Discover

Find and import historical decisions from your existing tools.

Overview

Your team has already made hundreds of decisions buried in Jira tickets, GitHub PRs, and Slack history. Discover uses AI to find these decisions and import them into Align.

How It Works

┌──────────────────┐     ┌──────────────────┐     ┌──────────────────┐
│   Connect        │ ──▶ │   Scan           │ ──▶ │   Review         │
│   Data Source    │     │   AI Analysis    │     │   & Import       │
└──────────────────┘     └──────────────────┘     └──────────────────┘

1. Connect - Link your Jira, GitHub, or other data source
2. Scan - Align's AI analyzes content for decision-like patterns
3. Review - You review suggested decisions
4. Import - Confirmed decisions become searchable snapshots

Supported Sources

Eleven connectors can be scanned:

Source	What's Scanned
Jira	Issues, comments, epics
GitHub	PRs, issues, ADRs, discussions
Slack	Channel history
Microsoft Teams	Channel and chat history
Confluence	Pages, decision logs
Zoom	Meeting transcripts
Linear	Issue and project history
Notion	Pages shared with Align's Notion app
Azure DevOps	Work items and pull requests
Microsoft 365 Documents	SharePoint site pages and library or OneDrive files
Google Drive	Drive files

A single scan request can cover up to all eleven at once.

Notion scans only the pages you share

An admin connects Notion once for the workspace. Notion's consent screen asks which pages and databases to share, and Align reads only those and everything under them. Align's Notion app has the "Read content" capability and nothing else, so it cannot edit a page. Notion does not return a page you did not share, and Align fetches only the pages Notion lists. To widen a scan, share more pages with the Align integration in Notion, then scan again. A scan reads each page's title and the first 300 blocks of its body, up to 8,000 characters.

GitLab and raw git come in through the CLI

Discover cannot scan these. They have no hosted connector. The free Align CLI imports them with align import gitlab and align import git.

Running a Discovery

Step 1: Configure Source

1. Go to Discover in Align
2. Click New Discovery
3. Select your source (e.g., Jira)
4. Configure:
   • Projects/Repos - Which ones to scan
   • Date Range - How far back to look
   • Filters - Labels, issue types, etc.

Step 2: Run Scan

1. Click Start Discovery
2. Wait for AI analysis (typically 1-5 minutes)
3. Progress is shown in real-time

Step 3: Review Results

For each suggested decision:

• Review the AI-generated summary
• Accept to import as a Decision Snapshot
• Reject if it's not actually a decision
• Edit to refine before importing

Step 4: Import

1. Select decisions to import
2. Click Import Selected
3. Decisions are now searchable in Align

What Discover Looks For

Decision Signals

Align's AI looks for patterns like:

• "We decided to..."
• "Going with option X"
• "Approved: ..."
• "After discussion, we'll..."
• "Sign-off received for..."
• "RFC accepted"

Document Patterns

• Architecture Decision Records (ADRs)
• RFCs (Request for Comments)
• Design documents
• Meeting notes with decisions
• Sign-off comments

Issue Patterns

• Issues with "decision" labels
• Closed issues with approval comments
• Epics with architectural scope
• Discussions with resolution

Discovery Settings

Date Range

Range	Best For
Last 30 days	Recent decisions, quick scan
Last 90 days	Quarter's decisions
Last year	Major architectural decisions
All time	Full historical import

Confidence Threshold

• High - Only very clear decisions (fewer false positives)
• Medium - Balanced (recommended)
• Low - More suggestions (requires more review)

Best Practices

1. Start Narrow

Begin with a specific project and recent timeframe:

Project: platform-core
Date range: Last 90 days

2. Use Filters

Focus on high-value content:

• Filter by labels: architecture, adr, decision
• Filter by issue type: Epic, Story
• Exclude: Bug, Task

3. Review Carefully

Not everything is a decision:

• ✅ "We decided to use Kubernetes" → Import
• ❌ "Discussed Kubernetes options" → Reject (no decision yet)
• ❌ "Fixed the bug" → Reject (not a decision)

4. Iterate

Run multiple discoveries:

1. First pass: High confidence, last 30 days
2. Second pass: Medium confidence, last quarter
3. Third pass: Specific projects/labels

Bulk Approval

When you're ready to import, select multiple suggestions and approve them at once. Align uses a two-phase approval flow to make this feel instant:

Phase 1: Instant Links

During the scan, Align automatically detects cross-platform references between suggestions - for example, the same Jira key (PROJ-123) appearing in both a Jira ticket and a Slack message. When you approve, these precomputed links are promoted instantly (< 1 second), and you'll see "Linked" status messages appear immediately.

Phase 2: AI Analysis

For decisions without obvious cross-references, Align's AI analyzes them in the background to find semantic relationships (supersessions, conflicts, duplicates). This runs only for items that weren't already linked in Phase 1, keeping the process fast.

What Gets Linked Automatically

Pattern	Example	Confidence
Jira key in multiple sources	PROJ-123 in Jira + Slack	100%
GitHub PR reference	org/repo#42 in GitHub + Jira	100%
Same topic, different tools	Architecture discussion in Slack → ADR in GitHub	AI-detected

One-click Tier review

The "Tier review" group lets a person accept a whole batch of high-confidence suggestions in one action. That records that a person accepted the batch. It does not record that they read each suggestion. The group is off for new accounts, and their suggestions stay in the list for one-by-one review.

After Import

Imported decisions:

• Appear in your Align dashboard
• Are fully searchable
• Link back to source evidence
• Are automatically linked to related decisions across platforms
• Can be edited and enriched