AI in SDLC Report by IT One and Skolkovo

Alexander Polomodov and Dmitry Nemov review the Russian AI in SDLC report by IT ONE and Skolkovo. The team paired international research with 25 interviews, roughly 50 hours with technology leaders. This reveals practice behind public claims, although the market can outdate a tool inventory before publication.

Map the SDLC through engineering jobs rather than products: understand a system, design a change, write code and tests, verify, and deploy. Companies need safe model access, experimentation, and practice sharing. Year-over-year evidence shows the shift from completion to code generation and agents.

Leaders converge on a centralized platform that supports model switching and access control. Maturity varies: tests, documentation, and review already accelerate, while deployment remains closed and conservative. AI may recombine roles, but it does not remove the need for codebase context, architecture, and engineering gates.

Most companies measure impact poorly. A practical ladder moves from tool use to job outcome, then impact and cost; demanding ROI earlier is meaningless. Shadow AI proves demand but adds risk. Scale business-valued cases using deterministic checks, evals, constraints, and feedback loops.