SAP confident of AI in ERP, techies ponder challenges

SAP has used the ASUG (Americas SAP User Group) convention in Orlando to attempt to further justify the use of Artificial Intelligence (AI) in modern Enterprise Resource Planning (ERP) deployments.

Already a widely voiced theme from SAP, the company has gone one step further in terms of describing the degree to which AI controls will penetrate the types of systems its software is used to build.

“We want to automate half of all ERP business processes within the next three years using AI-powered functionality,” said Bernd Leukert, member of the executive board of SAP SE for products & innovation. “SAP S/4HANA Cloud is the only product in the ERP market to give companies the breadth and depth of intelligence they need to leapfrog their competitors to stay on top of their game.”

This leads us to SAP calling SAP S/4HANA Cloud ‘the only tried and true intelligent ERP’, so can the company validate and substantiate that claim?

Part of the actual software at work here is the SAP CoPilot digital assistant, a technology designed to put conversational user experiences into ERP. It is, if you go with the marketing gloss, the world’s first ‘hands-free’ ERP.

But is a promise of AI in ERP and a selection of fancy (albeit no doubt intelligently engineered) chatbots enough to really make these enterprise systems smarter?

The jury is out

The ERP AI naysayers claims that we’re years away from having the data and machine learning ability to train the algorithms and automating ‘that many’ processes [that would typically be needed] within an ERP stack.

Imagine all the on-premises installs with their own process configurations, now imagine the time required to train an algorithm to understand them all, with limited access to datasets. That makes ERP AI really tough, doesn’t it?

The ERP AI proponents disagree and say that we can already automate around 80% of SAP ERP tasks without using screen scrapers.

These same proponents remind us of the opportunity here, that is – why should ‘human’ people do things like invoice matching when this kind of process can be carried out by machines and the people don’t enjoy the work in the first place? This is ERP work that is definable enough to be able to point the AI engines at to target the heart of the task.

It’s all the data, dummy

“There is merit to the comments of both the naysayers and proponents. This is because the proliferation of AI inside of ERP (and everywhere else) will come down to the data. AI will be realised when there is enough data being processed the right way to teach the machines and software to automate and leverage the capabilities of AI,” said Daniel Newman, principal analyst at Futurum Research.

“As we continue to collect and use more data and train the machines we will see AI’s potential realised…  and I believe it will happen much sooner for companies that make the investments in it. For those that wait, they will likely fall behind the competition,” added Newman.

In the latest release of SAP S/4HANA Cloud, project managers now can take advantage of AI-powered project cost forecasting capabilities to reduce budget overruns and make more accurate resource investment decisions.

In other use cases, a retailer can use dynamic pricing to intelligently adjust prices according to demand. A business manager can tap into staffing predictions based on the requirements and status of a given project or job.

It is certainly important to remember that as we build layers of machine learning (ML) into systems designed to power what will ultimately form the AI intelligence to help ERP systems work better, there is no intelligence without the data.

Deeper still, unless we the humans a) get access to the right data in the right process configurations and b) physically code the right functional scripts into codebases feeding ML and AI, then there’s no smart anything in the future.

SAP CEO Bill McDermott

Data Center
Data Management