Generative AI and Traditional ML: A Recipe for Analytics Success

In the latest installment of the Augmented Ops podcast, we spoke with Dr. Lisa Graham, CEO of Seeq Corporation, which develops software tools for industrial process analytics. In the episode, entitled "Gen AI, Democratization, and the Future of Industrial Analytics," Dr. Graham draws on her extensive experience and shares her views on the future of data and analytics. Starting her career as a process engineer and end-user of Seeq and other BI tools before joining the company and rising to the position of CEO, she provides nuanced insights into the role of data in driving process improvement, the continued importance of traditional machine learning, and the value that generative AI promises to bring to the realm of analytics.

Her insights reveal the challenges faced by industrial companies, particularly the need to turn the massive amounts of raw data that they collect on a daily basis into actionable insights that allow engineers to continuously improve their processes.

Dr. Graham underlined the importance of data in driving frontline operations, emphasizing its key role in driving operational efficiency and process improvement. "It's important that we recognize data and analytics are a critical part of operations," she states, “I mean, flat out from operations to process, equipment, monitoring. That's what enables the key intelligence for driving what everyone's talking about.” As she explains, it is crucial to extract actionable insights from data in order to be able to effectively improve product quality, reduce energy consumption, and ensure profitability.

Despite the abundance of data in industries, Dr. Graham pointed out a common challenge: turning this wealth of raw data into useful, actionable insights. She identified a scenario she referred to as "DRIP" — a common pitfall of many companies which refers to being Data-Rich, but Information-Poor. This situation, in which businesses possess a wealth of data but lack the capability to effectively utilize it, is a significant barrier to operational advancement.

Dr. Graham shares a number of examples of how manufacturers have transformed their operations through taking advantage of their data. In one instance, she describes a company with thousands of assets, but only had visibility into a few hundred of them. This meant they were be able to optimize that relatively small handful of assets, but they don't have the insights they needed to optimize the rest. By taking advantage of a powerful tool like Seeq to aggregate their data sources and conduct analysis, they were able to move to exception-based monitoring across all of their assets, enabling them to have greater insight and make significant optimizations to their processes.

One of the most important tools for turning the vast amounts of data being collected into valuable insights has been machine learning (ML). Dr. Graham notes that "traditional machine learning [is] continuing to demonstrate amazing business value, including in time series and analytics."

When it comes to the much-hyped subject of generative AI, she posits that the advent of generative AI is not about replacing traditional ML but complementing it. Generative AI has the potential to enable significantly more intuitive, user-friendly analytics tools that can further democratize data access across various organizational roles. By taking natural language inputs and generating the necessary SQL queries and visualizations, for example, this emerging technology promises to open up the field of advanced analytics to people who aren’t trained engineers or data scientists. This represents a massive stride in democratizing what was once a highly specialized skill held by few.

Dr. Graham envisions a future in which advanced analytics become an integral part of every worker's toolkit, regardless of their technical background. As industries continue to generate an ever-growing mountain of data, the role of analytics in driving operational excellence becomes more pronounced, and the businesses that thrive will be those that are able to most effectively take advantage of it. She believes that “in the next few years, especially in the next 24 months, the role of analytics is going to continue to become even more mission critical. So as we think about machine learning or generative AI or everything else that might be coming, it comes back to the role of analytics as the data-driven decisions continue to drive gains in productivity and sustainability.”

While she is bullish on the technology, Dr. Graham also highlights that generative AI capabilities have a number of important limitations that must be considered before businesses implement it into their processes. “Organizations must acknowledge its limitations and associated risks, including data challenges, a lack of transparency, and data privacy concerns,” she explains. These are not insignificant concerns, especially in an era where data is not only abundant but can also be sensitive and subject to strict regulatory standards.

Dr. Graham further emphasizes that "gen AI results need to be validated," explaining that the outputs of generative AI are only as good as the underlying data and models that they are based on. And while she does believe that it will significantly lower the barrier to entry when it comes to analytics, she stresses that “despite popular discourse, gen AI requires human oversight to function effectively; it doesn't replace the need for domain experts, but instead, it complements their expertise.”

"Gen AI is not magic," she asserts, arguing that the real value of the technology comes from integrating it into a larger toolbox for solving operational problems. As these tools continue to make their way into more and more products, manufacturers will have to carefully consider how they integrate this technology into their processes.