Manufacturers everywhere are grappling with three persistent challenges: a shrinking skilled workforce, the pressure to digitize operations, and the mandate to do more with less. In the latest episode of MetaPod, host Ron Crabtree speaks with Bryan DeBois of RoviSys about practical ways AI can solve plant-floor problems without disrupting uptime. The discussion focuses on three diagnostics leaders can use to find gaps and prioritize action.
1. Expertise Loss and the Risk of Tribal Knowledge Walking Out the Door
The “silver tsunami” is real. As veteran operators retire, the know-how that once lived in their heads disappears. Bryan cites a dramatic shift in tenure that many plants feel every day, and he frames a simple litmus test: if performance depends on who is running the line, you have a tribal knowledge problem.
To address this, Bryan’s team applies machine teaching to pair historical time-series data with expert decisions. By interviewing top operators and encoding their heuristics, they capture the subtle, unwritten tips that never make it into SOPs. Those skills then seed autonomous AI agents that learn in simulation through deep reinforcement learning. Trained safely offline, agents can match or even outperform veteran operators, and later guide novices through decision support recommendations on shift.
2. Decision-Making Bottlenecks: From Reactive to Proactive
Plants often rely on manual judgment, which leads to inconsistent choices across shifts. Bryan urges leaders to ask where decisions are delayed or disputed, and whether teams are reacting to problems rather than anticipating them.
Adopt predictive AI that can forecast outcomes like batch quality with high confidence before scrap or rework is inevitable. Start with AI in decision support mode so a human stays in the loop while trust and validation build. Only after sustained proof should teams progress to closed-loop control. This staged path respects plant realities and maintains uptime.
3. Missed Optimization Opportunities and the Limits of Dashboards
Many plants collect data but still invest according to the “loudest voice in the room.” Bryan shares how one client used analytics to shift its culture to data-driven prioritization. Another moved from static dashboards to action boards, where every chart triggers a specific action, and operators can flag nuisance alarms that feed continuous improvement at corporate.
What to do about it:
- Use analytics to guide investments toward high-impact hotspots rather than gut feel.
- Replace passive monitoring with closed-loop intelligence that recommends precise adjustments.
- Optimize for changing conditions instead of static set points to avoid leaving throughput, quality, and energy gains on the table.
Generative vs. Predictive vs. Autonomous AI
Bryan clarifies the alphabet soup:
- Generative AI covers tools like large language models as well as image and video generation.
- Predictive AI uses historical data to forecast future outcomes so teams can act early.
- Autonomous AI learns in simulation and can recommend or set set points in real time after validation, helping plants run consistently even as materials, equipment, and conditions change.
Connect and Learn More
If you have further questions or are interested in learning more about how to implement autonomous AI in your manufacturing organization, you can get in touch with Bryan by connecting with him on LinkedIn or by visiting rovisys.com. You can also get in touch with Ron via email at metapod@metaexperts.com or connect with him on LinkedIn.
MetaPod offers a wealth of knowledge and expertise for operational executives and organizations seeking to optimize their operations and supply chains, as well as common challenges such as digital transformation, the forever labor shortage, and doing more with less. Listen and subscribe today for more insightful and engaging discussions on operational excellence, leadership, growth strategies, and organizational transformation.