BA / FA / PO Reference · 2026

Functional Specs
Toolkit

Purpose-built tools beat generic LLMs for consistent spec output. Structure acceptance criteria with EARS "shall" statements when specs feed AI coding agents. AI drafts — humans approve, always.

7 tools compared 4 EARS types 3 prompt templates 21 citations 7 min read

Top picks →

Kiro AWS greenfield pipelines ChatPRD stakeholder-facing PRDs Copilot4DevOps Azure DevOps native

Tool Matrix

7 tools · choose by integration, AI-agent fit, and draft speed

AI-agent optimised Partial Human-first

Kiro AI ✓

AWS greenfield; EARS → code pipeline. Generates requirements, design doc, and task list before any code runs. ^[9]

⚡ Minutes

AWS-native; limited brownfield support

ChatPRD Human

Stakeholder-facing PRDs; PM coaching; retains organisational memory across sessions. ^[5]

⚡ 5–10 min

No customer evidence layer

Keeborg AI ✓

Full spec stack for AI coding agents — generates CLAUDE.md and .cursorrules alongside 8 interconnected spec documents. ^[3]

⚡ 90 sec

New entrant; niche ecosystem

Copilot4DevOps Partial

Azure DevOps-native FRDs, user stories, and test cases. Elicit feature generates child stories from parent work items. ^[18]

⚡ Minutes

Azure DevOps only

Jama Connect Partial

Regulated industries; INCOSE/EARS inline scoring. Advisor checks requirements against EARS rules as you type. ^[6]

⚡ Minutes

Enterprise pricing

aqua Human

Voice-to-requirement; rapid BA field capture. Converts voice notes into structured requirements in 15 seconds. ^[7]

⚡ 15 sec

Shallow on complex logic

GitHub Spec Kit AI ✓

Portable, MIT-licensed, agent-agnostic spec-driven dev framework. Works with any AI coding agent. ^[10]

⚡ Minutes

High review burden; many files

EARS Quick Reference

Rolls-Royce → adopted by Airbus, Bosch, NASA, Siemens ^[12] · use when specs feed AI coding agents

Core pattern: While <precondition>, when <trigger>, the <system> shall <response>. — Mix clauses as needed.

Ubiquitous

The [system] shall [response].

eg: The system shall encrypt all PII at rest using AES-256.

Event-Driven

When [trigger], the [system] shall [response].

eg: When a payment fails, the system shall notify the user within 5 seconds.

State-Driven

While [state], the [system] shall [response].

eg: While unauthenticated, the system shall redirect to /login.

Conditional

Where [feature enabled], the [system] shall [response].

eg: Where MFA is enabled, the system shall prompt on each new device.

Spec-Driven Development Pipeline

Specs become agent-executable contracts, not handoff artifacts ^[10]

Spec-first

Specs precede code, then are discarded. Lowest overhead, but specs drift from implementation immediately.

⚠ highest drift risk

Spec-anchored

Specs persist alongside code and update as features evolve. Cross-role authorship keeps them current.

✓ practical target

Spec-as-source

Humans edit only specs, never code directly. AI derives implementation from specs alone. Currently experimental.

⚗ experimental

📋

Requirements

user stories + EARS ACs

→

🏗

Design Doc

architecture + constraints

→

✅

Task List

reviewed before code runs

→

🤖

AI Generates

code from spec contract

→

👁

Human Review

intent, not syntax

Prompt Templates

Role + Context + Task + Constraints + Output format ^[16] · works with Claude, ChatGPT, Copilot

📝 Meeting Notes → Functional Requirement

You are a senior BA. Here are raw stakeholder notes: [paste notes] Convert to one EARS-formatted functional requirement covering: - Functional context - Business logic (IF/WHEN/THEN, numbered) - Happy path (numbered steps) - Unhappy path (numbered steps) - Open questions

📖 User Story + Acceptance Criteria

You are a product owner. Create a user story for: [feature description] Format: "As a [persona], I want [goal] so that [benefit]." Add 5 acceptance criteria in EARS event-driven format (When/Then). Flag edge cases and missing non-functional requirements.

🔍 FRD Gap Analysis

You are a senior requirements engineer. Review this FRD: [paste document] Identify: (1) Ambiguous or missing requirements (2) EARS structure violations (3) Missing non-functionals: performance, security, a11y Output as a numbered list with severity: high / medium / low.

Platform-Native Integrations

Avoid context-switching — use what's already in your ecosystem

Jira / Atlassian Rovo

Drafts acceptance criteria from Confluence docs; flags gaps in user stories. Draws on connected documentation automatically. ^[17]

→ "Generate AC" in ticket view

Azure DevOps / Copilot4DevOps

Generates FRDs, user stories, and test cases from parent work items. One-click backlog generation from a PRD. ^[18]

→ Elicit from parent work item

Confluence Rovo

20+ agents: summarise, Q&A, content outline. Converts meeting notes into structured requirement pages automatically.

→ Convert notes to requirements page

Notion AI

Agents run autonomously for up to 20 minutes. Process sprint retrospectives into updated backlogs; 85–90% reduction in content creation time. ^[19]

→ Process retrospectives → backlog

Critical Rules

What AI cannot replace ^[6]

In regulated industries (ISO 26262, DO-178C, IEC 62304) human review is a structural requirement, not a preference. For all teams: AI drafts → human approves → refinement session before sprint backlog.

Stakeholder politics or unstated constraints
Legal, compliance, or safety weight of a requirement
Tradeoff calls when requirements conflict
Validating generated language matches real intent

Two-audience discipline ^[1]^[2]

Every spec in 2026 serves humans and AI coding agents simultaneously. Optimising for only one audience breaks the other.

Human layer: plain language, rationale, stakeholder narrative.
Agent layer: structured headings, "shall" clauses, explicit out-of-scope list, non-functionals.

Pattern: write human-first, then add a machine layer — .cursorrules, CLAUDE.md, or EARS-structured ACs.