[1/2] AI in the SDLC: Research by IT One and Skolkovo (Category AI)
I recently attended the presentation of research findings by IT One and Skolkovo, where I took part in a panel discussion. Afterwards, I read all 59 pages of the report. Its structure is roughly as follows:
- Global practice: current and coming changes to the SDLC under AI. The authors reviewed reports by other respected organizations; I link to those below when discussing the findings.
- AI in Russia’s SDLC: the market, services, practices, and expectations. The authors examined publicly available services in Russia, plus major announcements even when the products were not yet widely available.
- Technology leaders’ views on AI in the SDLC: interviews with CTOs, CIOs, and development leaders at large companies. They interviewed 50+ respected industry figures. I could not find the list of interviewees in the report, so we will have to take their word for it.
- Forecasts, conclusions, and recommendations on the opportunities and risks of AI in the SDLC: the authors’ views on how AI will develop and how to integrate it into software development.
Let us start with global practice:
- The AI development-tools market: $6.9 billion → $29.6 billion by 2032, or 4× growth. The biggest effects are in development and testing. Source: Spherical Insights.
- Some 62% of developers already use AI and another 13.8% plan to, according to 2024 data. Managers estimate lower adoption, but the trend is accelerating. Source: StackOverflow 2024. I reviewed the 2024 report and also covered the newer 2025 report.
- Value is shifting from individual efficiency, or writing code faster, toward team productivity: a 10% coding-speed improvement now, and potentially 25–30% higher team productivity in a few years through “AI at team level,” whatever that means. Source: Mia Platform.
- SDLC acceleration: already 15–20%, according to Forrester, with 30–50% potential over the medium term, according to a Medium article by Sathish Rama, Director Gen AI at PayPal.
- Process transformation is expected over 1–3 years; 32% of technology leaders say the actual effect has already exceeded expectations. The cited source is MIT, but no specific article is linked.
The authors also mention the METR study, in which AI could slow experienced engineers down on complex tasks and repositories. I discussed it in posts 1 and 2, and even on the RIMS podcast, arguing that the study was designed for a particular result and represents a very narrow case.
Interestingly, they found a maturity model from the UK government’s DEFRA: “The AI-Powered SDLC: A Comprehensive Technical and Cultural Maturity Assessment Framework.” At first I could not see what a department responsible for environmental protection, food policy, and rural development had to do with digital transformation. Then I looked into it and understood :) Since the mid-2010s, DEFRA has become a leader in digital transformation in the British public sector, helped by active collaboration with the Government Digital Service (GDS) and the Central Digital and Data Office (CDDO). It has thousands of data sources, from satellites to farmers’ reports, and a very diverse user base. That makes it a place where the British government actively tests AI/ML applications and agent systems.
The AI risks highlighted in Western reports include:
- Code and data leaks, and compliance problems.
- Erosion of engineers’ skills and blind trust in suggestions.
- Growing technical debt and vendor lock-in.
AI in software development is clearly hotter than ever. In the West, seemingly everyone is publishing reports, forecasts, products, and more, mostly with a positive outlook. But many point out that the big efficiency gain will come from moving beyond individual copilots and integrating AI into company processes. Next, we will look at Russia.
#Software #Engineering #Productivity #DevEx #AI #Management #RnD #Leadership #Economy