State of AI4SDLC
Research + AI-native delivery
Slide contents
1. State of AI4SDLC
Research + AI-native delivery
2. Talk Route
Research, delivery, rollout
Part 1: AI4SDLC Research 2025.
Part 2: delivery loop.
Parts 3-4: organization, task management.
Part 5: rollout and metrics.
3. 01. AI4SDLC Research 2025
How AI affects software development — meta-analysis and engineer survey
4. How We Looked at AI4SDLC
Practices, culture, trust
Where AI is already daily.
Where adoption remains rare.
How speed, quality, trust connect.
What turns personal gains into team impact.
5. Meta-Study: Data Sources
50+ studies, four source types
Large-scale surveys of developers and tech leaders.
Telemetry and product/engineering logs.
Field experiments and controlled task studies.
Internal corporate research and interviews.
6. AI4SDLC (2025)
Survey of engineers and tech leads
Self-assessed productivity, quality, trust impact.
AI usage frequency by SDLC task type.
Variable sample size across questions.
Descriptive analysis without causal conclusions.
7. Study Limitations
How to read the results
Sample reflects study context.
Process changes can affect productivity.
Industry metrics remain inconsistent.
Novelty and adaptation skew trends.
8. Meta-study patterns
Recurring 2023–2025 research patterns
80%+ growth · 10h saved — Default, but uneven — Daily tool; gains depend on task context.
-1–7% stability/throughput — Coding outpaces delivery — Bottlenecks: review, tests, integration.
verify-first + platform — ROI through process — Trust depends on SDLC maturity.
9. Engineer survey signals
AI4SDLC Survey 2025 signals
58% often/always — Everyday AI tool — Code generation is daily practice.
64% vs 32% — Local effect dominates — Productivity outpaces quality and team throughput.
49% / 60% — Trust and expectations diverge — Low trust; business expects impact.
10. AI Is Daily, but Not Everywhere
High-risk tasks still need humans
Mass scenarios
58% often use generation.
Rare scenarios
24% code review; 18% legacy.
11. Personal Speed Is Not Delivery
Coding speed ≠ value
64% report productivity growth.
Team effect needs mature process.
DORA 2024: AI adoption reduced throughput.
DORA 2025: throughput up; instability stayed.
12. Trust Requires Systemic Validation
Verify, don't trust as engineering norm
Quality
32% quality better; 14% worse.
Trust
49% don't trust AI code.
Verification culture enables impact.
13. Bottleneck Moved to Delivery
Main takeaway for team leads
Commit → prod is still long.
Review, tests, releases limit flow.
Documentation becomes context quality.
Winners rebuild the work system.
14. 02. Classic PDLC → AI-Native
SE 1.0 → SE 2.0: system rebuild
15. SE 1.0 vs SE 2.0
Role-based vs agent-based SDLC
SE 1.0 · Role-based SDLC
Roles pass work through SDLC.
Brownfield keeps control and handoffs.
SE 2.0 · Agent-based SDLC
Engineer + agents compress the loop.
Greenfield is the experiment zone.
16. The Team in SE 2.0
Team lead owns context and guardrails
Role-based team
Streams run through role-based SDLC.
Brownfield delivery remains core.
Agent-based team
Team lead owns context, queue, approvals.
Agents get bounded autonomy.
17. Signals from Market Leaders
Google, Meta, AWS, DORA
Google: AI inside engineering loop.
Meta: JIT tests for diffs.
AWS: roles compress with guardrails.
DORA: payoff depends on platform.
18. SE 2.0 Needs Discipline
Coding speedup can create rework
Weak control → rework.
Noisy tests → instability.
Agents outrun review.
Foundation: branch protection, CI, signing.
19. SE 1.0 ↔ SE 2.0
Bidirectional practice transfer
SE 2.0 → frontier — Agentic orchestration, JIT tests, auto-docs. Greenfield and R&D.
SE 1.0 → foundation — Checkpoints, developer platform, safe delivery.
Practice transfer — 1.0 gives guardrails; 2.0 returns automation.
20. 03. AI-Native Organization
SDLC and company model both change
21. Why the Next Step Is AI-Native Organization
Faster code; slower decisions and handoffs
Layers compress — After SDLC roles, organizational layers shrink too.
Decision latency — Approvals and handoffs become bottlenecks.
Leverage over headcount — Focus shifts to output per employee and platform.
22. Org Signals at Market Leaders
Wider span, fewer levels, stronger platform
Wider span
Meta: AI org ratio reaches 1:50.
Nvidia: ~50 CEO reports.
Fewer levels
Amazon: +15% IC/manager ratio.
Microsoft, Google: AI-first tooling.
23. Flattening Needs Platform Support
Without platform, chaos follows
Need golden paths, feedback, policy-as-code.
Otherwise managers absorb bottlenecks.
Ratios 1:50 are not templates.
Managers design human-agent loops.
24. 04. AI-native task management
Atlassian and Linear show: not just coding changes, but the work unit itself
25. Task management: the next bottleneck
Tickets are no longer manual forms
Before the ticket — Incidents, chats, and docs become task sources.
Triage doesn't scale — Fast flow turns backlog into queue.
Status is computed — Task stores links, history, next steps.
26. Ticket tracker to contextual work graph
The work unit itself changes
Classic task management
A person creates the task manually.
Context moves from chats and docs.
AI-native task management
System creates work item from signal.
AI suggests team, assignee, labels.
27. Signals from Atlassian and Linear
Intake → enrichment → delegation
Atlassian
Jira: AI-assisted work items and summaries.
Rovo/agents add context and handoffs.
Linear
Triage Intelligence suggests owner and labels.
AI agents work; human owns outcome.
28. DevOps/platform: What to Do
Task management as delivery platform
Connect incident, support, docs.
Automate intake; human-owned approval.
Trace intent → PR → deploy.
Measure signal-to-action lead time.
29. 05. AI-Native Rollout in 2026
New operating model, not IDE pilots
30. AI as Operating Model Layer
Inside the standard delivery loop
AI enters the workflow — PR, review, tests, CI/CD = AI surfaces.
Agents get bounded tasks — Bounded tasks under human control.
Management language changes — Throughput, risk, economics.
31. Big Tech Moves AI-Native to System
Workflow embedding + measurement platform
Workflow embedding
Microsoft: AI reviewer inside PR flow.
GitHub: coding agent opens draft PR.
Platform, evals, economics
AWS/Amazon: AI linked to e2e delivery.
Stripe: maturity measured through evals.
32. How Big Tech Measures the Shift
Adoption → throughput → quality/risk → economics
usage, IDE, CLI, agents — Adoption + workflow — Who entered engineering.
bugs, tests, incidents — Quality / Risk — Defects, tests, incidents.
cycle time, CTS-SW, ROI — Throughput + economics — Did flow improve? Did AI pay?
33. Takeaways for Teams
Managed AI-native delivery system
Embed AI into regular workflows.
Give agents bounded autonomy.
Keep approvals and accountability human-owned.
Measure delivery: throughput, risk, economics.
34. Key Theses
What to take with you
AI accelerates artifacts; impact needs mature verification.
Bottlenecks shift to task management and handoffs.
Manage AI-native via delivery outcomes.
ai4sdlc-research.space · tellmeabout.tech
35. Materials
Methodology and sources
AI4SDLC Research 2025
DORA AI / State of DevOps
GitHub, Microsoft, Google, AWS
Stack Overflow, JetBrains, Atlassian
36. How Was the Talk?
DevOps Conf
Please share your feedback — it helps us get better
Alexander Polomodov, Technical Director & Fellow, T-Technologies
37. Thank You!
DevOps Conf
If you want to explore the materials mentioned in the talk, head to the "Book Cube" channel — all links and resources are collected there
Alexander Polomodov, Technical Director & Fellow, T-Technologies
@Book_Cube