Test and deploy your agent

Previewing the agent before customers see it

Preview is the testing ground between building your agent and letting customers talk to it. It lets you step into the customer’s shoes, ask real questions, and see exactly how the agent responds before any change reaches production. This is especially important because a single mistake in a live answer can undermine trust, and Preview gives you a controlled space to catch those issues early.

What Preview does

Preview runs against the same Engine that powers your live agent. Living Knowledge, Memory, Context, and Procedures all behave identically, so what you see in Preview is what customers will experience. The only difference is that Preview sessions do not affect your Analyze metrics for the live agent, so you can test freely without skewing your data.

The feature is currently in Beta and will be available soon. When it is live, you will find it under “Test > Preview”.

Key concepts

Preview is built around three main elements:

• Conditions - The setup that shapes the session: Audience, topic, Living Context state, and surface (marketing site, product, helpdesk). These are the same conditions a real customer would arrive with. You can use full URLs with domain names when previewing components.
• Conversation panel - Where you type messages and see the agent respond. This is identical to what a customer sees.
• Inspection panel - A side panel that shows what the agent is seeing internally: Sources used, Procedures available, and Living Context values. Use this to diagnose unexpected answers.

Every Preview session is saved automatically in a session log. You can share a link with a colleague or come back later to review.

What you can do in Preview

• Type questions and watch the agent respond
• Trigger Procedures to see them run end to end
• Switch Audience, topic, or surface to test different paths
• Inspect which Sources the agent used
• Share session URLs with colleagues for review

When to use Preview

Use Preview in these situations:

• After updating a Source, to confirm the change shows up correctly
• Before publishing a new component or Moment configuration
• Before any pilot review meeting, as a quick sanity check
• When the agent is giving a confusing answer in production and you want to reproduce the issue

When not to use Preview

Preview is designed for one human, one session. It is not the right tool for:

• Testing at scale - Use Simulations instead.
• Scoring the agent against expected answers - Use Quality reports.

How to start a Preview session

Open “Test > Preview”. Set the Audience, topic, Living Context, and surface. Then click “Start Preview”.

To test how the agent treats a specific customer, set Living Context to that customer’s profile. The agent will then see the same plan, history, and signals it would see in production.

To check which Sources the agent used, open the Inspection panel next to the chat. Every answer lists the Sources it pulled from.

To share a session, copy the session URL from the address bar. Anyone with access to your Unless account can open it.

Why Preview and production answers differ

If the agent answers differently in Preview than in production, the cause is usually a difference in conditions. Check the Audience, topic, and Living Context first. If those match production, the next thing to check is whether the live Sources have changed since you set up the Preview.

Conclusion

Preview is your safety net before every deployment. It lets you verify answers, Procedures, and personalization settings in a controlled space, so you can ship changes with confidence. When you are ready to test at scale or score performance, move to Simulations and Quality reports.