Case Study: How One Mid-Market Manufacturer Turned Analog Power Signals Into AI-Driven Operational Gains — Guidewheel — March 18, 2026

By Staff Reports - March 18th, 2026

Mid-market manufacturers are entering the AI era with less patience for experimentation and more exposure to consequences. The sector is being pushed to invest by cost pressure and competitive urgency, but it is also being forced to confront an uncomfortable reality. A large share of AI initiatives still stall in pilots, fail to operationalize, or produce unclear returns, even as budgets keep rising.

Gartner has reported that at least 50% of generative AI projects are abandoned after proof of concept because of poor data quality, risk controls, surging costs or unclear business value. Meanwhile, a 2025 survey by RSM US LLP found that 91 percent of middle-market companies are using generative AI in some capacity, but many are still working to integrate it into core operations, underscoring both strong adoption momentum and the ongoing challenge of translating pilots into scaled business value.

That mismatch between spend and value realization is shaping how the mid-market adopts AI. Instead of beginning with automation, many plants are beginning with observability. They are instrumenting operations so that AI can be fed reliable signals, and so that frontline teams can act on what the data infers. 

That adoption logic was on display in a BizTechReports vidcast interview with Lauren Dunford, CEO of Guidewheel, and two leaders from Alleguard’s Greenville facility, plant manager Heath Evans and maintenance manager Steve Billock. Their discussion centered on a pragmatic idea. If factories are going to “use AI,” the first step may be turning analog machine signals into digital data that people can trust, classify, and use to run the plant operations.

Vidcast Interview w/ Lauren Dunford, Heath Evans, and Steve Billock

 At a facility operated by Alleguard in Greenville, Michigan, that signal is amperage. Alleguard is a North American manufacturer of engineered styrofoam products used in industrial, commercial and consumer applications. While the company name may not be widely recognized, its products are commonly found protecting appliances, automotive components, pharmaceutical shipments and in retail stores that carry disposable coolers.

Evans and Billock said the perennial struggle for manufacturers is to optimize uptime, improve overall equipment effectiveness and reduce cost and operational risk. The question was how to harness emerging technologies, including AI, in a way that could be integrated into daily operations and augment, rather than disrupt, the capabilities of the team.

 That challenge showed up as a lack of real-time visibility into performance.

 “You didn’t know if you were winning or losing until the end of the shift,” Evans said.

That is where Guidewheel’s approach came into focus.

Dunford said the premise was straightforward. The plant began by capturing the electrical heartbeat of equipment. Noninvasive sensors were clipped around power lines to measure amperage and power draw. Algorithms translated fluctuations in that signal into machine states such as running, idle or down. Over time, the system incorporated plant context, including operating schedules, and could layer in additional sensor inputs, while remaining compatible with older equipment that lacked modern control systems.

“What we’re capturing is these real-time, tiny little fluctuations of that electrical heartbeat,” Dunford said.

By converting those fluctuations into structured data, the plant was able to see machine performance as it happened rather than reconstruct it after the fact.

Billock described it in practical maintenance terms.

“If you want to know how much a motor’s pulling, you put your amp clamp on the wire,” he said. “All Guidewheel’s done with this particular clamp is taken that signal that we all use to troubleshoot every single day and turned into an ongoing window into real-time data.”

Billock said the team deployed clamps broadly across the plant’s production equipment and then began developing a working understanding of how to interpret the signals under different operating conditions. Each press generates distinct electrical patterns depending on whether it is preheating, cycling or fully producing parts. By monitoring hydraulic systems, total incoming power and hopper activity, the team learned to distinguish between machines that were energized and those that were actually manufacturing. Without that contextual interpretation, preheating cycles could be mistaken for productive runtime, distorting performance data.

From Signal Capture to Operational Discipline

 That technical foundation was only the beginning. The harder task was standardizing how the data was classified, shared and acted upon across shifts and roles. Evans said the plant began in what he called a crawl-to-walk phase, focused on uptime and downtime classification. The goal was to build data integrity before addressing more complex optimization challenges.

“We wanted to ensure the downtime’s getting tagged properly, coded properly,” he said. “Because if we don’t have good data integrity going in, then trying to create solutions for the problems, we’re going to be chasing our tail.”

The plant’s deployment timeline also reflected a mid-market pattern. Leadership wanted to replace paper-based reporting with technology, but it did not start by rebuilding equipment or tearing out systems. Evans said the Greenville site served as a beta plant, trialing the system on one press in October. After extending the pilot, the plant expanded deployment the week before Thanksgiving, supported by Guidewheel staff and then carried forward by Billock’s maintenance team.

 That sequence matters because mid-market manufacturers often face a credibility hurdle with new technology. If a system is seen as “another dashboard” imposed from outside, operators may comply superficially or resist quietly. The Greenville team emphasized that adoption accelerated once operators and supervisors could all see the same version of the truth on the floor.

Guidewheel monitors display machine status across the plant. Tablets at equipment allow operators to classify downtime in real time. Alerts can trigger when downtime persists beyond defined thresholds. Evans said the visibility changed behavior quickly.

“Nine times out of 10, if you give somebody a challenge and they know what the target is, they’re going to go after it and they’re going to win,” he said, “provided you give them the right tools.”

Over the first three months, Evans said the plant’s uptime moved from the low 70 percent range to about 86 percent. He said overall equipment effectiveness, which averaged in the mid-70s in the last quarter of 2025, reached 88.9 percent month to date in February of 2026.

Those gains, the participants said, did not come from a single predictive breakthrough. They came from faster feedback loops and consistent problem definition. Each morning, Evans, Billock and supervisors review downtime events and clean up anomalies in tagging. Each week, they identify the biggest downtime driver and focus on it.

Billock said the team focuses each week on its single largest downtime driver rather than trying to solve every issue at once. The weekly routine also highlights a subtle but consequential shift in plant culture. Visibility tools can easily become punitive if they are used only to find failure. To that point, the team uses the same data to reinforce what is going well.

“It doesn’t just have to be a negative,” he said. “Use it to show them how good they’re doing.”

Evans described a moment that underscored the cultural shift underway. A team lead sent him a photo of the scoreboard showing an hour in which presses were running at 100 percent. She sent it, he said, not as a report, but as a reflection of how the team was performing in real time. The visibility allowed operators to see the results of their work as it happened, reinforcing ownership and pride on the floor. 

The most vivid example of usability illustrated how operational visibility had moved beyond the plant floor. By converting electrical signals into structured digital data, the system made machine performance securely accessible to authorized personnel in real time, rather than confining it to physical equipment or end-of-shift reports.

Billock said that shift has changed how he begins his day.

“I’ve got a three-month-old baby,” he said. “Every morning we get up, I get my coffee, he gets his bottle, and we sit there and go through Guidewheel.”

From his phone, he reviews the prior two shifts before arriving on site. He can see every downtime event, compare it with supervisor notes and identify patterns that may require attention. Instead of walking into the plant to reconstruct what happened overnight, he arrives with context and priorities already in mind.

The difference, he said, is not convenience alone. It is confidence. “It’s so surprisingly accurate,” he said.

In practical terms, information that was once locked inside individual machines and paper logs is now available securely to the people responsible for acting on it. That mobility shortens response times and extends accountability, while allowing plant leaders to remain connected without being physically tethered to the factory floor.

A Universal Signal Across Diverse Machines

 That portability of information reflects a deeper architectural decision.

Guidewheel’s bet is that amperage and power draw provide a universal denominator. If the system can infer machine states from electrical signatures, it can establish apples-to-apples visibility across heterogeneous environments without requiring deep integration into every machine’s control stack.

That design choice also intersects with risk management. Manufacturers have become prominent targets for cyber incidents, and many mid-market operators remain cautious about connecting operational technology to corporate networks. Dunford said the system can be deployed in ways that avoid touching existing networks or equipment controls, which can reduce perceived exposure.

“It’s air gapped,” she said, describing configurations that keep the monitoring separate from the plant network while still sending data to the cloud.

Dunford described a future in which the system learns both locally and comparatively. Each machine has a unique history and usage pattern. But as more machines across more plants are monitored, benchmarking across similar equipment could improve pattern recognition and predictive accuracy. She cited incidents where anomalies have helped teams find overheating risks before they became larger problems.

The Greenville case suggests that the frontier for many mid-market manufacturers may not lie in developing fully automated factories. It may hinge on having analog behavior of machines translated into digital formats that informs AI systems and human decision-making at the same time.