How to Build a Winning AI Business Case in Manufacturing

The manufacturing world is crowded with failed digital transformation projects. Grand AI initiatives promised revolutionary change but delivered little more than expensive dashboards and frustrated operations teams. So what separates the wins from the washouts?

Pushpal Jagdale has lived through both sides of this equation. As IT Business Relationship Manager of Manufacturing & Operations at Novelis, he’s navigated the treacherous waters between IT vision and operational reality. His track record spans discrete manufacturing at Honeywell and metals production at Novelis (environments where promises must translate to measurable results on the factory floor).

In a recent conversation on The Manufacturing Intelligence Podcast, Pushpal revealed his blueprint for AI business cases that actually survive the boardroom and thrive in production. The foundation isn’t technology: it’s trust.

Before Pushpal ever mentions artificial intelligence, he spends time where the real work happens. His early career writing standard operating procedures, running Kaizen events, and analyzing cycle times taught him a fundamental truth: transformation initiatives succeed or fail based on their impact on frontline workers.

“When you come with a big transformation initiative, you have to ask yourself: how are you improving the life of someone in the plant?” Pushpal explains. “That empathy matters. Because if you don’t earn trust at the ground level, you won’t scale anything.”

This isn’t just feel-good philosophy; it’s a pragmatic strategy. The operators, technicians, and supervisors who run production lines every day possess institutional knowledge that no algorithm can replicate. They understand the nuances of equipment behavior, the workarounds that keep lines running, and the real bottlenecks that constrain throughput. Ignore their insights, and even the most sophisticated AI system will optimize for the wrong variables.

Rather than positioning AI as a revolutionary leap forward, Pushpal frames it through the lens of continuous improvement: a language every manufacturing professional understands. His approach centers on three core questions that have guided lean manufacturing for decades:

Can you make the same output with fewer inputs? This might mean using computer vision to reduce quality inspection labor, or predictive analytics to minimize unplanned maintenance resources.

Can you increase capacity with the same resources? Perhaps through AI-optimized scheduling that reduces changeover times, or machine learning models that identify optimal process parameters for higher yield.

Can you reduce risk without compromising throughput? This could involve AI-powered safety monitoring systems or predictive models that prevent equipment failures before they impact production.

By connecting AI capabilities to familiar manufacturing KPIs (OEE, scrap rates, uptime, safety incidents), Pushpal translates abstract technology into concrete business value. “If it doesn’t serve the goals of the business,” he says, “it’s just noise.”

The temptation in digital transformation is to think big: to envision connected factories humming with intelligent systems. Pushpal learned early that sustainable change happens incrementally. During his tenure at Honeywell, he didn’t launch a global digitalization program. He started with one surface-mount technology (SMT) line.

“We cataloged every asset, noted what could talk to the network, and ran a small proof of concept,” he recalls. “That let us show value fast, gain trust, and get buy-in to expand globally.”

This pilot-first methodology serves multiple purposes. It de-risks investment by proving concepts before major capital allocation. It creates local champions who become advocates for broader adoption. Most importantly, it generates real data about what works in your specific environment, rather than relying on vendor promises or industry case studies.

At Novelis, this approach has enabled measured expansion from initial pilots to AI applications spanning predictive maintenance, automated SOP guidance, water chemistry optimization, and even robot dogs conducting hazardous area inspections.

When Pushpal walks into a budget meeting to request $3-4 million for an AI initiative, he doesn’t lead with technical specifications or competitive benchmarking. He channels the mindset of the VP of Operations, who controls the purse strings.

“If someone came to me asking for that kind of investment, I’d want to know: How does this help me meet demand? Improve traceability? Reduce rework?” he explains.

This perspective shift transforms AI business cases from technology projects into operational improvement initiatives. This framework distills every proposal to three fundamental questions:

What operational problem does this solve? Be specific. “Improving efficiency” isn’t a problem statement; “reducing unplanned downtime on Line 3 that costs us $50,000 per incident” is.

What is the measurable impact? Quantify benefits in dollars when possible, but don’t ignore soft savings like improved safety or regulatory compliance. Just be honest about what you can and cannot measure.

How does it make work easier, safer, or faster for the team? If your AI solution makes someone’s job harder or creates additional administrative burden, it’s not going to survive contact with operational reality.

Perhaps Pushpal’s most critical insight involves the unglamorous foundation that enables AI success. “There is no data without strategy and no strategy without data,” he emphasizes. “Most AI projects fail not because the models are wrong, but because the input data isn’t usable.”

This means doing the hard work before implementing AI: cataloging assets, standardizing protocols, ensuring IT/OT alignment, and establishing data governance processes. It’s less exciting than training neural networks, but it’s what separates successful deployments from expensive experiments.

At Novelis, this foundation-first approach has proven essential. Without clean, contextualized data flowing from production systems, even the most sophisticated algorithms produce garbage outputs. But with proper data infrastructure in place, AI applications can deliver immediate value and scale across multiple use cases.

To overcome natural resistance to AI adoption, Pushpal employs what he calls the “co-pilot strategy.” Rather than implementing AI systems that replace human decision-making, he starts with tools that augment human capabilities in obvious ways.

“Once they see how much time they save, they get curious,” he notes, referring to simple applications like meeting transcription and summarization. “That opens the door to broader factory applications.”

This approach builds familiarity and trust gradually. Workers experience AI as a helpful assistant rather than a threatening replacement. They see tangible benefits in their daily work, which creates receptivity to more sophisticated applications. From there, organizations can expand into predictive maintenance, process optimization, and autonomous monitoring systems.

Looking toward 2030, Pushpal maintains balanced expectations about AI’s manufacturing trajectory. He envisions a hybrid landscape where foundational upgrades continue alongside accelerating AI maturity in ready organizations.

His predictions include executive teams relying on AI for strategic planning, frontline workers collaborating with intelligent systems, expanded investment in knowledge capture, and the proliferation of AI co-pilots across functions. But he’s also realistic: “We won’t be lights-out everywhere by 2030. But we’ll be smarter, faster, and safer where it counts.”

This pragmatic outlook reflects hard-won experience. Manufacturing transformation happens in quarters and years, not weeks and months. Success comes from consistent progress rather than revolutionary leaps.

Ultimately, Pushpal’s methodology succeeds because it prioritizes trust over technology. By starting small, solving real problems, and scaling what works, organizations can build confidence in AI applications while minimizing risk and disruption.

The best AI business cases in manufacturing don’t showcase cutting-edge algorithms or impressive ROI projections. They demonstrate understanding of operational challenges, respect for frontline expertise, and commitment to sustainable improvement. They prove that AI can make work better, not just different.

In an industry where promises often exceed delivery, this trust-first approach offers a path forward that honors both technological possibility and operational reality. The factories of tomorrow will be built by leaders who understand that transformation starts with people, not processors.

Listen to the full podcast episode here.