Quick Start Guide

This guide walks you through the fastest path to a working AI agent, with clear checkpoints so you can go live with confidence.

1. Confirm your agent and channels

If you are new to Applied, your Applied Labs team will provision your workspace and your first agent. Open Agents to review:

The agent name and general instructions.
Which channels are enabled (chat, email, SMS, or phone).
The default escalation behavior.

If you support multiple channels, each agent can be configured with its own tone, workflows, and escalation settings.

2. Add knowledge the agent can trust

Open Connectors -> Knowledge and add Responses. Each Response can be:

Draft while you test.
Live once you are ready to use it in production.

Tag responses with a topic and intent so you can measure performance later. If you already have docs or help center content, connect Content Sources so those materials are available to the agent.

3. Build workflows (optional)

Use Agents -> Flows when the agent needs to take actions or collect structured details. Examples include:

Ask for an order number, then look up the order in a connector.
Verify eligibility, then send a confirmation message.
Escalate to a human if a specific condition is met.

Make a Flow Active only when you are ready for it to run in live conversations.

4. Test before you go live

Open Test Coverage to create scenarios and confirm the agent behaves as expected. You can:

Organize scenarios by topic, intent, or flow path.
Group scenarios into Benchmarks for repeatable tests.
Import scenarios from CSV for faster setup.

Use pass/fail outcomes to identify gaps before deployment.

5. Deploy safely

When you are ready, create a revision in Agents -> Deploy and choose whether to make it live immediately. Revisions let you:

Track what changed over time.
Roll back quickly if needed.
Test Draft content without affecting production.

After deployment, monitor the Inbox for live conversations and check Analytics for early quality signals.

If you have a developer, the web chat widget can be embedded with two lines:


<script src="https://cdn.applied.guide/artifacts/applied-chat.umd.js"></script>
<applied-chat agent-id="YOUR_AGENT_ID" />

Common options


<applied-chat
  agent-id="YOUR_AGENT_ID"
  user-id="CUSTOMER_ID"
  theme="light"
  size="popup"
  metadata='{"context": {"name": "Alex", "email": "alex@example.com", "accountTier": "pro"}}'
/>

Use user-id and metadata to link conversations to the right customer.
Use size="full" when you want the widget to fill a container.

The metadata attribute accepts a JSON object with these top-level fields:

context — contact attributes (name, email, and any custom fields you want the agent to see about the user). Custom contact fields must be nested inside context, not placed at the top level of metadata.
groups — array of contact group names.

Update metadata after load


window.applied.setMetadata({
  context: {
    name: 'Alex',
    email: 'alex@example.com',
    accountTier: 'pro',
  },
});

You can control visibility with window.applied.show() and window.applied.hide() or open the widget programmatically with window.applied.setOpen(true).

Need help?

Contact support@appliedlabs.ai and we can help you set up channels, integrations, and best practices.