AI Applications to Transform Oversight and Supervisory Activities in Agribusiness

In 2022, Russia produced more than 5 million tonnes of poultry meat, 4.5 million tonnes of pork and 1.6 million tonnes of beef. In addition, almost 5 million tonnes of aquatic bioresources were produced, and milk production reached 32 million tonnes. The past year was also marked by a record grain harvest of more than 150 million tonnes. What’s more, 1.5 million tonnes of fish, 650,000 tonnes of meat, and 45.5 million tonnes of grain were exported. Each year, Russian products are bought by 130–160 countries worldwide. Given this, digitalization has become essential in a global market to support national biological security, guarantee the safety of primary produce and processed goods, and ensure that goods comply with requirements through oversight, supervision and licensing. Since 2006, 10 information systems have been developed and launched by the Federal Service for Veterinary and Phytosanitary Surveillance (Rosselkhoznadzor). Thanks to these programs, supervision can occur without having to interact with businesses across all basic areas of the agricultural industry (animal husbandry, industrial processing, crop farming, veterinary medicine, and agrochemistry). Today, it has become possible to automate the identification of unscrupulous entrepreneurs and consumer fraud schemes through the use of neural networks. One AI model recognized by the OECD as an outstanding global innovation is capable of analysing 46 million documents per hour. This translates into a 90,000-fold increase in inspectors’ productivity. Digital supervision allows the state to quickly identify violations, and helps honest entrepreneurs work in a competitive environment and openly maintain their reputation. It is becoming clear that there needs to be a legislative definition of digital public administration. This should encompass supervision and oversight, and should enshrine the procedure for employing artificial intelligence when detecting violations. Is it possible for inspectors to make decisions based on information provided by AI models? What interaction tools in digital public administration should facilitate prompt action on incidents detected by systems that go beyond the purview of the agency that identified it? Is the government in a position to delegate procedural steps to neural networks?