AI tackles customizations in SAP clean core migrations

No easy task to get to the clean core

Many SAP customers are aware that adhering to clean core principles will enable modernization, said Tim Wintrip, CEO at Lemongrass Consulting, a firm with U.S. headquarters in Edison, N.J., that provides cloud modernization for SAP and other systems. However, this can be difficult to achieve if customers haven't assessed their core systems.

The problem with the clean core is that customers have these very complicated estates -- especially the larger clients -- that have tens of thousands of custom objects.

"The problem with the clean core is that customers have these very complicated estates -- especially the larger clients -- that have tens of thousands of custom objects," Wintrip said. "For example, some of the [SAP ABAP] Z-code has been there for 20 years. Sometimes it's not documented anymore, but you can't just ignore it, because it may be doing something."

To combat this, Lemongrass released Clean Core AI Accelerator, a tool designed to help SAP customers simplify the transition to a clean core by providing deep analysis of their customizations. It was designed to ingest system information and enable users to find and document all of their customizations, according to Wintrip.

"This ingests all of SAP's release notes and every document that SAP has ever released on S/4HANA, so you can look at that Z code and then create documentation that anyone can read," Wintrip said.

Further, Clean Core AI Accelerator can recommend removing a Z-code customization because it's standard in S/4HANA or whether it can be replaced with an app in SAP BTP.

Kyndryl, a global provider of IT infrastructure services such as SAP cloud migrations, is partnering with tiny AI firm Nova Intelligence to provide agentic AI tools to enable implementation of a clean core.

Like Lemongrass, Kyndryl is looking to help its customers untangle the Gordian knot of customizations that are daunting clean core initiatives, according to Michael Bradshaw, global practice leader for applications data and AI at Kyndryl.

"Everybody is scared to deal with the heritage of what they built, because they know how long it took, how much money it took, and nobody wants to spend that same amount of money just to go to the next version of the software," Bradshaw said.

Earlier this year, Kyndryl applied Nova agents to its own SAP system applications and moved more than 33,000 lines of custom Advanced Business Application Programming (ABAP) code into clean core functions in 10 days, according to Bradshaw. A traditional process could have taken more than six weeks.

Kyndryl then partnered with Nova to include the agents in its Data Transformation Suite, a set of tools for SAP migration projects that it released in the spring.

The Nova agents work on SAP ERP Central Component and S/4HANA system data to find and catalog custom Z-code. It then creates a functional spec using the SAP data dictionary and repositories such as SAP knowledge bases to make recommendations on whether customizations should be moved into a clean core S/4HANA cloud environment.

"Some customers that have a million-plus lines of ABAP code that would take teams of programmers many months to go through and analyze and then work with business analysts to put into context and determine the business outcome that's being driven by this customization," Bradshaw said. "This attacks the root cause of what's been the reason that customers are slow to move."

Enterprise architect sidekick

The tools are poised to help SAP customers deal with the clean core barriers, according to analysts.

David Groombridge, an analyst at Gartner, described Lemongrass Clean Core AI Accelerator in particular as a "great development" because customers have been looking for AI-based tools that can take on this work.

The traditional approach for many customers has been to throw people at the problem by bringing in large systems integrators (Sis), which can be costly, he said.

"If you employ a global SI, they will often bring in a lot of consultants on a day rate to identify all those customizations and then migrate them manually over to a new format," Groombridge said. "For many large enterprises, that number is going to be six or seven figures in dollars."

Clean Core AI Accelerator will work best as a sidekick for enterprise architects who are doing the migration planning and implementation, he said.

"It's not actually executing the migration, it's coming up with a complete listing of the customizations and for each one of them a recommendation," Groombridge said. "A sensible migration team is not just going to take that at face value because AI is good, but it's not without its flaws, and there may be valid reasons why you don't want to accept those recommendations."

There is a potential market for these tools, according to Holger Mueller, vice president and principal analyst at Constellation Research.

"Getting to a clean core can be a massive, expensive and potentially impossible endeavor for an enterprise," Mueller said. "The good news is that AI has arrived and can help parse through all the code repositories, analyze usage and even rewrite some code."

However, this category of tools is in its infancy, making it difficult to project the full value yet, he said.

"Like at every start, the vendors pick the low-hanging fruit," Mueller said. "Which is fair but makes it impossible to assess how much of any of these tools are helping enterprises to move to the clean core. The alternative is a from-scratch vanilla implementation."

Jim O'Donnell is a news director for Informa TechTarget who covers ERP and other enterprise applications.