AI in SDLC: the way from assistants to agents (AI column)
Written. extended versionwhich continues my previousIntegrate AI into development processes in a large company" Then I set a frame and talked about how to approach the implementation of AI in a large company. And this time I focused on moving to agency scripts and discussed the following topics:
Introduction and discussion of the topic of the report
- A history of assistants. (GitHub Copilot, Cursor) Moving to Agents and Why They Look Like a Revolution Examples of agents and tools (Claude Code by Anthropic, OpenAI Codex)
- Infrastructure and protocols (MCP, A2A)
- Economic prerequisites of agency hype Examples of successful cases from Google AlphaEvolve Future of work and agency levels in Stanford study Managing and Regulating Agents by Google Deepmind Evolution of SDLC to Agency Software Engineering 3.0 Demo agent regime inside T-Bank on the example of the game5 letter Development Ecosystem and Engineering Goals in Google’s Measuring Developer Goals Internal Development Platform and Spirit and our Approach to AI Development Agent mode and work with data inside python notebook Agency mode for quality assurance and creation of test cases
- Code review agent.
- An agent to look for vulnerabilities (safeliner) Measurement of effectiveness and evaluation framework Framework for assessing assistants and agents from the DX platform Results of using assistants/agents in T-Bank
Discussion and links to all sources for study are available in my tg channel
The podcast release is available in Youtube, VK Video.
P.S. Take a survey to participate in our review of the impact of AI on development.
#AI #PlatformEngineering #Engineering #Software #Processes #Productivity