CIO of St. Petersburg Yulia Smirnova - about neural networks in the civil service: how AI becomes a new competence of power

𝓲

Фото: St. Petersburg Informatization and Communications Committee

State apparatus and AI: not a revolution, but the evolution of thinking

Until recently, the introduction of AI into government structures seemed like a distant prospect or an ambitious experiment. Today, this is becoming the new norm. According to the Tenderplan procurement search system, in 2024 the volume of tenders for software and hardware complexes with AI accelerators in Russia increased by 150% in money and by 120% in pieces. Most of these purchases - 2.3 billion rubles - are accounted for by government agencies. And this is not an accident. The state is rapidly mastering AI not for the sake of hype, but as a tool to increase efficiency, transparency and proactive management.

When we say "AI in the civil service," it is important to understand: this is not about replacing an official, but about expanding his capabilities. AI becomes a new "employee" in the management team - an analyst, adviser, sometimes even a mediator between a citizen and the system. State AI systems can recognize patterns, predict loads, route requests, manage queues, run services without human input. But most importantly, they change the very logic of work: from reactive response to proactive help.

Where AI is really needed and where it is not yet

Artificial intelligence has no magic. Its strength is in speed, scale and ability to identify patterns where a person simply does not have time. The greatest effect of AI is given in zones of high repeatability, standardized processes and massive data. These are processing citizens' appeals, automatic routing of applications, analytics of large amounts of information, predicting demand for public services, resource allocation, logistics and planning. Where logic and statistics work, AI is indispensable: it reduces costs, speeds up processes and increases accuracy.

But there is also a downside. Human control is not just about algorithms. This is also a choice, sympathy, moral responsibility. There are areas where AI should not go deep yet. Ethics, politics, interpersonal conflicts, crisis support situations, working with vulnerable groups - all this requires empathy, intuition, the ability to hear between the lines. In such contexts, not only the result is important, but also the process: as it was said, with what tone, in what moment. Here, even the perfect algorithm can turn out to be cold and inappropriate.

AI is not a boss or a judge. He is an assistant. It is able to simplify the routine, take on millions of tasks of the same type, give analytics, predictions, even recommendations. But it's still up to the person to make a decision. Especially where we are talking about someone's fate, about values, about the choice between "correctly" and "according to the law, but soulless."

Competent AI architecture is not total automation, but a clear understanding of the boundaries: where the machine is stronger and where a person is needed. This is a culture of collaboration between the human mind and the power of the algorithm. This is the new maturity of a digital state in which technology amplifies humanism rather than displaces it.

𝓲

Фото: Rawpixel/Freepik

State as a neural network: philosophy of architecture

The transition to artificial intelligence is not just an infrastructure upgrade, not the next stage of digitalization, but a radical shift in the logic of management itself. The state building the AI system is changing not only the tools, but also its own form. We move away from a hierarchical, pyramidal model of power - with a vertex, circulars - to a network, distributed architecture, inherently close to the device of the neural network.

AI infrastructure is not only about hardware, clouds and computing. This is about a different model of statehood. It acts not directive, from top to bottom, but through interfaces - digital contact points between the state and the citizen, between the system and the person. In this model, the power itself becomes "transparent" - it works not because it was "ordered," but because the data suggested: this is where the problem is, this is where the resource is, this is where the narrow neck that interferes with people.

A neural network state is a system where a decision is not made in isolation, but is born as a result of the analysis of many interconnected signals: from economic indicators to sentiments on social networks, from weather to population migration. This is not an authoritarian artificial mind, but a distributed intelligence in which each manager becomes not an administrator and executor of regulations, but an architect of flows - data, tasks, assistance, resources. His task is not to "command," but to direct, balance, calibrate.

In such a system, errors are not accumulated, but neutralized in real time. Inefficiency becomes the exception, not the rule. The main thing is to build a transparent, flexible and learning system, where the state does not resemble a machine, but a living organism with well-developed reflexes. AI in this sense is not the goal, but the nervous system of this new statehood.

Petersburg experience: "Victoria" and other AI solutions in action

In St. Petersburg, AI is no longer a theory. One notable example is the Victoria voice assistant integrated into Service 122. She accepts requests for a doctor's call, informs about the status of applications, records for medical examination. During peak periods, Victoria is capable of handling up to 4,000 calls at a time - a load that no live team can cope with. This is not just a service, but an element of a new architecture of interaction with citizens: fast, accurate and humanized.

Intelligent video analytics integrated into the city's video surveillance system has become an invisible but powerful aid in ensuring order and comfort for residents. Modern algorithms work around the clock, automatically tracking what is happening on the streets, in public spaces, at social facilities. Thanks to video analytics, cameras do not just capture the image - they analyze the situation in real time: they identify clusters of people, recognize potentially dangerous incidents, record violations of the rules of improvement and traffic, and control the state of the urban environment. If a non-standard situation is detected, the system instantly sends a signal to operators, allowing services to respond faster and more accurately. This technology helps prevent accidents in the early stages, increase the speed of troubleshooting infrastructure problems, and most importantly, create a sense of security and confidence for residents that the city really cares about everyone.

The increase in the number of tenders for AI accelerators is a signal not only of technological demand, but also of sovereignty. The state is committed to keeping critical computing power in its hands. That is why St. Petersburg has become one of the first regions where a large-scale project for the purchase and localization of such systems has been launched. This is an investment in sustainability, independence and control over key algorithms, including decision-making systems.

St. Petersburg is not just a city where equipment and digital solutions are massively purchased. This is a city that thinks strategically and acts systematically. We are building a full-fledged digital transformation platform, in the center of which is a person, not technology. Artificial intelligence becomes part of the urban ecosystem, weaving into its key areas: from transport to healthcare, from education to housing and communal services.

AI helps to analyze traffic flows and reduce traffic jams, predict the needs of hospitals and optimize queues in clinics, adapt curricula to the needs of schoolchildren, predict accidents on engineering networks and respond to them in a timely manner. And all this is not for the sake of the effect of automation as an end in itself, but for the sake of improving the quality of life of every resident of St. Petersburg.

The neural network is not the lord, but a partner

AI in the civil service is not fiction, but a new reality. And it requires maturity, architectural thinking and responsibility from us. The official of the future is not the overseer, but the curator of digital ecosystems. Its tools are not only regulations, but also algorithms. His task is not to control, but to create conditions. And if we learn to use AI as a partner, and not as a threat, it is not power that wins, but every citizen. We are not replacing a person - we are strengthening his capabilities. We give the official more accuracy in making decisions, the doctor - more time per patient, the teacher - more tools for individual approach. AI becomes not a taskmaster, but a partner. And this is our conscious choice.