The figures presented at the forum are based on data from the Design 1.0 digest by DOM.RF, the housing development institution. Neural networks speed up the preparation of design documentation fourfold, cut its processing time by 50% and improve the accuracy of work volume calculations by 45%. Together, these effects produce savings of about 1.8 million roubles per project.
Dmitry Livshin, CEO of Cyber Business Consulting, considers the figure a benchmark rather than a ceiling: "The potential effect may be much higher than 1.8 million roubles per project. Such an amount looks possible, but I would not consider it an upper limit, since it depends on the scale of the facility, the volume of repetitive operations and which specific part of the work can be improved. For a small organisation, the costs of implementation may be high compared with the savings. In a large project, even a single task well chosen for AI can produce a much more significant result."
One of the industry's systemic problems is disordered source data. Many engineering decisions are made based on specialists' personal experience rather than recorded regulations. This makes it possible to automate a single operation but blocks the transfer of more complex processes to an algorithm. Olga Kutuzova, head of innovative systems development at Nanosoft, pointed to the need for preliminary work: "The construction industry is archaic. That is why one of the difficulties of implementing AI is the lack of ordered source data. Often, when work needs to be done quickly, companies do not seek to introduce innovations and see them as an extra burden. That is why it is important to decompose projects and choose specific tasks for automation."
Scaling pilot AI solutions to the entire organisation involves additional infrastructure costs. Igor Rogachev, CEO of StroIIka, noted that construction data are closely interconnected, so working with them requires additional software solutions: pilot test results are not always preserved in full deployment, and assessing the economic benefit becomes more difficult.
Another limitation is the impossibility of fully delegating design to an algorithm. Nikolay Flyagin, head of the technical expertise group at the CAD and GIS Solutions Department of Softline Solutions (Softline Group), gave a telling example: "AI proposes the optimal route for a pipeline, but the engineer notices that it passes through an area of permafrost. If local geological features are not taken into account, the pipe will have to be relaid in a few years. That is why design cannot be fully handed over to algorithms: they can propose options and perform calculations, but a specialist must evaluate the result and make the final decision."
AI tools are built into all modern BIM solutions and are not singled out as a separate class of products. Livshin formulated the practical logic of implementation: the company gets the opportunity to take on more profitable orders without a proportional increase in costs — and this, rather than one-off savings, is the long-term economic meaning of the technology. Kirill Semion, CEO of the NCC ISU, added that AI creates particular value when designing complex integrated solutions: the system makes it possible to see at once how changes in one node affect the entire project.
Related news:
- AI traffic grew 4 times at Ozon in a year — how AI traffic is changing marketplace sales
- Driverless trucks may enter the Russia — Belarus route — driverless trucks on the Russia — Belarus route
- Yandex agentic commerce: how a store should prepare its infrastructure — how a store can prepare for agentic commerce
- How neural networks are changing international logistics and foreign trade — where neural networks already work in logistics and foreign trade
