> ## Documentation Index
> Fetch the complete documentation index at: https://help.agentsy.pro/llms.txt
> Use this file to discover all available pages before exploring further.

# Conversation Experience

> Understand how visitors interact with your Agentsy AI Experience, from their first question to a meaningful next step.

An Agentsy AI Experience gives visitors a way to interact with your expertise before they decide whether to contact you. Instead of only reading a service description, visitors can ask questions, explore options, and understand how you approach their needs.

The goal is not to replace every human conversation. It is to make those conversations more informed and useful.

## The visitor journey

A typical experience can include the following stages. The exact flow depends on your configuration, including whether guest chat is enabled and whether registration is required.

**1. Discover the experience**

A visitor arrives at your branded AI website through a link, campaign, referral, or other channel. Your headline and supporting content explain what they can explore.

**2. Start a conversation**

The visitor asks a question or responds to the assistant's welcome message. The assistant should help with the visitor's immediate intent rather than require a long introductory questionnaire.

**3. Explore relevant expertise**

The assistant explains concepts, guides decisions, or helps the visitor consider options using its instructions and available knowledge sources.

**4. Continue with appropriate access**

Depending on your access settings, the visitor may need to register or sign in to continue. Guest chat, where enabled, can provide a way to try the experience before registration.

**5. Choose a next step**

When the visitor is ready, the assistant can suggest an appropriate human next step, such as using a booking or contact option you have configured.

> [!NOTE]
>
> Registration, guest access, and available contact actions depend on the AI Experience's settings. Do not assume every visitor follows the same sequence.

## What makes a useful conversation?

### Start with the visitor's intent

A visitor might want a quick explanation, help comparing alternatives, or guidance about what to do next. The assistant should respond to the question asked and request additional context only when it will meaningfully improve the answer.

For example, a visitor asking a boat advisor about the difference between a center-console and a cabin cruiser should receive a useful comparison before being asked about budget or preferred brands.

### Share expertise, not just generic answers

A well-designed assistant reflects the business's knowledge, approach, and boundaries. It may use uploaded educational documents, service descriptions, frameworks, or FAQs to provide more relevant responses.

The assistant should distinguish information supported by its sources from information it cannot verify. It must not invent current inventory, appointment availability, prices, policies, or other live facts without an appropriate source.

### Make the conversation easy to follow

Use concise explanations, clear language, and a manageable number of questions. Adapt to the visitor's language and familiarity with the subject when your instructions support this behavior.

Avoid pushing registration, booking, or sales messaging repeatedly when the visitor is still seeking legitimate guidance.

### Know when to involve a person

A human handoff is appropriate when the visitor wants to take action, requests something outside the assistant's capabilities, or needs individualized professional assistance.

For example:

* **Consulting:** A visitor wants to discuss a specific engagement or proposal.
* **Boat advisory:** A visitor asks to see boats currently available for purchase.
* **Insurance education:** A visitor wants personalized coverage advice or to proceed with an application.

The assistant can explain the next step and direct the visitor to a configured contact or scheduling option. It should not claim that a meeting has been booked unless that action has actually been completed.

## Example: From curiosity to a human conversation

Imagine a visitor exploring an AI Experience for a leadership consultant.

**Visitor:** "My team keeps missing deadlines. Where should I start?"

**Assistant:** "A useful first step is to distinguish whether deadlines are slipping because priorities are unclear, work is overloaded, or decisions are getting stuck. Which of those feels most familiar?"

**Visitor:** "Mostly unclear priorities."

**Assistant:** "Then you could start by identifying the team's three most important outcomes for the next two weeks, assigning one owner to each, and reviewing what work can be deferred. Would you like a simple framework for running that discussion?"

Later, if the visitor wants help implementing the approach, the assistant can suggest speaking with the consultant.

This approach gives the visitor a useful experience before asking for a commitment.

## Configure the experience around your goals

Before inviting visitors, review these elements:

| Element | What to check |
| - | - |
| Welcome message | Does it clearly invite visitors to ask a relevant question? |
| System instructions | Do they define the assistant's role, tone, limitations, and handoff behavior? |
| Knowledge | Are the reference materials accurate, relevant, and current? |
| Access settings | Is the intended public, private, or guest experience configured correctly? |
| Registration | Is the transition clear and appropriate for your audience? |
| Calls to action | Do booking or contact links lead to the correct destination? |
| Privacy | Does the assistant avoid requesting unnecessary sensitive information? |
| Mobile usability | Can visitors read responses, scroll, and reach the chat input comfortably? |

## Test the conversation from a visitor's perspective

Try these scenarios before launch:

1. **New visitor:** Ask a simple question without providing background. Does the assistant help immediately?
2. **Unclear request:** Ask a vague question. Does it clarify naturally rather than interrogate?
3. **Knowledge question:** Ask about a topic covered in your documents. Is the answer grounded and useful?
4. **Unknown information:** Ask for a fact the assistant cannot verify. Does it acknowledge the limitation?
5. **High intent:** Ask to speak with the business. Does the assistant guide you to a real next step?
6. **Registration:** If applicable, try the guest-to-registered journey and confirm that navigation behaves as intended.
7. **Mobile:** Test long answers, scrolling, and the on-screen keyboard on a phone.

> [!TIP]
>
> Test with questions real prospects would ask, not only questions written to match your knowledge documents. This reveals where the experience needs better guidance or source material.

## Improve the experience over time

Review the conversation information and engagement insights available in your Agentsy account, subject to your plan and permissions. Look for recurring questions, confusing responses, and moments when visitors request human assistance.

Use those findings to refine your system instructions, update knowledge documents, improve website messaging, and clarify next steps. Treat engagement metrics as signals for improvement, not guarantees of sales or conversions.

## Related guides

* [System Instructions](/create-and-configure/system-instructions)
* [Conversation Settings](/create-and-configure/conversation-settings)
* [Guest Chat and Registration](/create-and-configure/guest-chat-and-registration)
* [Testing Your Assistant](/create-and-configure/testing-your-assistant)
* [Human Handoff](/conversations-and-leads/human-handoff)
* [Conversation Insights](/conversations-and-leads/conversation-insights)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.