Is Your AI Strategy Realistic, or Just a Pipe Dream?

Luke Durcan, director of EcoStruxure at Schneider Electric, recounts a conversation with an executive from an industrial firm who stated: “We want to do some AI. We want to get some AI into our process as quickly as possible.”

When Durcan asked, “When?” the executive replied, “Probably July.”

Durcan observed that the executive clearly lacked a deep understanding of the processes, mechanisms, and requirements necessary to achieve AI integration. He noted that, in many industrial contexts, “AI” is often used as a buzzword rather than a concrete capability.

Experts in industrial analytics and data science consistently agree on a disciplined, step‑by‑step approach: contextualize data, then apply analytics, machine learning, and deep learning techniques to achieve real business outcomes.

Atif Kureishy, head of Teradata’s AI and deep‑learning initiative, defines AI as a suite of supporting techniques—analytics, machine learning, deep learning—tailored to solve specific business problems. “For example, deep learning—using neural networks, GPU‑based computation, and high‑dimensional data—enables increasingly accurate predictions,” Kureishy explains.

Industries that have embraced these techniques early include consumer tech, financial services, and insurance. Retail, telecommunications, and automotive are following closely. In manufacturing, the automotive sector is among the fastest adopters, driven by its focus on autonomous vehicles.

Durcan highlights the oil and gas industry as a pioneer in process sectors. “These organizations have invested in data, infrastructure, and technology for decades because it has consistently delivered value,” he says.

Further along the maturity curve, consumer packaged goods, materials, minerals, and mining sectors show steady progress, while discrete manufacturers—such as electronics producers—are already quite advanced.

So, what should lagging industrial firms do to catch up in the era of Industry 4.0, smart factories, and AI? And what next steps should mid‑tier companies take?

The first move is a self‑audit to ensure a robust data‑science foundation. Teradata often assists clients in building the foundational elements that have served its banking customers for 30 years. Many industrial companies must identify the contextual data they possess, calibrate sensors, and focus on “data‑science 101”—talent, tooling, and environment.

According to Gartner’s 2023 research, more than 87% of organizations across all sectors exhibit low business‑intelligence and analytics maturity.

At the outset, a manufacturer may deploy a network of sensors across its operations to capture the conditions a material experiences during production. With contextual data in hand, the company can detect anomalies that precede defects and scrap. “That’s not full prediction yet, but it’s a better characterization of the process,” Kureishy notes.

Durcan stresses that technology must be integrated with people and processes. “In a typical brown‑field facility, seasoned employees hold invaluable process knowledge. Your AI tools must complement that expertise, delivering incremental value,” he says.

Once a firm has a solid data‑science base, it can explore more advanced techniques like neural networks. As maturity grows, the focus shifts from describing events to correlating variables and establishing causation. “When A happens, B occurs, so we know C will materialize,” Kureishy explains. “This predictive insight allows proactive intervention.”

Further sophistication enables prescriptive recommendations to optimize or fix processes. The apex of this journey is fully automated anomaly detection and resolution—essentially, an automated “terminator” of defects, as Kureishy joked.

Durcan cautions that the AI journey has no final destination. “There’s no nirvana on the horizon; the landscape keeps evolving,” he says.

Finally, industry leaders must prioritize data integration. Data alone is just the starting point for predictive or analytic models. Understanding the asset hierarchy, model, and context is critical. From there, organizations can build detailed data flows and infrastructure, enabling descriptive visualizations and operational responses. “That’s what 90% of people use data for,” Durcan adds. “The first step is the hardest.”