Slide contents

1. 1. AI in SDLC: Cursor isn't enough

   From a point assistant to platform context and measurable use cases

2. 2. Alexander Polomodov

   Technical Director & Fellow, large fintech

   Fintech architecture and engineering practices

   AI in SDLC at scale

   Code of Leadership · @book_cube

3. 3. From assistant to platform

   Why point tools fall short

   How big tech embeds AI

   What already works inside Spirit

   Where build-or-buy splits

4. 4. 01. A tool ≠ a system

   Local acceleration does not redesign enterprise delivery

5. 5. Assistants change coding, not the system

   Autocomplete became a baseline expectation

   Vibe coding accelerates prototypes

   Production still needs engineering constraints

   Company context remains outside

6. 6. Protocols expose context

   MCP

   Models call data and tools

   One interface replaces N integrations

   A2A

   Agents coordinate tasks

   Async and distributed workflows

7. 7. Platforms turn tools into developer jobs

8. 8. Goals come before AI

   Goal — What the engineer is completing

   Hindrance — Where flow and feedback break

   AI use case — Which action can be delegated

9. 9. 02. Five cases — one pattern

   AI amplifies a formal process and accumulated context

10. 10. Formal rules make AI useful

11. 11. The reviewer needs a second model

12. 12. An LLM alone was not enough

13. 13. Metrics mature with adoption

14. 14. On-call agents need mature context

15. 15. 03. AI inside Spirit

    Company context, professional use cases, and quality evaluation

16. 16. AI embeds into the platform

17. 17. AI is already a platform bet

    Respondent shares in State of Platform Engineering in the Age of AI

    45% — Core strategy — Central component

    34% — Important — Important, not central

    83% — Already using — GenAI in dev stack

18. 18. 04. Scale changes the answer

    Buy, build, or connect public models to company context

19. 19. Company size changes the answer

20. 20. Cursor is only the first step

    Rules and context precede models

    Platforms connect AI to real work

    Use-case metrics test the effect

    Build-or-buy depends on scale

    AI amplifies the engineering system already in place

21. 21. What was verified

    Source PDF · R2 SHA-256

    Talk recording · auto-captions

    Google · ByteDance · Uber

    Booking · Datadog · Red Hat

22. 22. Thank you!

    polomodov.tech

    All slides and links are in the Telegram channel

    Alexander Polomodov, Technical Director & Fellow, large fintech

    @book_cube