Machine
Learning
Training
Grounds
intelligently automate
Goodwill of Western Missouri and Eastern Kansas is a nonprofit organization that provides direct employment, workforce development services, and training in the Kansas City area. We are well known for our resale stores, which are one method of both executing and funding our mission services. Our mission is to empower people to discover their potential and adapt for the future through the power of work. We serve individuals with disadvantages and different abilities, individuals who are under- and unemployed. Through jobs and career advancement, individuals become self-sufficient and communities become prosperous. We have been committed to this effort for 125 years.
We have a plan to revolutionize our retail and mission models by embracing future state technologies. We are developing one system that will expand and improve donation production, and enhance the customer experience. The system will automate the identification of items using robotics; use machine learning to sort for best use of commodities among e-commerce, salvage and stores; create a digital inventory of donations to be sold via omni-channel e-commerce and in the store, simultaneously; and leverage mixed realities to create a customizable, treasure hunting experience in our stores using a cell phone. This is innovative, as the first technology to classify discrete random clothing by size, type, brand and quality. And our unique model as a business and a service provider allows us to leverage our 14 stores, and Goodwills’ 3200 stores, to make a business case with a mission mind.
We have proof of concept on the ability to use machine learning to accurately identify clothing by size and type – with a 90% confidence rate over 328 images.
Our business model plans implementation out for five years. We are currently in the middle of the first stage – proof of concept. Now that we know we can identify the basics of size and type, are testing how specialized a machine learning sort system can be, and at what confidence level. We recently put out an RFP to test other categorization including fabric type, quality, and brand. Our chosen supplier - York Exponential - will create the algorithm, execute the test, and initialize a data lake for donations we receive.
The next step is developing a good, better, best pricing that leverages dynamic sale point data from online platforms, so we can price items by demand and maximize revenue generated, all of which will be integrated in a tagging system.
After that, we must conduct a process mapping assessment to design an automated system and build a prototype that uses automated garment-on-hanger systems and robotics to augment our production workforce, powered by an "A.I. black box" and camera system for categorization and integration. The mechanical functionality is under evaluation with the potential opportunity to utilize these systems to move diverse garments, varying in shape and size from women/men’s tops, women/men’s pants, women’s dresses, children’s clothing from infant size and up.
Within three years, we should be able to have our digital inventory placed into omni-channel e-commerce pipelines while utilizing RFID-type technology to track through donation life-cycle from processing, to the sales floor, and POS system. In four to five years we should be able to roll out our augmented reality and virtual reality shopping platforms (in-store, on mobile, and at home) by monetizing/leveraging the data we will collect on used goods.
Throughout this process, we will make our work and systems available to our employees and our participants. We will partner with industry to create pipelines of talent trained in systems such as AI, hardware automation, and robotics. We will partner with educational institutions to build and award credentials and certifications. We will have on-the-job learning available for Goodwill employees, interns, and participants. And we will remain focused on the most vulnerable populations.
