Something shifted in Fort Worth last month. The AI sessions at Corrugated Week 2026 weren’t about whether AI belongs in a box plant. They were about what’s already working. One converter described an agent that searches decades of help-desk tickets and ERP knowledge articles in plain English. A shipping manager used AI to read a daily report and cut roughly 27 truck stops a week by consolidating partials. On the floor, a vibration sensor flagged a loosening component before a six-figure failure.
Board Converting News boiled it down to a playbook most operators would sign: start with a real problem, give AI trusted data, keep a human in the decision, measure, then expand.
The constraint isn't the model. It's the plumbing.
Read closely, the event panelists named the real bottleneck, and it wasn’t algorithms. It was restricted access to machine data, expensive interfaces and disconnected software. The sensor can tell you something is unusual. Only context can tell you what it means: which job is on the machine, what speed the quote assumed, who’s running it, what fixed this last time.
That context lives in your ERP, your scheduler and your MES. If those are three products from three vendors, every AI project starts by rebuilding the stitching. The plants with the strongest stories at Corrugated Week had spent years connecting PLCs, MES and ERP into one governed data layer before asking AI to do anything.
From equipment performance to plant performance
This is where MES is heading, and it’s the direction Amtech laid out at our user conference this past October in Nashville. Most MES products, ours included until now, optimize equipment: uptime, speed, waste. The next step is plant performance. A quote wins a job at an assumed machine speed. If the machine runs slower, the order is at risk of being late and over cost, and most teams find out at job close. The platform we’re building, Team Performance, takes live MES and ERP signals and alerts the people who can fix it while the job is still running.
That only works because Amtech’s quoting, scheduling, production monitoring and ERP already share one data model. After 40-plus years building software specifically for corrugated, that’s the asset AI needs most.
Why it matters this quarter
Box shipments were down in Q3 and new capacity keeps arriving. In that market, margin comes from execution, not volume. For owners, that makes AI ROI a function of integration cost: the fewer seams, the faster the payback. For VPs of Operations, the value is closing the loop during the job, not reporting on it afterward. For CIOs, every bolt-on is one more interface to maintain and one more surface to defend at a time when ransomware is the top operational threat to packaging manufacturers.
What to do today
Pick one job that ran long last week and trace it: Did the quote’s assumed speed make it to the schedule? Did the schedule make it to the floor? Did the floor’s actual speed make it back to costing? Every handoff that’s manual is where AI will stall. Then inventory your plant-floor software. If MES, scheduling and ERP come from different vendors, count the integrations you’re maintaining and ask what each one would cost to rebuild for an AI use case.
If you’d rather see what a connected quote-to-production flow looks like in a corrugated plant, talk to Amtech about EnCore, MES and scheduling for corrugated. Bring that one long-running job. We’ll show you where it would have been caught.
