Jazz Versus ROI: Positive Signs

From what I've seen in this space, manufacturers have been investing heavily in AI. I've seen large enterprises engaging in pilot programs in areas like predictive maintenance, quality inspection and supply chain optimization. That said, I would say that these are more like experiments than real pilots.

I spend a lot of time with our customers and industry leaders, and unfortunately, actual ROI is unclear. While some organizations claim gains, especially productivity-related ones, many initiatives stall at the proof-of-concept stage. The result is an increasing gap: AI is widely explored but not consistently scaled—and definitely not cohesive across the organization.

This discrepancy often comes down to execution. Companies that treat AI as a strategic capability tied to real operational problems and proceed with outcome-based pilots tend to see ROI. Those that approach it as an experimental or isolated tech initiative often struggle to generate value.

What Is Slowing Us Down, And What Can We Do?

There are several barriers at play here.

1. Unrealistic Use Cases: Many organizations start with overly ambitious or vague goals. These visions, while compelling, are difficult to implement and rarely deliver short-term value. You have to spend millions to even understand long-term returns. Instead, start with basic use cases, especially around diagnostics and productivity-based solutions. Trying to start with cognitive and predictive use cases as the first step may jeopardize your ability to measure returns quickly.

2. Poor Data Visibility: AI depends on having access to data. In many factories, data is siloed, inconsistent or incomplete, making it difficult to train reliable models. The old philosophy of "bring your data to a warehouse" is not true anymore. Disparate systems perform functions that were designed to be best-of-breed. Instead of centralizing data, evaluate AI frameworks that can leverage disparate data sources, provide specific intelligence where needed and enable the visibility of both problems and issues across diverse datasets. This allows better, inclusive decisions (while you work to remove silos). Don't wait for a master data plan to emerge before you see value from AI.

3. Skills And Cultural Gaps: A shortage of AI expertise, combined with resistance to change on the shop floor, can slow adoption. There needs to be a willful investment in skills engagement. Perhaps more important than using tools is investment in understanding the ethical, legal, privacy and security impacts of leveraging this disruptive technology.

The Problem With Unrealistic AI Examples

A major contributor to stalled adoption is the influence of unrealistic AI narratives. I have seen promise-driven marketing from vendors that are not grounded in reality, including promises of fully autonomous production lines or instant optimization of audit and policy documents.

The clear downside of believing the hype is that we will overinvest in complex solutions before they prove their value and/or skip foundational steps. As a result, we will be tempted to abandon AI initiatives when early results fall short.

In the limited time that AI has been in the market, the most successful AI implementations that I've seen are incremental, targeted and grounded in specific business problems. Rather than aiming for transformation overnight, we must emphasize incremental progress with measurable outcomes.​

AI adoption in manufacturing is accelerating, but true value comes from practical implementation. By focusing on realistic use cases, building strong data foundations and scaling proven solutions, manufacturers can bridge the gap between experimentation and ROI. Practical AI is not just a starting point—it is the pathway to sustained, transformative impact.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?