Cannabis Industry: Predicting the Future with Artificial Intelligence (AI)
By Cameron Bravmann, Business Operations at Green Belt Strategies
The Company
Company X is a player in the cannabis arena. After successfully winning a license for cultivation in a competitive state, they spent the better part of 18 months designing, building, permitting, and finally operating a large-scale commercial cannabis cultivation facility. Over time, Company X has vertically integrated, incorporating additional aspects of the supply chain, to include extraction, manufacturing, distribution, and retail sales. Company X has now been operational for nearly three years. Along with startup time, they have been at this for the better part of 5 years. In order to stay competitive, Company X is looking for ways to improve their operations.
The Challenge
Company X now has their eye on moving into the Multi-State Operations (MSO) game. In order to be successful at this, they will need to be able to emulate the success they have garnered at home. However, merely building and identical building in another place, implementing the same Standard Operating Procedures, and cultivating the same varietals is no guarantee that they will have the same success in the new location. The cannabis industry is rife with companies that have attempted such a feat, and have fallen out of favor in some cases, and out of business in others.
Company X is looking to expand at a time when retail prices are falling, and competition is fierce among the players. In order to successfully move into the MSO category, Company X needs to better understand its production process, so that it can create a carbon copy of the environmental attributes that have helped it becomes a success in its first location, so that it can replicate that success elsewhere. When Company X started to dig into the various parameters of its production, it came to realize that in spite of its current success, there were far too many unknown variables in their process, which needed to be better understood, in order to successfully make the transition to MSO.
Up until this point, Company X has been relying on an ERP system that allows for diligent tracking of the multitude of inputs/outputs that are related to its production process, including: Vendors, Customers, Sales, HR, Finance, Product, Compliance, Scheduling, and Supply Chain. They have been able to drive down their cost of production to remain competitive in an increasingly intense marketplace, but have noticed variability in their output, including potency and yield, as well as (at times) pest infestations which have caused damage to crops and finances. Currently these effects do not appear to have any pattern, however, Company X wants to be able to control for these, in order to better predict outcomes, and to create a more standardized product across multiple venues.
Company X is considering Artificial Intelligence (AI) to help it improve its understanding of its production process, so that it can better predict outcomes, create a consistent product, and better manage scarce resources and apply them correctly. According to some, “The global artificial intelligence in agriculture market is witnessing a CAGR of 26.2% during the forecast period (2019-2024).”1 Others, including Nicolas Martin, assistant professor – Department of Crop Sciences (Illinois) states “We’re trying to change how people run agronomic research. Instead of establishing a small field plot, running statistics, and publishing the means, what we’re trying to do involves the farmer far more directly… We can detect site-specific responses to different inputs. And we can see whether there’s a response in different parts of the field.”2
The first steps in the process are for Company X to dig into their current understanding of their Key Performance Indicators (KPIs), in order to decide where the biggest value gains can be made. Company X has decided to work with an AI development group, in order to better understand plant characteristics. After working with industry consultants and the AI development group, Company X has discovered that it has been the beneficiary of fortunate circumstances in many cases. Through several sessions, it was determined that Company X was unable to answer some basic questions about its operation, mostly concerned with environmental controls, including temperature, humidity, and air quality, as well as with regards to its Integrated Pest Management (IPM) program, which has been hit or miss mostly.
From time to time, Company X has experienced plant morphology due to these factors but has taken no concrete steps to address them. Moreover, unpredictable pest infestations can cause havoc in the operation, grinding production to a standstill. In order to create a unified experience for its customers, as well as a predictable production schedule free from unforeseen problems, Company X needs to understand environmental factors better, so that it can replicate its current success in other jurisdictions, with a focus on production predictability, notably product quality and consistency.
The Strategy
Company X has begun working with industry consultants in order to develop new KPIs related to the environmental conditions at its production facility. In addition to tracking things like labor hours, consumable inputs, and utility costs, Company X will now be collecting/studying data related to temperature, humidity, nutrient concentrations, and light intensity. Additionally, Company X will begin looking at and analyzing crops for pests and disease. The premise is that Company X needs to start creating large data sets, in order to create Machine Learning (ML) algorithms, which will be “trained on the data sets and then learns on its own”. 3
The application of ML into agriculture is nothing new, but its application in the cannabis space is in its relative infancy. In the past, cannabis producers benefited from high prices due to the illegal nature of their business. As legalization sweeps the country/world, producers will be moving from an illicit marketplace mentality into one that more closely resembles that of the agriculture/horticulture industry, where the final product price more accurately reflects input costs. This will of course lead to a drop in the price of the final product. In order to remain competitive, Company X will need to adopt practices that take some of the daily operating decision making out of the hands of its supervisors and front-line employees, and puts it in the “hands” of the algorithm, which “teaches itself to flag something as small as an individual insect, long before human would usually identify the problem.”4
80 Acres Farms, a Cincinnati-based indoor grower, has recently opened a fully automated facility, which is monitored by AI at every step of their production process. “We can tell when a leaf is developing and if there are any nutrient deficiencies, necrosis, whatever might be happening to the leaf… We can identify pest issues, we can identify a whole variety of things with vision systems today that we can also process.”5 By being able to identify things of this nature, through the use of cameras and computational programming, Company X ought to be able to move ahead of its peers when it comes to process control, and this could help to create a moat around its business.
Key Benefit Areas
Company X has begun to develop a new set of KPIs for this next stage in their business cycle. By establishing new metrics, they have been able to tighten controls across production, and are starting to reap the benefits of doing. Some of the areas that have been most dramatically affected include:
Nutrients
Company X has been able to dial in nutrient concentration based on plant needs. Whereas prior to implementation, all plants were treated as a batch, and there was no particular methodology when deciding on batch composition. After implementation, Company X was able to identify three distinct nutrient programs: High, Medium, and Low. By organizing batches with plants that all meet a certain nutrient profile, Company X has managed to increase productivity of strains. Additionally, instances of nutrient deficiency and toxicity have disappeared, now that this information has been brought to bear on the operational workflow.
Light Intensity
Similar to the nutrient profiling, three distinct groups of plants have been identified in terms of light intensity: High, Medium, and Low. This has added another level of granularity to production. Now plants can be grouped by how much light they require, along with their desired nutrient concentration.
Integrated Pest Management (IPM)
Plants are now continually monitored for signs of pest infestations. The use of AI makes it possible to monitor plants 24/7, something that is simply too difficult for humans to do. Moreover, Company X is now able to identify things early, meaning that steps can be taken to address small issues before they become big problems. In the past, a generic IPM program was rolled out on a scheduled timetable, which was effective a lot of the time, but not always. Under the new regime, Company X is able to identify issues before they become problems, allowing for quick mitigation, and no disruption to the production schedule.
Key Cost Areas
Key costs included both personnel and the tools required for AI; notably cameras for ongoing observation, program developers to write the application, and personnel to organize the data sets, which will be only semi-structured at first. Industry consultants provided the internal support for the AI program, in order to minimize the impact that would be felt on the staffing side. However, consultants are not inexpensive. Moreover, there is the challenge associated with trying to run daily operations, while simultaneously building a forward-looking business predicated on data gathering, which needs to be able to function in the background, while daily operations continue in earnest.
During the development of the software, the consulting group re-wrote current standard operating procedures, in order to assist in the forthcoming training that would be required for all permanent staff, in order to bring them up to speed on the new processes. Additional data scientists have assisted in developing KPIs, and scrubbing the data sets, in order to start to make meaning out of it.
Lessons Learned
Company X has benefited greatly from the adoption of its new Artificial Intelligence (AI) platform. The first year after implementation saw great strides made with regards to product quality and consistency. Pest infestations have been mitigated to a point of non-issue. Operational workflow has benefited from better understanding plant requirements, creating opportunities to group varietals based on feed and light requirements, as well as photo-period constraints, allowing Company X to optimize its throughput. This has had the compound benefit of making it possible for both the Sales/Marketing department to see what is coming down the pipeline, as well as providing the opportunity for them to request varietals based on consumer demand, allowing the production side of the business to work in better harmony with those requirements, creating a better relationship with its Customers, as well as its vendors.
Moreover, by using AI techniques to better understand their operating environment, Company X has had the ability to expand into new territories, without fear of loss of quality/consistency with their product. So while it is relatively easy to quantify the monetary benefit of a successfully run business that operates in multiple states, the ability to maintain consistent and predictable quality across the business portfolio has the added benefit of increasing “goodwill”, an intangible attribute that can exponentially increase company value.
Calculating the ROI
Company X captured Direct Benefits through increased productivity, including precise nutrient dosing based on plant feed requirements, precise light management based on plant photoperiod requirements, and improved IPM outcomes based on early detection and eradication. All of these have contributed to a 20% increase in yield in Year 1. It is expected that this will be built upon in Year 2 and 3. Moreover, Company X has tighter controls over its final product. Learning and assimilating this information into their business practices has made it possible to execute a MSO growth plan, with continuity and consistency of quality. Company X has also captured Indirect Benefits, including staff productivity and improved customer satisfaction. While harder to quantify, it is undeniable that this has had a symbiotic effect on the business, both up/down stream, and with regards to its future growth strategy.
- “Global Artificial Intelligence in Agriculture Market, 2020-2025”. Globe Newswire, Feb 28, 2020. https://www.globenewswire.com/news-release/2020/02/28/1992500/0/en/Global-Artificial-Intelligence-in-Agriculture-Market-2020-2025.html
- “New artificial intelligence algorithm better predicts corn yield”. University of Illinois College of Agricultural, Consumer and Environmental Sciences, Feb. 20, 2020. https://www.sciencedaily.com/releases/2020/02/200220130500.htm
- Daniel T. Allen. “Farmers are using AI to spot pests and catch diseases — and many believe it’s the future of agriculture”. Business Insider, Nov. 8, 2019. https://www.businessinsider.com/farmers-artificial-intelligence-in-agriculture-catch-disease-pests
- Daniel T. Allen. “Farmers are using AI to spot pests and catch diseases — and many believe it’s the future of agriculture”. Business Insider, Nov. 8, 2019. https://www.businessinsider.com/farmers-artificial-intelligence-in-agriculture-catch-disease-pests
- Daniel T. Allen. “Farmers are using AI to spot pests and catch diseases — and many believe it’s the future of agriculture”. Business Insider, Nov. 8, 2019. https://www.businessinsider.com/farmers-artificial-intelligence-in-agriculture-catch-disease-pests