AI-Powered Platform¶

Kiket integrates artificial intelligence at every step of the software development lifecycle. Our AI capabilities are designed to enhance productivity while keeping humans in control of critical decisions.

AI Feature Overview¶

Capability	Description	Availability
Task Generation	Generate actionable tasks from requirements	Team+
Smart Assignment	AI-powered task assignment suggestions	Team+
Developer Aide Suite	Code review, bug detection, test generation	Professional+
Knowledge Automation	Meeting summaries, release notes, stakeholder briefs	Professional+
Predictive Analytics	Sprint completion forecasting, risk heatmaps	Professional+
Semantic Search	Natural language search across your workspace	Team+
Import Wizard	AI-assisted migration from other tools	All plans

Core AI Infrastructure¶

Agent Architecture¶

Every AI feature is powered by the Agent framework:

Agent Manifests define capabilities, context requirements, and approval rules
Prompt Templates use versioned, testable prompts with variable injection
Context Broker assembles relevant data (issues, code, analytics) automatically
Human-in-Loop workflows ensure oversight for sensitive operations

Learn more: AI Agents & Prompt Templates

Model & Provider¶

Kiket uses Google Vertex AI with Gemini models:

Primary: Gemini 2.5 Flash for low-latency operations
Fallback: Automatic failover and retry logic
Cost controls: Token limits per operation and organization

Data Privacy¶

AI processing happens within your Kiket tenant
No data is used for model training
Configurable redaction policies per organization
Full audit logging of all AI operations

Task Generation¶

Transform high-level requirements into actionable tasks:

# Example: Generate tasks from a user story
Input: "As a user, I want to reset my password via email"

Generated Tasks:
- Implement password reset email template
- Add forgot password endpoint to auth controller
- Create password reset token model and migration
- Build reset password form component
- Add rate limiting for reset requests

Features: - Confidence scoring (0.0–1.0) on each generated task - Automatic acceptance criteria from templates - Low-confidence tasks route to approval queue - Context-aware: considers existing issues, sprint history

Configuration: Set thresholds in your workflow definition:

ai_task_generation:
  enabled: true
  confidence_threshold: 0.65
  require_approval_below: 0.65

Smart Assignment¶

Get AI recommendations for task assignments:

How it works: 1. Analyzes team skills matrix and proficiency levels 2. Considers current workload and upcoming PTO 3. Reviews historical velocity and similar task outcomes 4. Suggests optimal assignee with confidence score

Assignment Rules: - ≥80% confidence with low workload impact → auto-assign (if enabled) - <80% confidence → route to approval queue - High risk or no capacity → suggest pairing partner

Skill-Aware Matching:

Task: "Implement OAuth2 authentication"
Required Skills: authentication, security, API design

Recommendations:
1. Alice (95% match) - Expert in auth, moderate load
2. Bob (82% match) - Good security skills, available
3. Pair: Charlie + Dana (78% combined) - Learning opportunity

Developer Aide Suite¶

AI-powered developer productivity tools available in issue and PR views:

Code Review Assistant¶

Analyzes code changes for issues and improvements
Suggests optimizations and best practices
Identifies potential security concerns
Available via command palette or PR integration

Bug Detection Scout¶

Scans code for common bug patterns
Highlights logic errors and edge cases
Prioritizes findings by severity
Integrates with issue tracking

Test Generator¶

Generates unit test scaffolds from code
Covers happy paths and edge cases
Supports multiple testing frameworks
Outputs to clipboard or file

Story Point Estimator¶

Suggests story point estimates based on historical data
Considers complexity, team velocity, and similar issues
Supports multiple scales (Fibonacci, T-shirt, linear)
Escalates to planning poker when uncertain

Access: Available on Professional+ plans via the Developer Aide panel in issue views.

Knowledge Automation¶

Automate documentation and communication:

Meeting Summaries¶

Ingest meeting transcripts from calendar integrations
Generate structured summaries with decisions and action items
Route to approval before publishing
Deliver via configured channels (Slack, email, in-app)

Release Notes¶

Draft release notes from merged PRs and closed issues
Include highlights, breaking changes, and migration notes
Support review/edit workflow before publication
Export to docs-site or changelog

Stakeholder Briefs¶

Generate executive summaries for different audiences
Persona-aware tone adjustment (executive, customer, audit)
Template library for common briefing formats
Quota: varies by plan (Professional: 3/month, Business: 20/month)

Learn more: Knowledge Automation Guide

Predictive Analytics¶

Data-driven forecasting for planning and release management:

Sprint Completion Probability¶

Predicts likelihood of completing sprint on time
Factors: velocity, scope, team capacity, historical patterns
Updates automatically as sprint progresses
Visual confidence intervals in milestone view

Release Risk Heatmaps¶

Aggregates risk signals across releases
Categories: blockers, incidents, dependencies, deployments
Color-coded severity (low/medium/high)
Drill-down to contributing factors

Bottleneck Detection¶

Identifies workflow stages with congestion
Measures average age and WIP ratio
Suggests automation or process adjustments
Feeds into assignment recommendations

What-If Scenarios¶

Model impact of capacity changes
Adjust velocity assumptions to see projected timelines
Compare baseline vs. stretch scenarios
Export projections for planning meetings

Access: Available on Professional+ plans via release workspace and executive dashboards.

Semantic Search¶

Natural language search across your entire workspace:

Search: "issues about authentication that are blocking the release"

Results:
- AUTH-123: OAuth token refresh failing (Blocker, linked to v2.0)
- AUTH-456: Session timeout not working (In Progress)
- Related: Security review document from last sprint

Features: - Embeddings-powered relevance ranking - Entity recognition (issues, people, dates) - Saved search views with AI-suggested filters - Cross-project search with permission scoping

Import Wizard¶

AI-assisted migration from Jira, GitHub Issues, and other tools:

Smart Field Mapping¶

Analyzes source system schema
Suggests field mappings with confidence scores
Handles custom fields and complex types
Preview transformations before import

Duplicate Detection¶

Compares incoming items with existing issues
Similarity scoring to identify potential matches
Merge or skip options for duplicates
Audit trail of deduplication decisions

History Reconciliation¶

Maps status transitions to Kiket workflows
Preserves activity timeline and comments
Links attachments and relationships
Validates data integrity post-import

Access: Available on all plans via Settings → Import.

Pricing & Quotas¶

AI features are included with subscriptions:

Plan	AI Features	Monthly Quota
Starter	Basic (duplicate detection, auto-labeling)	100 operations
Team	Standard (generation, assignment, search)	500 operations
Professional	Full (Developer Aide, Knowledge, Predictive)	2,000 operations
Business	Full + priority processing	10,000 operations
Enterprise	Full + dedicated capacity	Custom

What counts as an operation? - Each agent invocation (task generation, code review, etc.) - Knowledge document generation - Predictive forecast calculation

Not counted: - Semantic search queries - Duplicate detection (per-issue) - Auto-labeling

Governance & Controls¶

Organization Settings¶

# AI settings in organization configuration
ai:
  enabled: true
  features:
    task_generation: true
    smart_assignment: true
    developer_aide_suite: true
    knowledge_automation: false  # Disabled for this org
  approval_required:
    - low_confidence_tasks
    - stakeholder_communications
  redaction_policy:
    enabled: true
    patterns:
      - email
      - phone
      - ssn

Audit Logging¶

All AI operations are logged: - Agent ID and version - Input summary (redacted) - Output summary - Token usage and cost - User who triggered the operation - Approval chain (if applicable)

Disabling AI Features¶

Organization admins can disable any AI feature: 1. Navigate to Settings → AI Features 2. Toggle individual features on/off 3. Changes take effect immediately

Getting Started¶

Check your plan – Most AI features require Team or higher
Review settings – Configure which features are enabled
Try the command palette – Press Cmd/Ctrl+K and search for "AI"
Read the guides – Each feature has detailed documentation

Additional Resources¶

FAQ¶

Q: Can I use my own AI models? A: Not currently. Kiket uses Vertex AI for consistent quality and security. Enterprise customers can discuss custom model options.

Q: Is my data used to train models? A: No. Your data is used only for your organization's AI operations. We do not use customer data for model training.

Q: What happens when I hit my quota? A: AI operations will return a quota error. You can upgrade your plan or wait for the next billing cycle.

Q: Can I see what data was sent to the AI? A: Yes. All AI operations are logged in the audit log with input summaries (respecting redaction policies).

Last updated: December 2025