The AI spec layer: the Rikki Operating Model

Rikki compresses your AI adoption timeline. One adaptive model covering people, process, and organizational memory — calibrated to where you are today, ready to grow with you.

The Rikki cycle: Envision, Create, Deliver, Evolve

The Orchestration & Governance cycle is circular and bidirectional — not a linear pipeline. Any stage can flow to any other. Envision can loop back from Create when specs need refining. Deliver can jump to Envision when production signals spark new ideas. Evolve feeds everything. Rikki sits at the center, orchestrating and governing across all stages simultaneously.

Think of how a real project works: you envision, start creating, realize something needs re-envisioning, adjust, create again, deliver a piece, get feedback, evolve your understanding, and envision the next iteration. That's the Rikki cycle — not a waterfall with new names.

Rikki Spec · Govern Trace · Learn Envision Context → intent Create Specs → software Deliver Software → production Evolve Outcomes → memory 🔵 Intent gate 🟡 Review gate 🔴 Ship gate

Solid arrows: primary flows. Dashed arrows: cross-phase jumps. Rikki connects to every stage — orchestrating, governing, and learning continuously.

Four operating models

Organizations adopt AI at different levels. Each model defines how people, process, and organizational memory work — and how the Rikki cycle runs within it. Rikki adapts to whichever model fits your organization today.

Model 1

AI-assisted

Individual AI tool adoption. No organizational standard. Traditional roles intact.

People

  • Traditional roles: PM, dev, QA, ops
  • Each person selects their own AI tools
  • Knowledge concentrated in individuals

Process

  • Existing methodology (Agile, Waterfall, or ad hoc)
  • Manual artifact creation
  • Governance applied inconsistently or post-hoc

Org memory

  • Fragmented across tools and individuals
  • Static artifacts that decay immediately
  • No decision-to-implementation traceability
EnvisionPM pastes transcript. Rikki generates PRD with Customer, Target State, Acceptance Criteria, and MoSCoW priorities. 🔵 Intent Gate: human confirms before build specs are written.
CreateRikki generates SPEC, Validation Report, and Plan. 🟡 Review Gate: human confirms tech stack and locks the SPEC as single source of truth. AI agents build from the Plan.
DeliverHuman + agents work through Validation Criteria. 🔴 Ship Gate: human confirms all ACs and TAs pass before production. Monitoring and rollback plan in place.
EvolveHuman authors Evolve Note: did it land as intended? Learnings feed back to Rikki. Follow-on conversations trigger the next Envision cycle.
Rikki tier: Rikki Skill (free). Declare your stack, paste a transcript → PRD with Acceptance Criteria, Build Spec, Validation Report, and Plan. Three human gates keep judgment in the loop.
Model 2

AI-integrated

Team-wide AI adoption. Roles shifting. Structured specs replace informal requirements.

People

  • Role boundaries blur — broader scope per person
  • New competencies: AI-augmented engineers, integration leads
  • T-shaped generalists outperform narrow specialists

Process

  • Structured specs consumed directly by AI agents
  • Governance embedded in workflow
  • Spec quality prioritized over code volume

Org memory

  • Conversations, docs, and code unified in one engine
  • New members onboard from accumulated context
  • Artifacts linked to source material — stay current
EnvisionTeam ingests conversations and enterprise data. PRDs grounded in existing codebase context. 🔵 Intent Gate confirms customer, scope, and Acceptance Criteria before specs are generated.
CreateAI agents build from locked SPECs and Plans. 🟡 Review Gate: human confirms tech stack. Validation Report surfaces blockers before code is written.
DeliverValidation Criteria checklist run by human. 🔴 Ship Gate confirms all Acceptance Criteria and Testable Assertions pass. Full traceability on release.
EvolveEvolve Notes capture what landed vs. Target State. Shared context engine grows. Each cycle compounds memory and sharpens the next PRD.
Rikki tier: Rikki Skill + Rikki Platform. Enterprise data ingestion. Searchable shared context. Spec-to-code alignment tracking across sessions.
Model 3

AI-native

AI agents execute. Humans orchestrate, validate, and make judgment calls.

People

  • Orchestrators replace traditional hierarchies
  • Each person leveraged 3–5x through agents
  • Core value: judgment, taste, customer empathy

Process

  • Continuous flow replaces fixed cycles
  • Agents build; Rikki verifies against specs
  • Humans intervene at judgment gates only

Org memory

  • Full traceability: conversation → spec → code → production
  • Verified outcomes improve future specs automatically
  • Knowledge persists independent of any individual
EnvisionProduction signals and conversations trigger PRD generation. Full organizational context synthesized. 🔵 Intent Gate confirms customer and Acceptance Criteria.
CreateAgent fleet builds from locked SPECs and Plans. 🟡 Review Gate: human confirms tech stack. Continuous alignment tracking flags drift against the locked SPEC.
DeliverAutomated Validation Criteria. 🔴 Ship Gate: governance monitoring active. Rollback plans generated. Compliance automated.
EvolveVerified outcome loop: Evolve Notes feed back to Envision. Organizational memory becomes the competitive asset.
Rikki tier: Rikki Platform. Spec-to-code alignment. Drift detection. Production governance. Verified outcome loop feeding back to Envision.
Model 4

AI-first

AI is the primary execution layer. Minimal team, maximum output, full governance.

People

  • Highly leveraged teams or individual founders
  • Focus: vision, customers, strategic judgment
  • Team size chosen by ambition, not overhead

Process

  • Intent expressed conversationally; agents deliver end-to-end
  • Governance, compliance, and quality automated
  • Enterprise-grade traceability at any team size

Org memory

  • Complete decision and outcome history, searchable
  • Each project improves the next via verified outcomes
  • Memory is the moat
EnvisionFounder converses. Rikki generates complete PRD with Acceptance Criteria from memory + new intent. 🔵 Intent Gate: founder confirms in minutes.
CreateFull agent orchestration. SPEC locked at 🟡 Review Gate. Agents build, test, and validate from Plan — minimal human involvement beyond gate approvals.
Deliver🔴 Ship Gate: Validation Criteria confirmed. Automated deployment with compliance. Monitoring and rollback in place from day one.
EvolveEvolve Notes close the loop. Production signals and customer feedback loop back. Every cycle makes Rikki smarter about this organization.
Rikki tier: Rikki Platform — Enterprise. Full orchestration. Automated compliance. One person, enterprise-grade governance.

Organizational memory replaces static documentation

Static documentationOrganizational memory
Written once, maintained manuallyGenerated from context, updated as code changes
Decays within daysLinked to source conversations — always traceable
Locked in wikis, rarely referencedSearchable by people and AI agents
No link to what was builtSpec-to-code traceability tracks divergence
Lost when people leavePersists independent of individuals
Same format regardless of contextAdapts to team size, industry, and operating model

Org structures Rikki supports

The right structure depends on your size, maturity, and industry. Rikki supports five patterns and helps you transition between them.

StructureSizeCharacteristicsModel
Split-leadership pods8–15Traditional roles with Code Architect and AI Integration Lead. Clear quality gates.1–2
Generalist teams4–8T-shaped builders across functions. AI fills skill gaps. Results-based.2–3
Orchestrator + agents2–4Humans design workflows and review. AI agents execute.3
Founder + AI stack1–2Full AI toolchain. Enterprise governance at solo scale.3–4
Hybrid enterprise50+Compressed middle layer. Orchestrators in pods. Dual-track.1–3
Rikki does the thinking. Every company building with AI faces the same questions: how to structure teams, what process to follow, what artifacts to produce, which governance standards apply. Researching this takes months. Maintaining it takes permanent effort. Rikki compresses the entire journey — calibrated to your profile, updated as the industry matures, ready from day one.