Improving product quality control with an AI agent

High quality and safety requirements

Thomas Regout is a manufacturer and supplier of telescopic guide systems for various applications. Not only for drawers and warehouse systems, but also for vehicles and ATMs, for example. This sets a high bar for quality requirements. Martijn Bouwens, Quality Engineer at Thomas Regout International: "We work according to various standards, such as ISO 9001 and IATF 16949, a standard specific to the automotive industry. A sector that imposes high safety requirements. As a result, we are required to register all deviations in products. We then follow up according to the 8D problem-solving method."

800 reports per year

This averages about 600 reports per year. "With our old Microsoft Access database, it was a challenge to filter information from it afterward," Martijn explains. "For example, to see if problems had occurred before and which solutions were chosen then. Moreover, it was always difficult to search for specific details, such as article numbers, types of deviations, and experiences with solutions." Generally, it was more efficient to simply ask colleagues. Hours and sometimes entire workdays were lost.

Faster and easier

At Thomas Regout, they wondered if an AI agent, based on Microsoft Copilot, could provide relief. For Axelio, it was a great test case. The first tests with the AI agent are now behind us, and Martijn is enthusiastic: "Axelio trained the AI agent to find answers in our database. We designed it so that it first looks broadly and then you can zoom in on details. The AI agent works especially quickly and easily. Moreover, you can be sure that you miss nothing. And the delivered data is always correct."

Preventing standstill

This makes the AI agent, according to the Quality Engineer, an important asset for a manufacturing company like Thomas Regout. "You want to ensure a solution as quickly as possible when an error is detected. This way, you can prevent standstill and longer delivery times," emphasizes Martijn. "Moreover, it's nice to notice that the agent takes everything into account. Especially when you work for a critical sector like the automotive industry."

Hitch in the cable

To achieve the end result, several barriers had to be overcome. Martijn explains: "First, we wanted to gain clear insight into the internal reports. Together with Axelio, we outlined exactly what the desired output was and how the agent could best search in our view. Initially, Axelio worked with two years of raw data from our old quality management system. In the first test, the AI agent seemed to work well. However, the quality of the retrieved information was too sparse. But fortunately, by that time, we were able to link our new quality management system, ManualMasters, to the agent."

"The nice thing is that the AI agent thinks proactively instead of reactively."

Quick switching

However, the team was not there yet. Martijn: "The data from our new system turned out to be indexed, making the information behind the codes unreadable. The nice thing is that Axelio does not give up at such moments and remains open and honest in communication. We put our heads together for half an hour, and the problem was solved. You realize that we understand each other very well. We decided to temporarily export all data from ManualMasters to an Excel database. We linked that to the AI agent. This made the data readable. Thanks to that step-by-step approach, we now also know for sure that it is worth naming all the codes from the indexed columns."

Growing expectations

Martijn shares that the AI agent fully meets all the expectations they had at Thomas Regout a year and a half ago. He laughs: "You only notice that those expectations are growing. Both from our side and from Axelio's. Many things that were not possible in Microsoft Copilot two years ago are now possible. As a result, you want to give increasingly complex tasks. The question for us now is actually: where do we draw the line?"

Proactive thinking

Thomas Regout can move forward in any case. "But there are still things we want to refine further with Axelio," Martijn adds cheerfully. "In the future, we want to use the AI agent more broadly within the organization. The nice thing is that it thinks proactively instead of reactively. This way, it can also work predictively for us and, for example, help a Design Engineer in designing new products, taking into account what could go wrong. If we could ever fix that, it would save us a lot of time and effort."