From AI Curiosity to Practical Execution in Corrugated

Key Insights from Two AI Breakout Sessions at Corrugated Week 2026
By Aleks Zlatic | Founder and CEO | Aurum Intelligence

Corrugated Week 2026, jointly hosted by TAPPI and AICC in Fort Worth, brought together the people, equipment and ideas shaping the next generation of corrugated manufacturing. Ai (as expected) was a central theme and great to see that there were two AI breakout sessions for this important topic affecting members. One conclusion came through clearly: our industry is moving beyond curiosity about artificial intelligence and toward practical execution.

The two discussions approached the opportunity from different directions. Tuesday’s session, AI Extracting and Leveraging Practical Solutions, examined business applications that converters can put to work now. I had the pleasure of moderating that conversation with Jason Hooston of Two10 Technologies, Jason Alvarez of Break the Uncertainty, Chris Blizzard of Acme Corrugated, Tom Trinchi of Jamestown Container and Philip Webb of Pakked. Wednesday’s Practical Applications for AI on the Plant Floor moved deeper into machine intelligence, predictive maintenance and the operating knowledge held by experienced plant personnel.

Taken together, the sessions offered a useful roadmap: start with a real operational problem, apply AI to a manageable body of trusted information, keep people involved in the decision, measure the result and then expand what works.

Practical value is already being created

The first panel made AI tangible through applications already improving daily work. Tom described an internal agent that searches decades of IT files, historical help-desk tickets and 350 ERP knowledge-base articles through natural language questions.  Instead of relying on one person to remember where an answer lives, employees can reach the  organization’s accumulated knowledge in seconds.

Chris shared an equally practical example from shipping. A manager used ChatGPT to analyze a daily report and  identify consolidation opportunities that were difficult to see manually. The result was a significant reduction in  partial shipments, including approximately 27 fewer stops per week. Philip demonstrated how AI can read quote requests and attachments received by email, extract the relevant specifications, create an estimate through the ERP  and return a branded quotation for review in less than three minutes.

These examples matter because they address work every converter recognizes: searching for information, rekeying  documents, preparing estimates, reviewing reports, and coordinating shipments.They also reinforce that a useful  first project does not require every system to be integrated or every data set to be perfect. A focused problem, a
controlled source of information and a clear success measure can be enough to begin.

Plant floor AI depends on context

The plant-floor panel extended that thinking from business processes to machines and maintenance. Kaleb  Bozorgzadeh of SUN Automation Group noted that sensors, cloud storage, and anomaly detection have become increasingly accessible. The harder question is no longer whether something unusual is happening. It is what the anomaly means, what action maintenance should take, whether the alert is valid and whether the intervention  created measurable value.

That interpretation requires human knowledge. The experienced operator who can hear that a machine does not  sound right possesses context that a generic model does not. AI can help preserve that expertise by connecting  machine condition, operator response, root cause, corrective action, and outcome. Over time, the system can progress from asking questions, to suggesting an answer, to recognizing a known condition and eventually predicting it. A useful analogy from the discussion was to treat AI like a new employee: give it limited responsibilities, reliable information, and regular correction before expecting greater autonomy.

Greg Tucker of Bay Cities provided a strong picture of what becomes possible when that foundation matures. Bay Cities has spent roughly six years connecting plant-floor and enterprise information, from PLCs and edge devices through machine systems, MES, and ERP, and into an enterprise data environment. The resulting use cases include real-time monitoring, predictive maintenance, digital twins, dynamic costing, AI-assisted scheduling, estimating analytics and customer profitability analysis. One example made the value especially concrete. SUN Automation’s Helios system detected vibration associated with a component becoming loose before it failed. A failure inside the machine could potentially have caused  hundreds of thousands of dollars in damage. The larger vision is a closed workflow in which a condition is detected,  a maintenance action is recommended, a work order is created, a part is ordered and downtime is scheduled before  the problem becomes a breakdown.

Start with the data you already have

U.G. Wilson of BW Papersystems emphasized that plant-floor AI does not always begin with an expensive sensor or IoT program. Purchasing records, maintenance reports, job histories, and production spreadsheets may already contain valuable patterns. A sudden increase in blade purchases relative to production volume, for example, could point to poor machine settings, excessive sharpening or an operator practice that deserves investigation. The same approach can turn AI into a troubleshooting partner. With controlled access to manuals, fault records, maintenance notes, job histories and vendor portals, an agent can investigate what settings changed before a fault,  whether the condition occurred previously and what corrected it last time. That is more useful than a generic  chatbot because it works with the history and language of the plant. Data ownership therefore becomes strategic. Panelists noted that restricted access to PLC information, expensive  interfaces and disconnected software can limit progress. Bringing needed information into a governed data layer gives a converter greater freedom to analyze its operations without forcing AI to query every production system directly.

A practical path forward

The strongest common message from both sessions was not simply to install AI. It was to connect technology to a  defined business outcome and to the people responsible for achieving it. Map the workflow with frontline employees, choose a narrow use case, protect sensitive information, validate the output, and assign an owner who  will act on what the system finds. Then measure the result in operational terms: fewer stops, faster quote turnaround, reduced searching and  rekeying, avoided downtime, lower inventory, improved margin, or better customer response. Early wins create confidence and help teams learn where more integrated applications will deliver the greatest return. For corrugated converters, the near-term opportunity is substantial. Predictive maintenance, troubleshooting,  knowledge retention, estimating, scheduling, and profitability analysis are no longer abstract concepts. The  technology is becoming more accessible; the differentiator will be how effectively each company adds context,  embeds the insight into a workflow, and closes the loop with human feedback. The plants that begin solving small,  valuable problems today will build the operating knowledge and confidence needed for much more capable systems tomorrow.

Corrugated Week 2026 was held September 28 to 30 at the Fort Worth Convention Center and was jointly hosted by TAPPI and AICC, The Independent Packaging Association.