Great BA work is not only writing requirements well; it is ensuring requirements are findable, understandable, and not duplicated across the project lifecycle. That is why Confluence and Notion are essential.
1. Why BA Need a Standard Knowledge Base
In AI projects, information changes quickly:
- Acceptance thresholds are adjusted after sprint reviews
- Data assumptions break when real data arrives
- New stakeholders join and need fast onboarding
Without a single source of truth -> each team member understands things differently -> expensive rework.
2. Confluence Space Structure for AI Projects
📁 [Product/Project Name]
├── 📄 Project Overview (purpose, stakeholders, timeline)
├── 📁 Requirements
│ ├── 📄 BRD — Business Requirements Document
│ ├── 📄 FRD — Functional Requirements Document
│ ├── 📄 Non-functional Requirements (performance, security)
│ └── 📄 Assumption Log
├── 📁 AI Specifics
│ ├── 📄 AI Feature Inventory (name, purpose, owner)
│ ├── 📄 Model Evaluation Criteria
│ ├── 📄 Data Dictionary
│ └── 📄 Bias & Fairness Checklist
├── 📁 Design
│ ├── 📄 User Journey Maps
│ ├── 📄 Flow Diagrams (BPMN/UML)
│ └── 📄 Wireframes (Figma links)
├── 📁 Testing
│ ├── 📄 UAT Plan & Scripts
│ └── 📄 Bug/Issue Log
└── 📁 Decisions
├── 📄 ADR Log (Architecture Decision Records)
└── 📄 Meeting Notes
3. Standard BRD Template for AI Projects
# Business Requirements Document
## [Feature Name] v[X.X]
| Field | Value |
|-------|-------|
| Author | [BA Name] |
| Status | Draft / In Review / Approved |
| Last Updated | YYYY-MM-DD |
| Stakeholders | [List] |
## 1. Problem Statement
[What problem is happening? For whom? What is the business impact?]
## 2. Business Objectives
- Objective 1: [Measurable]
- Objective 2: [Measurable]
## 3. Success Metrics
| Metric | Baseline | Target | Measurement Method |
|--------|----------|--------|--------------------|
| [metric] | [now] | [target] | [how] |
## 4. Scope
### In Scope
- [Item 1]
### Out of Scope
- [Item 1] — reason: [why]
## 5. AI-Specific Requirements
| Aspect | Requirement |
|--------|-------------|
| Accuracy Threshold | >= X% on test set Y |
| Fallback Behavior | When confidence < Z: [action] |
| Human Override | Who has authority? What trigger? |
| Data Freshness | Retraining every [period] |
## 6. Assumption Log
| ID | Assumption | Owner | Review Date | Status |
|----|-----------|-------|-------------|--------|
| A001 | [assumption] | [name] | [date] | Active/Invalid |
## 7. Constraints
- Legal: [GDPR, industry regulation]
- Technical: [API limits, data retention]
- Budget: [budget ceiling]
4. Linking Requirements with Jira/Azure DevOps
Best practice: every requirement in BRD should map to a corresponding Jira ticket.
In Confluence, use Jira macro to embed ticket status:
Requirements Section 3.1 -> PROJ-123 (Story: "As a user, I want...")
Acceptance Criteria 3.1.a -> PROJ-124 (Test Case)
When ticket status changes, Confluence updates automatically.
5. Notion as a Confluence Alternative: When to Use
| Criteria | Confluence | Notion |
|---|---|---|
| Atlassian integration | ✅ Native (Jira, Bitbucket) | ❌ Requires third-party |
| AI Assistant | ✅ Atlassian Intelligence | ✅ Notion AI is stronger |
| Flexibility | ⚠️ Template-heavy | ✅ Very flexible |
| Database/Spreadsheet | ❌ Limited | ✅ Strong (relational DB) |
| Free tier | ❌ Paid | ✅ Generous free tier |
| Best for | Enterprise, Jira-based teams | Startups, small teams |
Notion Database for Requirement Tracking:
Create a database with properties:
- Requirement ID (Text)
- Type (Select: Functional/Non-functional/AI-specific)
- Status (Select: Draft/Approved/Implemented/Deprecated)
- Priority (Select: P0/P1/P2)
- Linked Sprint (Relation -> Sprint database)
- AC Count (Number: acceptance criteria count)
6. Common Mistakes and How to Avoid Them
Mistake 1: Copy-pasting requirements across multiple pages
-> Use Confluence "Include Page" macro or Notion synced blocks
Mistake 2: Assumptions without owner
-> Every row in Assumption Log must include a responsible validator
Mistake 3: Old BRD versions not archived clearly
-> Use Confluence Page History and mark old pages as "ARCHIVED"
Mistake 4: Meeting notes not linked back to impacted requirements
-> Meeting note template should include "Requirements Impacted: [link]"
Conclusion
Confluence/Notion is the project memory. BA who invest early in structure save dozens of clarification hours later, especially when AI teams must explain "why the model behaved this way" six months after go-live.