Pilot (30-45 days)
A pilot is the fastest way to confirm fit and quantify improvements without changing your live routing process.
Key principle: Start in shadow mode. Clarity makes routing recommendations while your existing process stays in control.
Pilot scope
Typical scope
- 1 routing problem (“domain”)
- 20-200 categories
- 2k-25k historical labeled items
- 2-4 weeks shadow mode on live intake
What you provide
- Export of historical items with a final queue/team/category label
- A mapping of labels to your current queues/teams
- Optional: brief category definitions or runbooks
Success metrics (pick 2-3)
- Top-1 routing accuracy
- Misroute / reassignment rate reduction
- % auto-routable above a confidence threshold
- Time-to-first-assignment reduction
Deliverables
- Offline evaluation report (including a confusion matrix)
- Shadow-mode metrics snapshot
- Go/no-go recommendation for write-back routing
Timeline (example)
- Week 1 - Data intake + baseline evaluation
- Week 2 - Tuning + threshold selection
- Weeks 3-4 - Shadow mode on live intake + monitoring
- Week 5 (optional) - Write-back test (only if go)
What a “go” looks like
A pilot is a “go” when you can point to specific, measurable wins and a safe operating plan:
- Clear accuracy and misroute reduction
- Confidence thresholds tied to QA lanes
- Operational guardrails for exceptions
Next step
If you can export labeled historical tickets/cases, we can quickly confirm fit and outline the shadow-mode pilot.