It may seem odd that in a modern society we still rely on bees to a large extent for much of the food that ends up on our dinner plates, yet that is no overstatement, it is a fact. Even with all the advancements in agriculture that human civilization has made over the last century and a half, bees are still a vital part of food production. The chief reason for this, as many are probably already aware, is pollination. A Skolkovo resident may just have the solution to the problems facing commercial bee colonies and the answer lies in artificial intelligence.

Image source:

Honey bees are one of the most important pollinators for commercial crops and without them, our food supply would be in jeopardy. In recent years, articles have appeared in the news and social media about bee colony collapse, a phenomenon that occurs when between a third and up to 90% of worker bees abandon their hive and seem to just disappear. Experts have attempted to establish the cause of bee colony collapse, the proper term being colony collapse disorder (CCD), but the results of these studies remain mixed, with papers covering the issue pointing not to one but to multiple causes: habitat destruction, fungicide and pesticide use, pathogens, parasites, pests, and climate change to name a few. Others have pointed to the increased use of monocultures (single crops covering a large area) as leading to insufficient nutrition for bees, as well as stress caused by transportation for commercial pollination. The USDA Agricultural Research Service acknowledges the lack of proven cause of CCD, but also points to the possibility that it is not one, but rather a combination of factors responsible for the phenomenon.

It’s difficult to put a value on bees alone, but given that agriculture across the world relies heavily on pollinators, it is safe to say that their value is immense. The U.S. Food and Drug Administration noted that over 90 commercial crops grown in the United States rely on bee pollination and that approximately one-third of the food eaten by Americans comes from crops pollinated by honey bees, including apples, melons, cranberries, pumpkins, squash, broccoli, and almonds, to name a few. Such is the impact of bees on crops that the value they offer as pollinators is estimated to be anywhere from ten to twenty times the total value of honey and beeswax – $15 billion for added crop value in the US alone. Efforts are already underway to tackle and conduct further research on CCD and the European Commission has also set aside over a hundred million euros to support the beekeeping sector. Beekeepers in the US have offset the impacts of CCD by simply keeping more beehives, according to a Washington Post article. This has effectively doubled the price of honey since the mid-2000s and beekeepers are charging around twice what they previously charged to pollinate commercial fruit and nut crops. Put simply, the human labor required to manage and transport healthy bee colonies has increased significantly. Yet this need not – excuse the pun – “bee” the case with artificial intelligence and computer vision technologies, which we will get to in this article.

CCD has been an ongoing issue since it first appeared in the mid-2000s, and in Russia, the phenomenon occurred in approximately one hundred thousand hives (the country is home to 3.1 million hives as of 2021, placing it sixth in the world). Also, since monitoring bee colonies is labor-intensive, especially on large commercial farms, an automated system might be just what the food industry and beekeepers have been looking for, especially when hives might be at risk. The ability to spot the signs of a bee colony preparing to swarm and split – this is when half the colony migrates elsewhere to create a new hive – is crucial for beekeepers to take the appropriate preventive measures early on to keep pollination, and thus food production, at a stable level. It is also important to be able to spot issues in the colony related to the external environment such as temperature, pressure, humidity, and so on. In the context of CCD, it is a useful tool for accumulating data on the phenomenon. The new artificial intelligence and machine learning technologies required to monitor the factors affecting bee colonies both remotely and in real-time are already here. According to Maxim Denisov, the CEO of BumbTech (a Skolkovo resident), his company is the only known entity in Russia that has such a solution made for the task.

Maxim Denisov, CEO of BumbTech. Image source:

Maxim Denisov is himself the son of a beekeeper and, to use his words, he created the BumbTech company by chance. It was while helping his father manage the beehives at the family dacha south of St. Petersburg that he began to think of ways to make it easier. That is how he came up with the idea of creating a remote technology.

BumbTech AI Pollination Platform

The story began in 2016 when Maxim and his team needed a project to take part in an IoT (Internet of Things) hackathon run by Intel and the Higher School of Economics. They created their startup, developed the AI technology - BumbTech AI Pollination Platform - to monitor bee colonies, won the hackathon, and got one of Russia’s leading experts, Professor Alfir Gabdullovich Mannapov, onboard to help with pushing the company further in its development. Professor Mannapov is the head of the beekeeping department at the Moscow Timiryazev Agricultural Academy, and he helped the BumbTech team define the parameters needed to monitor honey bees and bumblebees. Maksim himself is an engineer but also underwent training in beekeeping to understand the needs of beekeepers, whether on a small- or industrial scale.

Professor Alfir Gabdullovich Mannapov (left) and Maksim Denisov (right). Source:

The BumbTech AI Pollination Platform is a combination of artificial intelligence, data analysis, machine learning, and computer vision. The system is installed using cameras and sensors placed in and near beehives to gather data, and thus accumulate data sets on the insects’ movements and the conditions in the surrounding environment such as temperature, lighting, air pressure, and carbon dioxide concentrations. The information is then uploaded to cloud storage where it is processed and beekeepers can access it. This makes it possible to monitor processes that lead to problems such as swarming and, with an accumulated dataset based on previous observations, the system can forewarn beekeepers; in this way, they can monitor their bee colonies remotely and spot potential issues before they take hold.

Where is this technology applicable? When BumbTech’s team began working on its project, they quickly realized that the issue they were trying to tackle was not specific to Russia, but elsewhere as well. Even in a small, localized setting, it is surprisingly difficult to keep tabs on bee colonies; without enough attention, half the colony can easily swarm and fly away to create another colony elsewhere, which is very costly for beekeepers, whether commercial or private.

The company has yet to find its place in the agritech market, but it is seeking out commercial clients. Russian beekeepers, unlike those in the west (the US in particular), are by and large small entities rather than industrial setups used to pollinate large areas of crops. That is why BumbTech is primarily focused on finding clients abroad. The company has scaled up its technology so that it can be implemented in an industrial setting with the idea of stabilizing the bee population and the pollination process on large farming complexes. In sprawling greenhouse farms such as those found in Spain and the Netherlands, bees pollinate crops on a grand scale and it is in such settings that BumbTech’s AI platform will be most applicable.

BumbTech Pollination Platform. Source:

According to Mr. Denisov, bumblebees are usually used to pollinate greenhouse crops, because they can tolerate the environment better – temperature, humidity, pressure (this is increased to prevent parasites entering greenhouses). Yet the BumbTech AI Pollination Platform isn’t just designed for bees and bumblebees; it can be adapted to monitor other insects as well.

Mr. Denisov stated that in using this AI tool, beekeepers will ultimately save money by helping keep pollination up, improving crop yields, and reducing defect rates. Beekeepers also won’t necessarily be compelled to keep as many colonies to compensate for CCD losses and will be able to accumulate datasets on it.

“An Italian farmer approached us at an agriculture expo in the Netherlands and was interested in our technology,” said Mr. Denisov, “because in springtime, when his cherry trees were flowering, the bumblebees were supposed to pollinate them and didn’t because it was still cold. The farmer didn’t notice because he was busy with other tasks. The result was a cherry orchard with no fruit. Our system allows you to monitor this to prevent crop loss. We can see with video sensors how many bees or bumblebees are going in and out of a given hive and can process this with machine learning. By observing their activity in this way over the course of days or weeks, we can run predictive analytics based on an accumulated dataset. In this way, we know when we need to make changes to the hive, and so on. This is all aimed at keeping pollination at a stable level.”

BumbTech has already received funding from the Skolkovo Foundation in the form of a five million ruble grant via the “Minigrant” program. They have since been focusing on perfecting their solution for greenhouse use through their work with LipetskAgro, a greenhouse complex located in southern Russia that grows vegetables all year round. The company also cooperates with another greenhouse farming complex in the Ryazan region to the southeast of Moscow called “Ryazan Vegetables.” BumbTech has a foreign partner in the Netherlands, although that project has been on hold since the coronavirus pandemic began.

“At the end of 2019, we ran a pilot project with our Dutch partner together with an independent greenhouse research center,” said Mr. Denisov. “We were supposed to run some more experiments last year, but that all changed when the borders closed; right now, we’re waiting for everything to go back to normal.”

Mr. Denisov stated that while the system initially only worked at 50% accuracy, but this has increased significantly as his team has developed it further to 90% accuracy.

“Our company is ready to sell our service for managing beehives in greenhouses. We are ready to go on the market. We have already done two pilots and we are in negotiations with foreign clients for pilot projects as well.”

The company charges an installation fee and a subscription for access to its database. Installing the system is straightforward but still requires specialists to do the job, and the company is currently working on a plug & play version that will make it easier for clients to install the system themselves.

“Our system is very flexible,” said Mr. Denisov. “It can be used in fields, gardens, city farms, and greenhouses. In one way, we are in fact competing with human labor. In Japan, for example, which is a highly developed technological society, they employ people to monitor bees in the greenhouses. They have to come and check the bee colony and count the number of bees going in and out of the hive over a given length of time. This is a huge burden because a greenhouse farm could be on average two and a half hectares or even larger. It’s difficult for laborers to check thirty hives and create a map. Computer vision and AI obviously greatly reduces the human error factor. We don’t see ourselves as replacing people, but as assisting them remotely and in real-time. Whether in another city or another country, commercial beekeepers and food producers could monitor their beehives as they pollinate food crops. Food producers need to plan their production and this relies on pollinators. We offer a tool to help them achieve this.”