What is spec-driven AI?

Explained

Spec-driven AI treats written specifications as the source of truth for how an AI system should behave. Requirements documents, acceptance criteria, API and tool schemas and policy rules are connected to the prompts, the retrieval content and the evaluation suites, so that a change starts in the spec and flows from there. This keeps behaviour traceable: anyone can see which requirement a prompt, a tool or a test comes from. It is a more governable alternative to adjusting prompts by trial and error, and it is closely related to spec-driven development, where AI coding agents build software from specifications.

How it works

• The spec states what the system must do and must not do, for whom and with which data.
• Tool and API contracts define exactly which actions a model or agent may take and with which inputs.
• Policies and guardrails, written as rules, set limits on content, data access and approvals.
• Prompts and agent instructions are derived from the spec rather than written separately.
• Evaluation cases are generated from the acceptance criteria, so tests check what the spec asked for.
• When the spec changes, prompts, tools and tests are updated together.

Spec-driven AI and spec-driven development

Spec-driven development applies the same idea to building software with AI coding agents: write the specification, turn it into a technical plan and a list of tasks, and let agents implement against it, with people reviewing at each stage. Open-source tools such as GitHub's Spec Kit package this flow.

Spec-driven AI extends it to the behaviour of AI systems in production. In both cases the spec, not the prompt or the generated code, is the artefact people review and maintain.

Why it helps

• Traceability: each behaviour can be traced back to a requirement, which supports audit and regulatory needs.
• Consistency: several prompts, agents and tests stay aligned because they come from one source.
• Safer change: a change is reviewed once in the spec and checked by the derived tests.
• Shared language: business owners can read and approve the spec without reading prompts or code.
• The cost is discipline: specs must be kept current, or they become one more outdated document.

Frequently asked questions

Further reading

• AI Design Platforms: The Four Pillars of Autonomous Product Innovation
• AI-Native Application Development: A C-Suite Guide
• Agentic software engineering: how autonomous AI builds enterprise code

Related terms

• API A defined interface that lets one software system request data or actions from another; it is how applications call AI models and connect them to business systems.
• Evaluation The process of measuring how well an AI model or application performs: accuracy, relevance, safety, bias, cost and other criteria.
• Prompt The input given to an AI model, such as instructions, context, examples and the question, that guides what it produces.
• Retrieval The search step that fetches the documents or passages most relevant to a query, before they are shown or passed to a model.
• Guardrails Runtime controls and policies that check what goes into and comes out of an AI model, to keep its behaviour safe, on topic and within the rules.
• Spec-Driven Development A way of building software in which an explicit, reviewable specification is written first and code, tests and docs are derived from it.