Peter Sayer
Senior Editor

SAP to add generative AI, industry smarts to CX tools

May 16, 20235 mins
Application IntegrationArtificial IntelligenceEnterprise Applications

A digital assistant incorporating generative AI functions is one improvement to SAP’s customer experience software the company will discuss at its Sapphire conference this week.

SAP sign
Credit: Peter Sayer/IDG

Every software developer is looking at how to incorporate generative AI in its products, even SAP.

The ERP vendor, which turned 50 last year, is developing a companion app for its software, to be called SAP Digital Assistant, which will use generative AI to help SAP users provide a better experience to their customers.

SAP is also looking to improve its customer experience tools in other ways, providing industry-specific accelerators to help some enterprises roll the software out faster, and helping online retailers reduce waste by offering tools to support the resale of refurbished or returned products.

Although SAP Digital Assistant will be accessible from SAP’s Sales, Service, Marketing and Commerce Cloud products, one of the areas its generative AI capabilities could be most useful is in customer experience management, helping sales staff and contact center agents respond to customer requests or identify their needs and draft pitches.

But it could also draw on the detailed information SAP’s enterprise applications hold about orders and inventory, for instance, to identify which products to promote or help field service engineers optimize repairs, said Ritu Bhargava, SAP’s chief product officer for industries and customer experience.

The assistant has little in common with the category-defining generative AI application, ChatGPT, according to Bhargava. “It’s not a reactive chatbot,” she said. “It’s a contextual prompt in the app.” SAP isn’t using GPT in its assistant, nor IBM’s Watson, which it recently said it will use across all other areas of its business. “This is our own in-house custom-built engine,” she added.

SAP Digital Assistant could also operate as a browser extension or an additional task pane in Microsoft’s or Google’s productivity software, she said, enabling it to evaluate previous interactions with a customer and advise on composing an email.

Early stages

The app is still in early development — it won’t enter beta testing until the fourth quarter — but Bhargava said pilot customers are already participating in its creation, and she hopes SAP can make a decision later this year about whether to release it.

Some of SAP’s customers are considering expanding their use of AI in the customer experience arena.

For jewelry maker Swarovski, it’s important that AI fits into the luxury experience its customers expect, CIO Lea Sonderegger said at a recent round-table event. The company already uses it to make product recommendations on its e-commerce site and is looking at using it for generating marketing messages. But such processes will never be entirely automated. “We use AI for key decisions, but the human should have the final word,” she said.

Another speaker at the same event also saw applications for generative AI in recommendations and marketing. Yulia Groza, VP of e-commerce technology at Levi Strauss, is evaluating the technology, in particular to assess the risk associated with it.

Concerns remain that generative AI tools, based on large language models like GPT3, could create privacy problems for enterprises and introduce or perpetuate biases because of the volumes of data about past interactions needed to train or refine their large language models.

Bhargava said that SAP is aiming for transparency in the way its assistant app makes recommendations, showing users the documents it relied on to make its recommendation, and enabling them to exclude some or give more weight to others.

But beyond that fine tuning of the model’s behavior, enterprises and their data will not be involved in the training of the AI model itself, she said.

SAP will train the models using a mix of open-source data and customer data in a multi-tenant environment, but each customer’s data will be processed independently and with no cross-polination of customer-specific insights from one tenant to another, a company representative said.

Improving returns

Later this year, SAP will offer another new tool to help retailers sell unwanted product. These may be items that aren’t selling well, nearing their expiry date, or returns. It will also identify underperforming products that can be promoted, and help track those being returned, both to facilitate reselling them, and also to prevent overproduction of unpopular goods.

“It’s not just about what you do with the returned goods,” Bhargava said. “You also, ideally, want not to have returned goods.”

SAP Recommerce will be made generally available in the fourth quarter, she said. The company hasn’t yet decided whether it will charge for the feature, or integrate the functionality into its existing SAP Commerce Cloud offering.

Accelerated roll-out

SAP has selected four industries to benefit from faster roll-out and integration of its customer experience tools: consumer products, retail, automotive, and utilities. For those industries, SAP will provide out-of-the-box integrations with common tools, and industry-specific data models to predict customer behavior. In retail, for instance, these models might help enterprises calculate a lifetime value score to help them focus on their best customers.

Retail is a broad category, so SAP will initially focus on two sub-categories, groceries and high fashion, where the company has a large installed base, and where there’s strong demand for the product, Bhargava said.