Not to be a spoilsport amidst the generative AI fanfare, but now is also the time to address the unique risks that this amazing technology brings to organizations as people embed it in ERP applications.
Teams in procurement, supply chain, HR, and finance have long relied on traditional machine learning automation to reduce tedious tasks and speed up daily productivity. A well-trained gen AI algorithm ratchets up the business results exponentially, producing new content – text, images, music, code – by enhancing and summarizing an avalanche of unstructured data in the context of someone’s job.
During a recent SAP Future of ERP podcast, hosted by Richard Howells, vice president of thought leadership for SAP ERP, finance and supply chain, ERP leaders talked about how organizations can balance the risks and opportunities at stake for companies as gen AI emerges.
Data quality: Responsible AI builds trust
Trained correctly, gen AI can accurately predict the next words in a conversation based on the probabilities that were established during the algorithm’s development and fine-tuned by human feedback. Like traditional computing models, gen AI is only as good as the data that built it.
“Low quality data training will result in a gen AI model with low quality output,” said Wim Rymen, partner at PwC Europe.
Rymen said that privacy and compliance risks only multiply with proprietary gen AI models. He advised companies to use responsible AI practices, building trust into models by design.
“People are a key element in responsible AI,” he said. “Organizations need to work with their talent and HR teams to develop training programs that make sure people at all levels of the organization are not only aware of the opportunity that gen AI brings, but also the risks and how to manage those. Companies that balance the risks with the rewards of this innovative technology will be able to instill trust in their organization, which is good for business and the company’s brand.”
Avoid unintended bias with embedded DEI
Bias surfaced in early AI-based models that experimented with some recruitment tools. Rymen advised organizations to work with diversity, equity, and inclusion (DEI) specialists when developing their own gen AI models.
“It’s important to incorporate ethical considerations in your model to avoid getting outputs that could be discriminatory or unfair,” he said.
Train employees for quality AI input
Billions of people who have signed into open AI tools like ChatGPT are already discovering the perils of bad prompts. Gen AI does not replace human thinking, insights, and creativity. To be effective, prompts need to be written coherently with the words, phrases, and sentence structures that allow the model to respond with quality outputs.
“It’s about asking your questions in the right way,” said Rymen. “Unsophisticated questions or prompts without the appropriate situational context may lead to misleading, inaccurate answers.”
Responsible AI requires education
Hallucinations from gen AI are rampant and scary, ranging from minor inconsistencies to outright fabrications. While it’s unclear how large language models (LLMs) arrive at erroneous conclusions, people have an enormous responsibility to rein in mistakes.
“Technology is only as smart as the user,” said Rymen. “People still need to be critical with the information that gen AI produces…to avoid starting to report hallucinations as facts that may result in highly visible errors for instance, in financial statement reporting. Train your users to be critical of the output in each use case.”
Gen AI is future of ERP
Given the staggeringly fast pace of gen AI advancements, organizations need to move quickly in creating enterprise-wide generative AI risk management strategies and governance policies.
“There isn’t an industry or sector that’s not thinking about how to quickly drive innovation and use cases from gen AI,” said Elizabeth McNicol, principal at PwC US. “The future of ERP is its ability to harness this powerful tool for improved decision-making, and to customize and simplify the user experience. It is probably one of the most significant changes that we’ll see as it relates to ERP, and it will create many new opportunities for automation. Managing and understanding the risks of Gen AI will enable our clients to really unlock the value that it can bring. We’ll see a lot of business reinvention.”
Learn how SAP S/4HANA Cloud provides ERP for every business need – from mission-critical operations to business model innovation.
Read the full article here
