Glossary
HAI Terms and Definitions
Complete reference of all terms used in the HAI Framework. The naming was chosen to require no legend — the terms are self-describing. Partnership language instead of numbered scales, culturally neutral across North America and Europe.
Who leads the work. Not a hierarchy — different work legitimately needs different modes.
AI-led
AI does the work, human ensures quality. Used for implementation of well-defined specs, routine tasks, and pattern-following work. The human’s role is specification and review.
Collaborative
Human and AI work together iteratively. Used for feature development, problem-solving, and design work. Both tracks are actively contributing throughout.
Human-led
Human decides, AI supports. Used for strategy, client relationships, novel decisions, and work requiring deep domain expertise. AI provides research and drafting support.
How much human attention and judgment the work needs.
Light
Quick review, routine approval. Standard patterns, minimal decisions needed. Minutes of human time per work item.
Engaged
Active participation, some decisions. Judgment calls required but in familiar territory. An hour or more of human time.
Intensive
Deep thinking, significant decisions. Multiple stakeholders, important choices, novel territory. Half a day or more of focused human attention.
Scope of the work. Maps to client-facing pricing tiers.
Small (S)
One component, one file area, straightforward. Hours of AI work. Typically included in subscription pricing.
Medium (M)
One feature area, multiple files, clear path. A session or two of work. Typically a fixed-price quote.
Large (L)
Multiple features or integrations, some unknowns. Multiple sessions. Fixed-price quote with clear scope document.
Extra Large (XL)
System-wide, architectural, many unknowns. Requires planning and parallel tracks. Scoped engagement with phased delivery.
How well-defined the work is. The most valuable assessment dimension — the strongest predictor of delivery success.
High
Clear acceptance criteria, testable conditions, defined scope. AI can execute with minimal iteration. Edge cases documented, context provided, non-scope items listed.
Medium
General direction clear, details emerging. Expect collaborative refinement. Approach is identified but specific implementation decisions remain.
Low
Vague intent, undefined scope. Research phase needed before specification can begin. For L/XL items, investing in spec quality before production is mandatory.
How predictable the outcome is and how well the scope is understood.
Risk: Low
Done this before, standard patterns, no external dependencies. Outcome is highly predictable.
Risk: Medium
Some unknowns, external dependencies, or new patterns. Outcome is mostly predictable with some potential for surprises.
Risk: High
New territory, multiple external dependencies, architectural decisions. Outcome has significant uncertainty.
Confidence: Low
Request phase, many unknowns. Normal at the start of work. Integration work with external APIs often starts here.
Confidence: Medium
Post-research, path is forming but details remain. Integration work caps at Medium until tested against live systems.
Confidence: High
Spec complete, implementation plan clear. The team has high certainty about scope, approach, and expected outcome.
HAI has two layers: universal methodology and implementation-specific details.
Core HAI
The methodology anyone can adopt: two tracks, work mode, involvement, spec quality, phases, retrospective, and social contract. Works with any tools, any team, any domain.
Extended HAI
Implementation-specific details: pricing figures, partner contribution tracking (Slicing Pie), environmental reference estimates, market lens, and specific tooling. Adopters should adapt these to their own context.
Other important terms used throughout the HAI Framework.
Specification
The human track: defining intent precisely enough for systems to execute. Includes direction, judgment, context, review, and accountability. Where value lives in a world where production cost approaches zero.
Production
The AI track: executing specification at speed and scale. Includes research, synthesis, code generation, iteration, and consistency. Requires specification to be valuable.
Retrospective
Post-completion review comparing initial assessment to actual outcome. The most valuable part of HAI — where specification skill compounds. Asks: was the spec good enough? What would we do differently?
Social Contract
The five commitments: use AI to work faster, humans specify and stand behind work, transparent about both contributions and costs, absorb AI costs, acknowledge environmental impact.
Assessment
A snapshot of work evaluated across all HAI dimensions (work mode, involvement, size, spec quality, risk, confidence). Updated at reassessment triggers, not constantly.
Phase
Work progresses through phases: Request, Research, Specification, Implementation, Review, Complete. Each phase has its own typical profile and confidence range.