Alexey Lebedev estimates the share
Posted: Mon Jan 20, 2025 8:00 am
Dmitry Svalov named the main obstacles to implementation as the complexity associated with the integration of AI with the current toolkit, the impossibility of transferring the entire context of the task to AI and the point nature of its application at the moment - as a result, a stable reduction in labor costs is possible by 5-7%, although in some cases the reduction can reach 30-50%, but such results are not yet guaranteed. In his opinion, the best results can be achieved when solving problems in software testing. But in general, as Dmitry Svalov emphasized, the potential for using AI is huge, confirming the well-known truth that new technologies are overestimated in the short term and underestimated in the long term.
of tasks that are entrusted south korea whatsapp resource to AI at 5-10%: "Here it is important to take into account the interpretation of the term "use". If we are talking about DevOps specialists who use AI tools, for example phind.com, as a technology consultant and debugging assistant, then the study's assessment coincides with what we observe in the market. About 40% of companies are indeed experimenting with the implementation of AI at different stages of the software life cycle. But at the same time, AI is not yet a basic technology for developers, and if by "use" we mean AI tools built into automated pipelines, the percentage of implementations is currently lower."
Sergey Lunegov warns that AI will not cope well with tasks that require a lot of creativity, as well as non-standard ones, about which little is written on the Internet.
Valentina Dmitrieva, Team Lead of the DevLab "KOMPAS-3D" "ASCON", named distrust as the main obstacle to the implementation of AI: "There is distrust of third-party APIs for language models due to security issues. This is one of the reasons why many companies have not yet implemented AI in DevOps processes. Thus, the implementation of LLM is carried out either by open-source projects or large companies that can deploy such models on on-prem infrastructure for the purposes of, for example, code review."
The effect, according to an ASCON representative, can be significant, although it is difficult to measure objectively: "At conferences, representatives of businesses implementing AI report an increase in development productivity by 5-30%. It is also noted that with the implementation of large language models, programmers write more unit tests, which, although it does not directly affect productivity, improves the quality of the software product."
According to the observations of Ilya Kulakov, Director of the Department of Advanced AI Projects at T1 Holding, in many companies the use of AI is at the experimental stage, and it would be incorrect to claim that AI processes cover 44% of DevOps tasks, although many companies use AI for DevOps tasks. According to him, the main effect of AI implementation is the reduction of the number of errors: "Successful integration of AI into DevOps requires qualified specialists in the team, infrastructure and the willingness to adapt to new technologies. To use AI, two conditions must be met: first, the task must be subject to "robotics", be formalized, routine and often repeated, and second, it is necessary to have a set of "clean" data for training the model. If the last condition is met, then at the current stage of development it can be solved by means of AI. It is also worth keeping in mind that AI reduces errors to a greater extent than labor costs, especially errors made by young specialists at the implementation stage and during subsequent use. AI also simplifies work with documents. The processes being developed are regulated by a large number of regulatory requirements that must be taken into account when developing documentation. Young specialists lack the knowledge and experience, and AI can help identify shortcomings and errors."
of tasks that are entrusted south korea whatsapp resource to AI at 5-10%: "Here it is important to take into account the interpretation of the term "use". If we are talking about DevOps specialists who use AI tools, for example phind.com, as a technology consultant and debugging assistant, then the study's assessment coincides with what we observe in the market. About 40% of companies are indeed experimenting with the implementation of AI at different stages of the software life cycle. But at the same time, AI is not yet a basic technology for developers, and if by "use" we mean AI tools built into automated pipelines, the percentage of implementations is currently lower."
Sergey Lunegov warns that AI will not cope well with tasks that require a lot of creativity, as well as non-standard ones, about which little is written on the Internet.
Valentina Dmitrieva, Team Lead of the DevLab "KOMPAS-3D" "ASCON", named distrust as the main obstacle to the implementation of AI: "There is distrust of third-party APIs for language models due to security issues. This is one of the reasons why many companies have not yet implemented AI in DevOps processes. Thus, the implementation of LLM is carried out either by open-source projects or large companies that can deploy such models on on-prem infrastructure for the purposes of, for example, code review."
The effect, according to an ASCON representative, can be significant, although it is difficult to measure objectively: "At conferences, representatives of businesses implementing AI report an increase in development productivity by 5-30%. It is also noted that with the implementation of large language models, programmers write more unit tests, which, although it does not directly affect productivity, improves the quality of the software product."
According to the observations of Ilya Kulakov, Director of the Department of Advanced AI Projects at T1 Holding, in many companies the use of AI is at the experimental stage, and it would be incorrect to claim that AI processes cover 44% of DevOps tasks, although many companies use AI for DevOps tasks. According to him, the main effect of AI implementation is the reduction of the number of errors: "Successful integration of AI into DevOps requires qualified specialists in the team, infrastructure and the willingness to adapt to new technologies. To use AI, two conditions must be met: first, the task must be subject to "robotics", be formalized, routine and often repeated, and second, it is necessary to have a set of "clean" data for training the model. If the last condition is met, then at the current stage of development it can be solved by means of AI. It is also worth keeping in mind that AI reduces errors to a greater extent than labor costs, especially errors made by young specialists at the implementation stage and during subsequent use. AI also simplifies work with documents. The processes being developed are regulated by a large number of regulatory requirements that must be taken into account when developing documentation. Young specialists lack the knowledge and experience, and AI can help identify shortcomings and errors."