Slide contents

1. 1. From Classic PDLC to AI-Native Development

   How the engineering process changes when AI becomes part of the production loop

2. 2. What This Talk Is About

   Transition strategy

   Part 1: Bimodal IT failure.

   Part 2: SE 1.0 vs 2.0.

   Part 3: DORA, Google, Meta, AWS.

   Parts 4–5: limits and practices.

3. 3. 01. Historical Analogy: Bimodal IT

4. 4. Bimodal IT: Two Gartner Modes

   Fast vs slow IT

   Mode 1: stable systems.

   Mode 2: experimental speed.

   McKinsey: Two Speed IT.

   Diagnosis became operating model.

5. 5. Where Bimodal IT Broke in Practice

   'Fast/slow' is too coarse.

   Fast loop still hits the core.

   DORA: speed and stability grow together.

   Useful diagnosis; weak operating model.

6. 6. 02. SE 1.0 ↔ SE 2.0

7. 7. Software Engineering 1.0

   Roles, stages, controls

   Requirements → code → tests → release

   Strength: artifact controls

   Weakness: context transfer, long cycle

   NIST SDLC — classic framing

8. 8. Software Engineering 2.0

   Work shifts to AI tools

   Issue → agent → PR → CI/CD.

   Copilot agent, Amazon Q, AutoGen.

   QA, security, analysis remain.

   Control: PR, CI, branch protection.

9. 9. SE 1.0 vs SE 2.0

   Two modes with bilateral exchange

   SE 1.0

   Roles and stages are separate

   Mature control points

   SE 2.0

   Roles compress into agent loop

   PRs in minutes, new quality loops

10. 10. 03. Industry Signals 2024–2026

11. 11. DORA: Local Productivity ≠ Systemic Acceleration

    DORA 2024: AI adoption reduced throughput.

    Reason: AI sped up IDE, not delivery.

    DORA 2025: correlation turned positive.

    ROI comes from platform quality.

12. 12. Google: AI as an Engineering Product

    Source: Google

    AI spans inner and outer loops.

    Features prove value in real flow.

    Funnel: latency, relevance, UX.

    Productivity is designed and measured.

13. 13. Meta: Just-in-Time Tests

    Source: arxiv.org/abs/2601.22832

    22,126 generated tests.

    JIT tests target a specific diff.

    Less manual review and flaky-test noise.

    Solutions return to SE 1.0 as standards.

14. 14. AWS: AI-Driven Development Life Cycle

    Source: AWS DevOps Blog

    AI generates requirements, code, tests.

    Roles compress, not disappear.

    Faster agentic loop needs guardrails.

    AWS outages exposed AI-tool risk.

15. 15. 04. SE 2.0: No Free Acceleration

16. 16. SE 2.0 ≠ Free Acceleration

    Local flow ≠ delivery

    More changes → rework.

    Faster code → instability.

    Fast PR → bottleneck.

    Local loop ≠ business gain.

17. 17. Load-Bearing Elements of SE 2.0

    Old guardrails become foundation

    Branch protection and PR review

    CI policies and policy-as-code

    Audit trail for changes

    Sandboxing, signing, SSDF

18. 18. 05. Strategy for a Large Company

19. 19. SE 1.0 ↔ SE 2.0 Exchange

    Thoughtful exchange, not victory

    SE 1.0 → SE 2.0

    Guardrails and security policies

    Version control and audit trail

    SE 2.0 → SE 1.0

    E2E delivery scenarios

    JIT tests and short feedback loops

20. 20. Three Transition Principles

    SE 2.0 — greenfield, R&D, orchestration.

    SE 1.0 — production backbone: platform, reliability.

    Success is measured by DORA, SPACE, DevEx.

21. 21. Key Takeaways

    What to take away

    AI changes speed and work organization.

    SE 2.0 is a capabilities factory.

    AI-native rebuilds creation, verification, delivery.

    A unified engineering approach is required.

    tellmeabout.tech

22. 22. References and Materials

    Key sources

    Gartner, McKinsey, NIST SDLC.

    DORA 2024/2025, AI Capabilities.

    Google AI, Meta JIT, AWS DLC.

    Copilot, Amazon Q, AutoGen, SSDF.

23. 23. Thank You!

    From PDLC to AI-Native

    For more materials on this topic, visit the "Book Cube" channel — all links are collected there

    Alexander Polomodov, Technical Director & Fellow, T-Technologies

    @Book_Cube