SAP Goes All-In on Artificial Intelligence

SAP's user conference, “Sapphire,” began on June 3rd.^rd Through 5^Number Held in Orlando, 12,000 SAP customers and partners attended the event, with an additional 15,000 watching remotely.

SAP (NYSE: SAP) is the world's largest enterprise applications provider. The German software giant has 27,000 customers and 300 million users relying on its software.

The biggest message coming out of Sapphire is that SAP is all in on AI. SAP is embedding its AI copilot, Joule, across its enterprise portfolio. 27,000 customers are using SAP Business AI on a regular basis, either as part of SAP business flows or by creating custom AI solutions with SAP's Business Technology Platform. And based on announcements at Sapphire, that number is set to grow rapidly.

The RISE with SAP service includes AI-powered cloud ERP managed and optimized by SAP. It is a feature-rich private cloud service from SAP. SAP's GROW service also offers similar capabilities for SMEs. It is a public cloud solution.

The company has already developed 50 use cases that leverage AI to help customers derive additional benefits from the solution. When SAP talks about AI, it means generative AI based on large-scale language models or AI based on machine learning. By the end of the year, the company plans to develop about 100 use cases. SAP CEO Christian Klein said, “We don't add use cases unless they add value.” This isn't for bragging rights. Every new use case is vetted to ensure customers get real value.

Joule is a role-based user interface that understands the process workers are engaged in and places them in the workflow where users frequently need additional help. Joule makes it easier than ever for users to get additional information or suggested solutions to their problems.

It's a next-generation user interface similar to Alexa or Siri. For example, you might ask Alexa to “Play Smooth Jazz on my station!” Joule, on the other hand, is a business AI. If there's a heat wave in the Southwest, a logistics manager might say, “Joule, show me all the refrigerated trucks that shipped in Arizona today!” Managers no longer have to drill down through web page after web page of dense tabular data to get the answer.

At Joule, Klein claims that AI can be embedded into 80 percent of the most common tasks. Most of the new in-context GenAI solutions are pre-trained on SAP's 200,000 pages of training and technical documentation. The business AI also understands SAP standard data. These two features mean that while GenAI's responses are not 100% accurate (no GenAI solution is 100% accurate), they are much more reliable than the answers you get when using GenAI on the public web.

Additionally, if a user doesn’t follow a recommendation or realizes that the answer provided is inaccurate, that feedback can lead to further training of the AI ​​model.

SAP argues that its approach to GenAI goes beyond the narrow AI use cases offered by best-of-breed application providers: End-to-end processes often span multiple applications. SAP can use its business technology platform to manage end-to-end processes, and as a result create AI use cases in the grey spaces between applications.

The Customer Permissions Conundrum

Most of Joule's use cases are developed using SAP's internal training data, and they produce great results without using customer data for training.

In other use cases, customer data may improve training or training may not be possible without this data. For example, when using GenAI to create job descriptions, aggregated and anonymous customer data may improve results. However, some customers believe that giving permission for SAP to use their data may benefit competitors. For example, an airline that creates job descriptions for mechanics and uses its own data to improve the job descriptions may not see value in allowing a competitor to derive value from its work.

While thousands of customers have given SAP permission to use their data for training purposes, many, perhaps most, customers will never allow this.

AI Use Cases

SAP has detailed AI use cases for code generation to create ERP extensions, providing managers with insight into compensation-related discussions, and sales forecasting capabilities that can be used to predict the sales rep-product mix most likely to drive sales.

But the supply chain use case was perhaps the most interesting. Optimization and machine learning already present in supply chain applications are core to deriving value from these solutions. Still, Joule will further strengthen SAP's supply chain offering.

Every application has a garbage in, garbage out problem. With AI, you can improve your master data, such as finding SKUs that are likely to be duplicates. You can improve the accuracy of the parameters you use. For example, lead times are often set and then ignored; however, lead times can change over time. Machine learning can help you detect and correct this important parameter.

Supply chain planning often suffers from the black box problem: users cannot understand the recommended production or inventory plans that result from an optimization run. While experienced users can dig into the log data and understand why a particular plan was generated, most users cannot. If users don't trust the answers, the application will not be used and will become an expensive inventory item.

Now, users simply ask Joule, “Why are our service levels declining?” According to David Vallejo, global head of digital supply chain at SAP, “Within three seconds, Joule gives us an incredibly relevant answer: First, demand is increasing. Second, this work center has a significant constraint.”

Joule can also provide potential answers: Often, when a problem arises, there are relatively few ways to address it. You can tell Joule to run an optimization for this limited set of scenarios and suggest the best answer.

SAP is also developing a logistics use case. Many carriers cannot provide advance shipping notices and other shipping data electronically. They provide it on paper documents. These documents are read and the core information is entered into logistics applications. But manual data entry is error-prone. To automate this problem, optical character recognition has been used. Optical character recognition uses automated data extraction to convert images of text into a machine-readable format. However, shipping documents can vary widely in format, so OCR can have a hard time extracting the relevant data. SAP believes that combining GenAI with OCR can significantly improve the reliability of the output.

Bottom line: Every enterprise application provider talks a lot about AI, but no one is betting as big on GenAI as SAP. SAP is all in on AI.