Specification-First Delivery

Build on the delivery system you already have

Specification-First Delivery with Existing Methods and Tools

You do not need to replace the methods and tools that already help your team deliver. Specification-First Delivery connects them to shared knowledge, explicit decisions, reviewed specifications, acceptance evidence, and accountable human judgment.

How it fits

Specification-First Delivery works with your existing tools

Start with the method or tool your team already uses. Each section shows what it continues to do well and how Specification-First Delivery adds the context, decisions, specifications, authority, and evidence that accountable human-AI delivery requires.

Agile / Scrum

Agile and Scrum organize iterative delivery, prioritization, feedback, and team cadence. Specification-First Delivery keeps those strengths while making the judgment and knowledge behind each increment usable throughout discovery, implementation, and acceptance.

What Agile and Scrum organize

  • Iteration planning, delivery cadence, and a regular rhythm for team collaboration.
  • Prioritization, backlog management, and choices about what to deliver next.
  • Discovery, refinement, reviews, and retrospectives that let teams inspect and adapt.
  • Feedback from delivery outcomes, users, stakeholders, and operations.
  • Small increments that produce visible outcomes and business value.

What Specification-First Delivery strengthens

  • Discussion and specification work become part of discovery, refinement, and delivery rather than a separate preliminary phase.
  • Backlog items retain their prioritization role, while reviewed specifications carry enough intent to guide implementation.
  • Definition of Ready can include resolved material ambiguity, ownership, and decision authority.
  • Definition of Done can include specification conformance, acceptance evidence, and returned learning.
  • Specifications evolve through feedback and delivery increments.
  • AI coding agents can implement within reviewed specifications, so teams retain judgment and knowledge rather than recreating material decisions during code creation.

Specification-Driven Development tooling

Specification-Driven Development tooling, including GitHub Spec Kit, Kiro, and similar tools, can organize substantial parts of specification-led software development: principles, requirements, clarification, technical planning, task decomposition, implementation, consistency checking, and verification. Specification-First Delivery is not a competing execution tool. It provides the broader engineering framework that connects those workflows to organizational knowledge, professional authority, accountable authorship, acceptance evidence, and continuity over time.

What workflow tooling can organize

  • Establish project principles, constitutions, presets, and reusable guidance.
  • Capture required outcomes, scenarios, constraints, and acceptance criteria.
  • Clarify ambiguity through structured refinement.
  • Produce technical plans, research, models, contracts, and implementation context.
  • Decompose work into tasks, dependencies, sequencing, and opportunities for parallel execution.
  • Support implementation through AI coding tools.
  • Check consistency among specifications, plans, tasks, tests, and implementation.
  • Preserve specification artifacts as context for future changes.

What Specification-First Delivery adds

  • Organizational knowledge beyond the immediate development workflow.
  • Explicit authoritative sources and decision ownership.
  • Direct cross-functional authorship and accountable professional judgment.
  • Relationships across functional, architecture, security, reliability, operational, technical, and test specifications.
  • Named human accountability for material decisions.
  • Acceptance evidence connected to the requirements and roles that own them.
  • Approved learning returned to the organizational Knowledge System.
  • Capability Continuity across changes in people, teams, models, and tools.

AI coding agents

AI coding agents can apply repository guidance, inspect code and documents, implement changes, and run available checks. Each agent begins without your organization’s history, constraints, and decision context.

What AI coding agents offer

  • A way to use governing instructions such as AGENTS.md and reusable skills as repository-level guidance.
  • The ability to inspect repository files and use available specifications and documentation.
  • The ability to draft technical approaches, implement code, and revise a change.
  • Task execution within the authority assigned by the invoking workflow.
  • The ability to run tests and delivery automation, then use the output to improve a change.
  • Implementation output that people can review, accept, or use as learning for future work.

What Specification-First Delivery adds

  • A cross-functional process for deciding when and how governing instructions and skills change, including durable repository constraints and the knowledge that must remain outside the repository.
  • Reviewed specifications, structured and reviewed by the relevant roles, that define the intended outcome and material decisions.
  • Technical specifications that establish appropriate implementation scope and constraints, reviewed by engineers through a documented prompting process.
  • Explicit execution authority and Job Delivery Graphs that make parallel work, dependencies, and decision rights visible.
  • Quality-owned test specifications and delivery-pipeline evidence against the agreed specifications.
  • Approved implementation learning returned to both the specifications and the organizational Knowledge System.

DevOps and CI/CD

DevOps practices and CI/CD pipelines automate the build, test, deployment, and operational feedback needed to move a change through its delivery path.

What DevOps and CI/CD continue to provide

  • Infrastructure as code, environment configuration, and repeatable deployment paths.
  • Automated build, test, security, and deployment checks.
  • Deployment records and operational feedback.

What Specification-First Delivery adds

  • Architecture and deployment knowledge, including system responsibilities, environment details, deployment paths, and dependencies.
  • Infrastructure specifications that state intended configuration changes alongside functional and technical work.
  • Explicit operational and security constraints for reliability, rollout, recovery, and environments before execution.
  • Pipeline evidence connected to the relevant specifications for acceptance and release decisions.

Begin with a real change

Start with your existing delivery workflow.

Keep the current methods and tools, then make their relationships visible through one real delivery increment.

  1. 1Keep the current methods and tools.
  2. 2Identify where knowledge, decisions, specifications, and acceptance evidence currently live.
  3. 3Connect them through one real change.
  4. 4Make ownership and authority explicit.
  5. 5Use existing tools to execute and verify the specifications.
  6. 6Return reviewed learning to the Knowledge System.

Specification-First Delivery makes the relationships among your existing methods, tools, people, decisions, and evidence explicit enough for accountable human-AI delivery.

Explore adoption guidance