How company deploys business around AI model
AMP Robotics is a promising startup working hard to make the waste recycling future happen. The company is structured around two key instruments AI system designed to distinguish different types of waste similar to the human eye and robotic arms that allow to pick particular elements of waste and sort it.
The current business model of AMP Robotics is to sell complex solution in the form of equipment installed on a conveyor belt moving with a certain speed and de-facto subscription on cloud-based AI services that makes the installed equipment work. The user of equipment is not the sole owner of the accumulated data, instead, it’s used to further enhance the algorithms on the AMP Robotics cloud.
The sorting of waste is very labor-intensive business and technological advancement is constrained by three main factors: 1) it’s hard to recognize a particular item and associate it with the material relevant for further processing, 2) current robotic manipulators are effective in handling solid materials with high speed and precision but it struggles to pick up a soft paper cup that could be jammed in a number of ways, 3) some garbage items are complex and consists of a number of components made from different materials. The AMP robotics solution is aimed at solving the first two problems that hugely expand the number of use cases.
Almost as a by-product company offers its’ customers AMP Insights – a data visualization platform that allows the company to use data gathered by the installed equipment: to assess the quality of the incoming waste stream, measure performance of operations or deploy an alert system. The process is labeled as “digitization of a waste stream”.
Another promising opportunity is identification of logos or waste components that are the part of goods or packages produced by a certain brand. This can help to create partnerships with brands interested in creating closed loops or just track the environmental impact of waste generated by the business. This could also be a valuable source of data for policies and regulations.
Current challenges and opportunities
Limited number of SKUs. The first problem is the conflict between a limited number of items that the system can currently identify and sort and the nature of waste coming from households and businesses. It represents a catch-22 style problem – to make it cost-effective, the company needs to be able to sort all recyclable materials, but to start doing that it needs to start with a few use-cases and gradually develop AI models to incorporate more and more SKUs. Now the company is testing the solution for three types of the incoming waste stream: municipal solid waste, electronic waste and construction, and demolition waste. The selection of those streams specifically focused on the relatively homogenous but high-volume type of waste.
Solid clean waste. As you could notice all current use cases assume solid waste, but a significant part of households and businesses waste nowadays, what we can see in dumpsters on streets around, is just a compressed mass contaminated with liquids and grease. There is no even preliminary idea how to handle this problem without imposing discipline in places of waste generation. At this stage, it looks rather as a regulatory task than a business opportunity.
Low labor cost in developing countries. For decades the United States was sending a significant part of waste to China, which benefited from low labor costs on waste processing facilities. The low cost of labor effectively prevented the emergence of capital-intensive robotized systems for identification and sorting of waste. However, today the situation is changing without any effort from an ecological organizations mainly due to economic growth in China, structural changes in the country’s economy and growing labor costs. The question now is whether the waste streams will be re-directed to other countries that have lower labor costs or the USA, as one of the drivers of technological advancement will decide to solve the problem.
The key issue for the company is to get the models right and this will require huge amounts of data and, what is more important, to develop an ongoing process for educating the algorithm about new items in waste. For the current phase company is doing the right thing focusing on a few use cases with a few big partners that consider it rather as a long-term ecology and social responsibility project. The nature of waste however differs region by region, thereby it’s important not just focus on a few anchor partners but to expand the geography and waste stream sources. So now company needs capital to expand geographically keeping focusing on a few types of waste streams that are technologically easier to handle.
The big prize of revolutionizing the waste management on this planet is unachievable without cooperation with governments. A good idea could be cooperation with the government of a country that is relatively small in terms of population and already achieved a high level of sorting culture on households side. That can possibly showcase creation of closed loop waste management system on a country level and stimulate other countries to change regulations and invest in infrastructure projects.
Re-direction of waste to countries with low labor costs is a type of problem that is hard to work with. However, the problem most likely will be resolved without focused effort from the management. The growth in the volume of operations and manufacturing of robotics, along with fixed nature of AI and software cost will most likely drop the cost for the company offering that will become more competitive with the “pushing abroad” option, which will have a big transportation component.