seeded by enterprise rivals such as Microsoft and Salesforce. Still, he said, the technology has promise.
One thing I like about Joule is that it will be embedded and work alongside many SAP applications.
Jon ReedCo-founder, Diginomica
"One thing I like about Joule is that it will be embedded and work alongside many SAP applications," Reed said.
SAP's large on-premises customer base might be in danger of missing out on some of the Joule functionality, given the vendor's focus on its cloud products. But they will have some access via BTP, SAP's development and integration platform, and it's not realistic for any vendor with a large on-premises footprint to integrate AI assistants across all its products, he said.
"It's understandable that SAP would start with [embedding Joule into] the cloud offerings," Reed said. "This is different than force-fitting people onto Rise with SAP in order to get it, and there are real technical limitations here with on-premises code lines."
It's still very early days for Joule, and the news is a play for the hearts and minds of Wall Street investors who want enterprise vendors to push AI technology, said Joshua Greenbaum, principal at Enterprise Applications Consulting.
SAP is also concerned that its existing customer base and net-new customers might be tempted to find similar AI technology from cloud hyperscalers, Greenbaum said.
"The hyperscalers have their own huge incentive to be in this market," he said. "Because, at a minimum, even if ChatGPT never delivers true business value, all the experimentation and model creating will use a ton of cloud cycles, and they'll make book whether this bet works or not."
This generative AI arms race has led to SAP playing defense and offense simultaneously, Greenbaum said. With Joule, the vendor has put the cart far out in front of the horse, he added.
"This technology is not available now, and the LLMs that it will be built on are still in development," Greenbaum said. "The fact that SAP has 20,000 customers who have agreed to give some version of their enterprise data to build the models is muddied by the fact that we don't know who these companies are, what their business processes are or what corners of the global economy they represent."
Right now, Joule is a promise, but there's not a lot more to it yet, Greenbaum said. It also reminds him of a previous situation when SAP bundled several advanced technologies into the SAP Leonardo platform.
"[Leonardo] ended with the idea that a lot of that functionality was going to be embedded and part of the commodity-level functionality that neither caused SAP to charge a premium nor caused the customer to get premium value," he said. "My concern is that we'll end up in that situation again, so it's truly up to SAP to start proving value now that they've put their money where their mouth is."
Jim O'Donnell is a senior news writer who covers ERP and other enterprise applications for TechTarget Editorial.
