Consistent ticket quality at every workflow stage

Request a Demo

Clarity Documentation

 

Embank Clarity is a Jira Cloud Forge app that automatically evaluates ticket quality using AI and enforces quality standards through graduated blocking, feedback, and templating.

Embank Clarity creates a quality gate without friction. Teams get:

  • Clear feedback on why a ticket isn’t ready
  • Guided improvements through structured comments
  • Flexibility to override with justification when necessary
  • Visibility into quality trends via private analytics
  • Control via role-based admin settings

The graduated blocking approach (hard block → warning → silent pass) balances enforcement with pragmatism—critical gaps stop work, minor gaps warn, and good work goes silently forward.

 

Key Capabilities

1. Quality Review & Scoring (0-10 Scale)

  • Evaluates ticket summaries and descriptions against quality criteria
  • Posts actionable review comments with structured feedback
  • Uses configurable LLM providers (OpenAI, Anthropic, AWS Bedrock)

 

 

Clarity Documentation

Embank Clarity is a Jira Cloud Forge app that automatically evaluates ticket quality using AI and enforces quality standards through graduated blocking, feedback, and templating.

Embank Clarity creates a quality gate without friction. Teams get:

  • Clear feedback on why a ticket isn’t ready
  • Guided improvements through structured comments
  • Flexibility to override with justification when necessary
  • Visibility into quality trends via private analytics
  • Control via role-based admin settings

The graduated blocking approach (hard block → warning → silent pass) balances enforcement with pragmatism—critical gaps stop work, minor gaps warn, and good work goes silently forward.

 

 

2. Graduated Blocking System

Three tiers of enforcement visible in the configuration:

  • Red Zone (≤4): Hard block—tickets cannot proceed without override
  • Amber Zone (5-7): Warning—tickets can proceed but with caution notice
  • Green Zone (≥8): Silent pass—no comments posted

The screenshots show this system in action:

  • A score of 4/10 triggers a hard block with the message “Action needed—this ticket may cause work”
  • Users see specific improvement areas (user intent clarity, action clarity, dependencies, done signal)
  • Overrides require justification and are logged for audit trails

 

 

3. Flexible Review Rules

Admin configuration includes:

  • Issue types: Epic, Story, Task, Bug, Improvement (all checked in example)
  • Status gates: Triggers at Backlog, Refinement, To Do, In Progress, In Review, Done
  • Commitment transitions: Blocks tickets moving to “In Progress” or similar commitment statuses if quality is insufficient
  • Cooldown: 1-minute default prevents review spam on rapid updates
  • Positive feedback: Optional toggle to praise high-quality tickets (8+)

4. LLM Configuration

  • Provider selection: OpenAI (ChatGPT shown), Anthropic, or AWS Bedrock
  • Model selection: gpt-4o in example, with fallback guidance (gpt-4o-mini noted as typical)
  • Secure API key storage with update capability

5. Project Scoping

  • Comma-separated project keys (e.g., “SCRUM”) limit which projects receive reviews
  • Enables targeted rollout and prevents org-wide disruption

6. Licensing & Analytics

  • License status display (Active/Inactive)
  • License type: Development or Marketplace
  • Billing managed through Atlassian Marketplace
  • Private analytics support for tracking quality trends

.

User Experience in Jira

Blocking Behavior

The screenshot shows:

  • Clear risk indicators with emoji (🔴 RED for high risk)
  • Explicit statement of risk level
  • Instructions to “confirm or clarify expected behavior or explicitly accept the risk”

Feedback Structure

Review comments follow a standardized format:

  • What’s Good: Highlights existing strengths (clear summary, acceptance criteria listed)
  • To Improve: Specific gaps identified (missing user story format, vague action steps, undocumented dependencies, lack of done signal)

Override Workflow

  • Users can override red-zone blocks by clicking “Unblock Ticket” action
  • Justification required: “risk acceptable, lets test it”
  • Logged with user attribution (@kieran in example)
  • Bypass status shown in subsequent comments: “⚠️ Clarity Block Overridden”

Configuration Highlights

From the settings screenshots:

Setting Value Purpose
Minimum Score Threshold 8 Comments posted only below this (unless positive feedback enabled)
Green Zone 8-10 Silent pass
Amber Zone 5-7 Warning; proceed allowed
Cooldown 1 minute Max one auto-review per issue per window
Backlog Treat as backlog-like ✓ “To Do” statuses treated as backlog when not in sprint
Blocking on Commitment ✓ Hard block on transitions to “In Progress”

Security & Data Handling

  • Encrypted API key storage in Forge app storage
  • Issue properties for idempotency and spam suppression
  • Audit logging for overrides
  • Analytics aggregate-oriented (avoids PII exposure)