Quickstart

The best way to understand Jardine AI is to run one real support question through the product from start to finish. Not a fake demo question, and not a perfect “happy path,” but something your team actually sees in the queue. Once you do that once, the rest of setup becomes much easier to reason about.

A lot of teams lose time here because they try to wire everything in one pass: every channel, every routing rule, every connector. You do not need that to get a meaningful win. What you need is one reliable path you trust.

Get One Flow Working End to End

Start in your dashboard and open Getting Started. If your organization is brand new, begin with the basics in Settings -> Organization so your team identity is set correctly. In the current product experience, organizations operate in single-workspace mode, so you can focus on one operating context instead of switching between workspaces.

Then move to Knowledge -> Library and add one useful source your agents already trust. A short billing policy, a refund policy, or a plan-change FAQ is perfect. For this stage, one clear source is better than many mixed-quality files.

After upload, wait until your document is indexed and active. Jardine can only ground answers on content that has finished ingestion.

Now open Knowledge -> Validate. In simple view, you get a lightweight summary. Toggle to Advanced Validation when you want to run full test chats and inspect evidence. Ask a real customer-style question such as:

Can I change plans in the middle of a billing cycle?

Read the answer like an operator, not like a demo reviewer. Is the wording clear enough for a customer? Does it match your actual policy? Do the supporting citations and grounding signals make sense? If the answer is weak, improve the source content first. In early setup, source quality almost always matters more than clever tuning.

At this point, you already have a concrete win: Jardine has answered a real question using your own knowledge.

Make It Operationally Safe

Once the answer looks good, the next priority is safety. The goal is not to automate everything immediately. The goal is to automate what is predictable and hand off what is sensitive.

Go to Routing. By default, Jardine handles most routing automatically. Keep that default behavior for your first pass. If you need tighter control, open Manual Controls and define explicit tags, destinations, and rules. A practical first manual setup is simple: keep a default tag, add one destination for human follow-up, and create one rule for clearly sensitive traffic.

Then connect one inbound channel in Settings. Pick only one channel for your first launch cycle so feedback stays easy to interpret.

• Email Inbox uses domain verification and forwarding setup.
• Intercom uses connection status and optional advanced routing profile controls.
• Zendesk uses OAuth plus webhook URL and webhook secret setup.

You do not need to activate all channels right away. One channel with clean behavior is better than three channels with unclear behavior.

Now run a small internal test. Ask teammates to send realistic messages, including one straightforward question and one sensitive message that should end up with a human. Confirm both paths behave the way your team expects.

When test messages start appearing in Conversations, inspect each thread. You can see classification, confidence, tags, escalation target, and message history. If ownership needs to change mid-thread, use Assign to AI or Assign to Human in the conversation view.

Roll Out Gradually and Learn Fast

After internal checks look stable, move to a limited rollout instead of a full cutover. A calm rollout sequence is still the safest one:

1. Internal traffic only.
2. Small external canary.
3. Broader traffic when behavior is stable.

During this stage, keep your review loop tight. If answers feel generic, go back to knowledge quality. If too many items escalate, review whether routing conditions are too strict or whether runtime context is missing. If sensitive items are not escalating, tighten routing policy immediately.

One thing worth remembering: quickstart is not the end of setup. It is the point where your team proves that the support loop works in your environment. From there, quality comes from iteration.

Teams that move fastest usually follow the same rhythm. They pick one high-volume question type, make it reliable, monitor behavior in conversations, and only then expand scope. They do not chase perfect configuration on day one. They build trust one workflow at a time.

If your first workflow is stable, continue with Core Concepts next. That page explains how Jardine makes decisions across knowledge, routing, ownership, and escalation so your future changes stay intentional.