“Trust your gut” will only get you so far.
Today, manufacturers need to trust their data, as well as agentic AI’s ability to make data-driven decisions. They also need to trust that those closest to the work—the people on the shop floor, not the data engineers—can build solution-finding software. It starts by asking the right questions.
The implementation of CreateOne’s AI-driven platform at Grupo Beta bears this out.
The vertical manufacturing operation in Tegucigalpa, Honduras, that produces two to three million garments a month uses AI agents and other digital tools to provide a single version of the truth, reducing the dreaded “data noise” and improving decision-making. With more than 2,000 interactions and 110 active users, from the shop floor to the CEO’s desk, the platform is quickly replacing “gut feelings” with data-driven decisions
The result has been reduced lead times, improved efficiency and enhanced supply chain management.
Grupo Beta’s rapid proof of concept was built in just three weeks—by the team itself. Users are able to prompt the AI to help identify—and solve—highly specific problems using natural language.
“This wasn’t a ‘cold start,’ but rather a platform that started right away understanding the customer’s business. And it is built on a novel architecture where we create an actual economic model of the business,” said Lawrence Lenihan, executive chairman and co-founder of Resonance, the parent company of CreateOne.
With trade fluctuations and variable input costs, the business needs and outlook for apparel manufacturing business change constantly, opening the need for a platform that is responsive to these challenges. “Once the answers could be trusted, people started asking for things that did not exist,” said CreateOne president Daren Hull.
For example, a production planner might ask, “Where are we going to run out of capacity?” A buyer might ask, “Which supplier prices changed against our open purchase orders?” A materials team might ask, “What will we need four weeks from now?”
Instead of receiving text answers from an AI chatbot, they receive a working tool built from the factory’s own systems and spreadsheets. This single source of truth is invaluable when so much of a company’s data is siloed, not to mention scattered on Excel, PDFs, even hand-written notes.
Knowing What You Don’t Know
The difference with smart agentic AI is that it can look for problems companies might not even think to ask about.
Using data from Grupo Beta, the platform helped put the company at the forefront of operational innovation in ways manual workflows couldn’t.
According to the company, rather than simply flagging issues after the fact, the AI proactively uncovered significant, recurring monthly savings. Initial deployment identified SKU optimization strategies to immediately recover more than $100,000 per month in lost revenue. Further, by consolidating and analyzing events across the business, the AI identified more than $425,000 in supplier chargebacks that had gone unnoticed.
Looking ahead, and benchmarking operations against Grupo Beta’s own best-performing weeks, the system demonstrated predictive intelligence by tracing supplier price changes back to individual purchase orders. It then alerted leadership to adjust budgets and protect margins across dozens of affected parts. This generated another $100,000 in monthly savings.
“People will ask, ‘I had these defects on this job yesterday, what can I do to fix it?’ and the AI will give out all the details—all the reasons why it was bad, what the defect rate was, what the percent rate was,” said Rodney Rozell, senior leader of manufacturing and operational supply chain organizations at Grupo Beta. “Then it will give another block that says, ‘Beyond Your Question.’ It will infer some things and say, you didn’t see this, you didn’t see this, you didn’t see this. You may want to look into it.”
Rozell cited a problem the manufacturer was having with different rates coming off different production cells. Because the platform was linked to Grupo Beta’s database, it was able to correlate the problems with the age of each respective machine.
“Then it handed out assignments to our engineers,” he said.
The ability to ask questions in natural language is also more important than companies might realize, especially across different levels of technical expertise, experience or education.
“Look at people in my trims warehouse, my fabric warehouse, or even Central America or Third World Countries in general. This has liberated them from issues of education level,” said Rozell. “They know what they need to get done, and they can go in and ask the system the questions, then tell me where the variance or error was made.”
That it’s all done via tablet or mobile phone means the platform can go wherever the workers go, and where problems lie.
The approach is less about “building a more efficient yesterday’s plan,” said Rozell, and more about creating an entirely new type of vertical integration, with connections that boost efficiency for everyone involved.
Hull concurred. “We all hope that AI will help these companies react to what is a new way of working. [For so long], people have been spending money on software and processes but is their performance getting better? We now have a system that’s making suggestions on how to get better, measuring how much was actualized, then making additional recommendations on how they can capture additional economic value, and creating a learning loop.”
