The people this book is for are not waiting for the field to stabilize. They are the interdisciplinary thinkers: the ones who see connections across domains, who understand flows and algorithms at an intuitive level, who can think in terms of design and systems but may never have written production code at scale. The shift this book supports is from executing tasks to designing systems, from producing outputs to producing evidence, from controlling every step to knowing which steps require your judgment. That shift is what makes AI useful in high-stakes, accountable work -- not speed, not automation for its own sake, but comprehension and accountability at the point where it matters.
The tenets and principles that follow operationalize that shift. They are not a checklist. They are a frame for thinking about design decisions when the right answer is not obvious. Use them as a lens, not a recipe.
Compressed Tenets¶
Strategic alignment — what matters.
Trust defines the mission
Data precedes models
Deliver capability quickly
Governance must enable execution
Humans remain accountable
Defensibility is required
Adapt continuously
Build internal capability
Experiment under control
Infrastructure enables delivery
Working Principles¶
Practitioner rules — how to act.
If it cannot be explained, do not use it
Slow governance prevents delivery
Adoption determines impact
Build lineage before scale
Reuse data before collecting more
Access defines capability
Security is a system property
Outputs require full provenance
Chapter-to-Principle Mapping¶
| Chapter | Primary Working Principle(s) | Primary Tenet(s) |
|---|---|---|
| Ch 1: Why Design Matters | All (introduction) | All (introduction) |
| Ch 2: Classification & Coding | Adoption determines impact | Deliver capability quickly |
| Ch 3: Data Wrangling & Standardization | Reuse data before collecting more | Data precedes models |
| Ch 4: Detection & Extraction | If it cannot be explained, do not use it | Humans remain accountable |
| Ch 5: Ensemble & Multi-Model | Outputs require full provenance | Defensibility is required |
| Ch 6: Parallel, Serial, Bottleneck | Adoption determines impact | Deliver capability quickly |
| Ch 7: Checkpoints & Recovery | Build lineage before scale | Experiment under control |
| Ch 8: Evaluation by Design | If it cannot be explained, do not use it | Defensibility is required |
| Ch 9: State, Drift, Validity | Outputs require full provenance | Trust defines the mission |
| Ch 10: State Mgmt & Provenance | Build lineage before scale | Defensibility is required |
| Ch 11: Workflow Orchestration | Access defines capability | Infrastructure enables delivery |
| Ch 12: Security & Supply Chain | Security is a system property | Trust defines the mission |
| Ch 13: Institutional Deployment | Slow governance prevents delivery | Governance must enable execution |
| Ch 14: Cost & Practicality | Slow governance prevents delivery | Infrastructure enables delivery |