Step 1
Open the common playbook
Start with the shared foundation used across the whole AI in Engineering process. This covers the rules that do not belong to a single specialization.
How to use the playbooks
The model is intentionally simple. There is one layer of guidance that applies everywhere, and then there are playbooks for concrete engineering contexts. This keeps shared rules in one place and avoids repeating tooling, pricing, and operating principles in every specialist document.
Step 2
Choose the stage or lane
Move to the part of the SDLC you are working on, or to the Data lane when the work is data-specific. Each area has its own playbook and local recommendations.
Step 3
Follow specialist guidance
Use the concrete playbook for your role, stage, and deliverable. That is where local standards, patterns, and practical recommendations live.
Playbook Flow
0
Foundation
AI Usage Playbook — Common
Shared guidance for all stages: tooling, model choice, pricing and session economics, privacy, security baselines, and rules for working with AI across engineering.
1
Specialization Layer
Stage-specific playbooks
Once the common layer is clear, move into the playbook for the stage you are in: discovery, architecture, development, quality, platform, or data.
Common playbook as the shared foundation
Why it exists
What belongs in common
- Tooling choices and the AI assistant landscape
- Model selection logic and price-to-quality trade-offs
- Pricing, session economics, and prompt caching
- Privacy, security, and organization-wide usage rules
- General working model for using AI responsibly across teams
Purpose
Keep cross-cutting guidance in one place instead of copying it into every specialist playbook.
Audience
Anyone working with AI across the engineering lifecycle, regardless of stage or specialization.
Outcome
Teams start from the same operating model before entering local recommendations for their area.
Engineering stages and their playbooks
Below is the intended map of the AI in Engineering process. Each stage gets its own playbook with recommendations tailored to that kind of work.
1
Planned
Discovery and requirements
How to use AI for framing the problem, clarifying scope, preparing requirements, and documenting assumptions.
2
Planned
Architecture and design
Patterns for architecture review, trade-off analysis, ADR creation, and design validation with AI support.
3
Available
Development
The specialist playbook for implementation work: planning with the agent, coding, review, verification, and safe delivery.
4
Planned
Quality and testing
How to use AI for test strategy, meaningful test generation, defect prevention, and review of edge cases.
5
Planned
Platform, DevOps, and release
Guidance for CI/CD, infrastructure, platform operations, and production-safe automation with AI assistance.
6
Planned
Operate and improve
Using AI in maintenance, support, incident response, optimization, technical debt reduction, and knowledge capture.
Dedicated Data lane
Parallel specializationData is not only one step inside the SDLC. It often behaves like its own lane with different constraints, workflows, and evaluation criteria, so it should have a separate playbook next to the engineering stages.
D
Planned
Data playbook
A dedicated guide for data engineering, analytics, experimentation, data quality, and AI usage patterns that are specific to data-intensive work.
- Data preparation, validation, and lineage
- Experimentation and reproducibility
- Evaluation standards for data and AI workflows
- Governance, privacy, and controlled access to sensitive data