Why firms hesitate on AI: Insights from SAP

Image created by DALL·E 3.

Every day, a new AI use case is born. Particularly with generative AI, the rapid pace of innovation is often leaving enterprises wondering what use case to adopt or where to even begin.

The challenge now, for tech providers like SAP, is to ensure that the capabilities they enable meet the immediate needs of their customers. After all, innovation is useless if no one is using it.

During a media conference, Philipp Herzig, Chief AI Officer of SAP, addressed key issues surrounding AI adoption, including democratisation, transparency, and interoperability.

Generative AI on a roll

At present, SAP has already developed around 30 use cases for generative AI, and plans to introduce 100 more throughout the year.

One example is its AI assistant Joule, which launched in late 2023. Embedded throughout SAP’s cloud enterprise portfolio, Joule sorts through and contextualises data across multiple systems to provide quick answers to business queries.

Philipp Herzig, Chief AI Officer, SAP. Image courtesy of SAP.

“SAP business AI means that we are embedding AI into our business applications and processes — supply chain, finance, HR, procurement, travelling expense, and all the various other business processes you usually find,” Herzig noted.

When asked about which use cases SAP prioritises for its R&D, the executive said it all boils down to value.

“We prioritise AI scenarios by value, because at the end of the day, you can do a lot with AI, but if you know the value isn’t there, or it’s not seen, or it’s maybe too expensive, it will not get adopted,” he said.

In addition, quality is also important when developing a particular use case for AI.

“We look at this through the lens of the customer’s return on investment, whether that was a part of the existing licensing or whether it’s part of our premium AI offerings. Ultimately, the value for the customer and what they can gain out of that is front and centre,” Herzig remarked.

On the customer side, many are still undecided about which of the available use cases to adopt, pondering whether to prioritise AI in HR, finance, or the supply chain.

Considering 2023 as primarily an exploration phase for generative AI, Herzig observed that customers often come in already with an idea of where they want to deploy AI.

“Customers, for example, have pain points in HR, because maybe they need to rescale. Or perhaps their focus is on the supply chain, feeling the pressure in this area and seeking AI solutions for relief—be it through demand forecasting, what-if analysis for planning, or streamlining procurement processes,” he said.

To solve the dilemma, Herzig reverts to the value perspective, noting which SAP applications a particular customer already has, and from there maximising opportunities.


While AI-enabled automation has significantly eased operational bottlenecks, eliminated redundancies, and saved enterprises countless hours on mundane tasks, AI has the potential to accomplish much more, Herzig stated.

Use cases, according to the executive, don’t have to operate in isolation.

“Consider this scenario: You’re working on an HR-specific task and wish to check the budget—a finance-related query. Typically, we’d access the finance system in S/4HANA for this. Now, with S/4HANA connected in the background, you can simply inquire, ‘Do I still have budget? How much is left?’ and receive an answer based on the cost centre and your user profile, because it’s all integrated with each other. For a deeper dive, a click can transition you into S/4HANA, directly to where budgets are usually reviewed. To continue the conversation, if you open Joule within S/4HANA to address a completely different application, it will recall everything you’ve discussed with Joule and SuccessFactors,” Herzig explained.

Meanwhile, there is growing apprehension among workers that AI might replace their jobs—not just writers, graphic designers, and filmmakers.

Now, even software developers are concerned, especially with the advent of tools like GitHub Copilot and Devin, which automate code generation, bug fixes, and more.

“People often freak out, fearing these tools will replace software developers. However, those who believe such tools will soon replace developers likely don’t fully understand software development. Fixing bugs or improving documentation and tests are just one part of it. Writing good software is inherently creative work, requiring constant innovation and often direct customer interaction to determine the next steps. It’s beneficial that these tools can handle bug fixes, freeing up time for developers to design the next feature or delve into what the customer truly desires,” he added.

Data transparency

Across industries, tech firms are applauding laws like the EU’s AI Act for its risk-based approach, instead of imposing strict regulations on algorithms. Similarly, an ASEAN guide has been introduced to help businesses in the region with their AI governance strategies.

For SAP, data privacy and security are paramount when designing new AI tools.

“When we consume GPT models from Azure, Gemini models from GCP, or those from Bedrock, we conduct thorough due diligence on processes such as data handling and compliance with our standards. Our contracts remain unchanged as we maintain the same level of data security and privacy SAP has adhered to over the past 50 years,” Herzig said.

Moreover, the executive emphasised the importance of upholding these standards with all partners and stakeholders to ensure a trusted ecosystem.

“It’s crucial that we don’t use customer data for training purposes. While we might use it for inferencing or to run queries, we never use it for fine-tuning. We may refine our models or adjust our programming language, primarily using open-source models in-house, but we strictly avoid using customer data for fine-tuning those well-known models. That is something we do not do,” he concluded.