Jardine AI logoJardine
Sign Up

Setup Guide

This guide is for teams that have finished first setup and now want Jardine to work reliably every day. Think of it as your "go from working to dependable" playbook.

By the end, you should have:

  1. A definitive idea of what Jardine is allowed to do and what humans are allowed to do.
  2. Clean, usable knowledge and operations material.
  3. A stable flow of tickets coming from your support channels to Jardine.
  4. Routing and escalation logic that your team trusts.
  5. A repeatable operating rhythm after launch.

Before You Start

You need:

  1. A channel already connected (Intercom, Zendesk, or Email).
  2. Admin access to your Jardine workspace.
  3. At least one teammate responsible for quality checks.

1. Decide what goes live first

Start small. Stable systems are easier to build when the first version is narrow and clear. Pick one channel and one support area first.

  1. Assign one channel owner.
  2. Assign one policy owner.
  3. Assign one person responsible for quality review.
  4. Define what "good" means for this phase (accurate answers, correct escalation, and consistent tone).

2. Prepare your source material

Before you tune behavior, make sure Jardine has the right material. This step reduces confusion later when replies feel inconsistent.

  1. Add product facts and FAQs to Knowledge Library.
  2. Add support boundaries and playbooks to Operations Library.
  3. Remove old or conflicting documents that should no longer guide responses.

You are in a good place when:

  1. Important FAQs are covered.
  2. Sensitive boundaries are explicit.
  3. There is no conflicting guidance across documents.

3. Confirm tickets are reaching Jardine

Do this before deeper AI tuning. If events are incomplete or delayed, every other quality check becomes misleading.

  1. Confirm your channel connection is complete (OAuth + webhook for Intercom/Zendesk, or domain + forwarding for Email).
  2. Confirm the required events/triggers are correctly configured on Intercom, Zendesk, or your email setup.
  3. Send test messages and confirm they appear in Conversations.

4. Configure routing and escalation rules

Use Routing to define clear, explainable behavior. If a human reads your rules and cannot explain them in 30 seconds, simplify them.

  1. Create a small initial tag set.
  2. Map sensitive tags to escalation destinations that have clear owners.
  3. Decide what should happen when Jardine is unsure or a case is high-risk.
  4. Avoid too many rules in the first pass.

You are ready to proceed when:

  1. Sensitive cases escalate consistently.
  2. Destination ownership is clear.
  3. The default path for unclear cases is predictable.

5. Check response quality

Use Tools Playground with realistic scenarios. Test both easy and difficult cases, not only the "happy path."

  1. Check routine answers for accuracy.
  2. Check sensitive prompts for correct escalation.
  3. Review tone and response clarity with your team.
  4. Fix major gaps before widening rollout.

6. Roll out in stages

Rollout works best when each stage has a clear purpose.

  1. Start with Generate AI Drafts so your team can review behavior safely.
  2. Run daily quality checks during this phase.
  3. Enable Auto-send AI Replies only for stable categories.
  4. Expand gradually, one channel segment or tag group at a time.

Operating Rhythm

After launch, consistency matters more than speed. Use a simple daily and weekly rhythm.

Daily

  1. Check Control Center.
  2. Resolve blocked or overdue items.
  3. Use Chat with Jardine for follow-up actions and operational checks.

Weekly

  1. Review escalations, misses, and surprises.
  2. Update knowledge and policy documents based on what you learned.
  3. Remove noisy or redundant rules.
  4. Re-test important scenarios.

If quality drops

When quality slips, avoid broad changes. Reset to control, then recover in small steps.

  1. Pause rollout expansion.
  2. Move sensitive paths back to review mode.
  3. Fix one layer at a time (content, routing rules, or channel setup).
  4. Re-test before expanding again.

Done Checklist

You are ready for broader rollout when your team can answer:

  1. Which channels are currently live?
  2. Which cases escalate, where, and why?
  3. Who owns daily quality and weekly maintenance?

Next: Troubleshooting.