As Woodi* evolved from a quaint carpenter shop into a bustling full-scale factory, its volume of orders, production, and shipping consistently swelled. While the company expanded, bringing in new staff and refining old processes, some elements remained unchanged. This constancy can be beneficial, such as in preserving core values and traditions; however, it can also pose drawbacks. For instance, key personnel often continue performing mundane tasks because they’ve developed a sixth sense for efficiency, and converting this intuition into teachable skills can be quite a challenge.
The problem
During a trade fair, representatives from Woodi and Digisalix discussed their respective fields and challenges. A significant opportunity to simplify a laborious and repetitive task through automation emerged from these discussions.
The task involved packing every order onto pallets for delivery, typically in a container on a lorry. This process is a critical cog in the furniture sales cycle, requiring meticulous planning of each pallet’s arrangement to ensure logical and efficient product placement. Surprisingly, this planning must occur before production or even the finalisation of the actual order to provide the sales department with accurate shipping cost estimates for quotations.
This repetitive packaging task presented a golden opportunity to transfer knowledge and achieve considerable time savings through automation, not to mention reducing dependence on a few ‘furniture Tetris grand masters’. Moreover, if goods could be snugly fitted onto four pallets instead of five, it was possible to cut shipping costs by up to 20%. More compactly packed pallets also meant fewer lorries essentially hauling air across the countryside, making this approach economically and ecologically sound.
The solution was clear: let a machine take over the planning!
But how to proceed? Did we need AI, a computer farm, or perhaps a neuro-wizard? Not in the slightest. A couple of PhDs from a nimble machine learning company concocted a traditional optimisation algorithm, and a user-friendly interface was built on top of it. This setup allowed anyone to review the machine’s output, tweak it if necessary, or teach the machine new packing techniques. As the palletising system is integrated with the ERP system, both sales and the packing departments receive the data simultaneously, and all this information is neatly archived for future use.
Smartificial Intelligence?
The role of AI in this project was insignificant. Chat-GPT might have thrown in a few snippets of code, but the system primarily relied on more conventional methods. We tackled the problem with a straightforward optimiser that uses previous orders as templates for new ones. The existing data dictates the packing rules, eliminating the need for any complex model training or upkeep, thus giving us the perks of machine learning without its usual headaches. Moreover, the system is continually updated as new packing solutions are added to the database.
Unexpectedly, the most daunting aspect of the project wasn’t developing the algorithm itself but ensuring seamless and logical collaboration between humans and machines. After several design iterations and rigorous testing, we finally nailed it, though that saga is best saved for another blog post.
In the end, by wielding relatively simple methods, we automated Woodi’s process, resulting in notable time savings and a reduction in stress. All it took was an open-minded, forward-looking client who embraced the potential of modern computing and a capable supplier. Thus, another burdensome conveyor belt task was significantly streamlined.
What should we simplify next?
*Woodi is a Finnish furniture manufacturer with a mission to create high-quality, sustainable, and inspiring environments for kindergartens, schools, and care homes across Europe.
