Test Your First Conversation

Test Your First Conversation

You have configured your AI agent, uploaded your knowledge base, and embedded the widget. Now it is time to run a real conversation and verify that leads are captured, scored, and displayed correctly in the dashboard. This guide walks you through a structured test so you can go live with confidence.

Before You Start

Make sure you have completed these steps:

• [ ] Created your Synaptiq tenant and configured basic settings (Quick Start Guide)
• [ ] Embedded the widget on your site or have access to the Preview mode (Embed the Widget)
• [ ] Configured your AI agent's greeting, personality, and qualification criteria (Configure Your AI Agent)
• [ ] Uploaded at least one document to your knowledge base (Upload Your Knowledge Base)

You do not need all of these to run a test, but the more you have configured, the more realistic the test will be.

Option 1: Test via Preview Mode

The fastest way to test is the built-in preview, which does not require the widget to be embedded on a live site.

1. Go to /admin and click AI Agent > Preview in the left sidebar
2. A simulated widget appears on the right side of the screen
3. The Debug Panel on the left shows real-time data extraction and scoring as the conversation progresses
4. Start chatting with the agent as if you were a potential customer

Advantages of Preview Mode:

• No need to deploy the widget to test changes
• The Debug Panel shows field extraction, confidence scores, and knowledge base references in real time
• Conversations from Preview mode are tagged as "Test" in the dashboard so they do not pollute your production data

Limitations:

• Does not test the actual script tag embed or page-specific context
• Does not test the JavaScript API or event listeners
• Visual styling may differ slightly from the production widget

Option 2: Test on Your Live Site

For a production-realistic test, use the widget as an actual visitor would:

1. Open your website in a browser (use an incognito window to avoid cached sessions)
2. Click the chat bubble to open the widget
3. Have a full conversation with the agent
4. When finished, go to the Synaptiq dashboard to review results

Tip: If you want to test multiple conversations without them merging into one session, open a new incognito window for each test. Synaptiq identifies returning visitors by browser fingerprint and session cookies.

Run the Standard Test Script

Use this script to exercise every important behavior of your AI agent. Play the role of a potential customer and follow these steps:

Phase 1: Initial Engagement

You say: "Hi, I'm interested in learning more about what you offer."

What to verify:

• The agent responds with a message that aligns with your configured greeting style
• The tone matches the personality setting you chose (Professional, Friendly, Casual, or Technical)
• The response is the right length (Concise, Balanced, or Detailed — per your setting)

Phase 2: Product Questions (Knowledge Base Test)

You say: "Can you tell me about your pricing?"

What to verify:

• The agent's response references actual information from your uploaded documents
• Pricing details are accurate and specific, not generic
• If you check the Debug Panel (Preview mode), you should see knowledge base chunks being retrieved with relevance scores above 0.6

Follow up with: "What integrations do you support?"

What to verify:

• The agent pulls from a different section of your knowledge base
• The answer is specific to your product, not a generic list of common integrations

Phase 3: Qualification Data Collection

Now let the agent guide you through qualification. Respond naturally when it asks questions. Provide this test data:

| Field | Test Value | |-------|-----------| | Name | Alex Johnson | | Email | alex.johnson@testcorp.com | | Company | TestCorp | | Job Title | VP of Sales | | Company Size | 200 employees | | Budget | $5,000-10,000/month | | Timeline | Within 30 days | | Use Case | Automating inbound lead qualification |

What to verify:

• The agent asks for each required qualification field naturally, woven into conversation
• It does not ask for everything at once like a form
• It acknowledges your answers before moving to the next question
• The Debug Panel shows fields being extracted in real time (Preview mode)

Phase 4: Edge Case — Reluctant Visitor

Start a new conversation and this time, push back:

You say: "I'm just browsing, not ready to share any info yet."

What to verify:

• The agent respects the pushback and does not aggressively demand information
• It pivots to providing value (answering questions, sharing useful content) rather than interrogating
• After providing value, it gently circles back to collect information

Then say: "Actually, can you tell me about your security practices?"

What to verify:

• If you uploaded security documentation, the agent provides a detailed, accurate response
• If you did not upload security docs, the agent triggers its fallback behavior (acknowledge and redirect, per your configuration)

Phase 5: Off-Topic and Competitor Questions

You say: "How do you compare to [a competitor in your space]?"

What to verify:

• The agent follows your custom instructions (e.g., does not badmouth competitors)
• It focuses on your product's strengths rather than attacking the competitor
• It uses knowledge base content to back up claims

Then say: "What's the weather like today?"

What to verify:

• The agent gracefully redirects back to the topic without being rude
• It does not attempt to answer irrelevant questions

Phase 6: Conversation Completion

You say: "This sounds great, what's the next step?"

What to verify:

• If you have calendar integration enabled, the agent offers to book a meeting
• If not, the agent provides a clear next step (e.g., "Our team will follow up via email within 24 hours")
• The agent thanks the visitor and provides a way to come back if they have more questions

Check the Lead in Your Dashboard

After completing the test conversation, go to the Synaptiq dashboard to verify the data was captured correctly.

Leads View

1. Navigate to /admin/leads
2. Your test lead should appear at the top of the list (most recent first)
3. Verify the following on the lead card:

| Field | Expected | |-------|----------| | Name | Alex Johnson | | Email | alex.johnson@testcorp.com | | Company | TestCorp | | Status | Qualified (if all required fields were collected) | | Score | A number between 0-100, reflecting the fields collected and their weights |

If the score seems too low or too high, review your weight configuration under AI Agent > Qualification > Scoring Rules.

Conversation Transcript

1. Click into the lead to open the detail view
2. Click the Conversation tab
3. You should see the full transcript with:
   • Every message from both the visitor and the agent
   • Timestamps for each message
   • A sidebar showing which fields were extracted and at which point in the conversation
   • Knowledge base citations — which documents were referenced in each agent response

Qualification Timeline

The Qualification tab on the lead detail page shows a timeline of when each field was captured:

[0:00] Conversation started
[0:12] Greeting sent
[0:34] Visitor asked about pricing
[0:35] Knowledge base referenced: pricing-guide.pdf (relevance: 0.89)
[1:02] Field extracted: name = "Alex Johnson" (confidence: 0.97)
[1:15] Field extracted: email = "alex.johnson@testcorp.com" (confidence: 0.99)
[1:28] Field extracted: company = "TestCorp" (confidence: 0.95)
[2:10] Field extracted: company_size = "200" (confidence: 0.91)
[2:45] Field extracted: budget = "$5,000-10,000/month" (confidence: 0.88)
[3:02] Field extracted: timeline = "Within 30 days" (confidence: 0.93)
[3:15] Lead qualified — Score: 87
[3:16] Notification sent: email, Slack

This timeline is invaluable for understanding how the agent collects information and where the conversation could be improved.

Verify Notifications

If you configured notifications in the Quick Start Guide, verify they fire correctly:

Email Notifications

• Check the inbox for the email address configured under Settings > Notifications
• The email should contain the lead's name, email, company, qualification score, and a link to the full conversation in the dashboard

Slack Notifications

• Check the Slack channel you configured under Integrations > Slack
• The notification should include a summary card with the lead's info and a link to the dashboard

Webhook Notifications

• If you configured a webhook URL, check your endpoint for the incoming POST request
• The payload follows this structure:

{
  "event": "lead.qualified",
  "timestamp": "2026-04-05T14:32:00Z",
  "lead": {
    "id": "lead_abc123",
    "name": "Alex Johnson",
    "email": "alex.johnson@testcorp.com",
    "company": "TestCorp",
    "score": 87,
    "fields": {
      "job_title": "VP of Sales",
      "company_size": "200",
      "budget": "$5,000-10,000/month",
      "timeline": "Within 30 days",
      "use_case": "Automating inbound lead qualification"
    },
    "conversation_url": "https://synaptiqintel.com/admin/conversations/conv_xyz789"
  }
}

Tip: If you do not have a webhook endpoint set up, use a free service like webhook.site to capture and inspect the payload during testing.

Common Issues and Fixes

Lead appears but score is 0

The agent collected fields but they were not mapped to your qualification criteria. Go to AI Agent > Qualification and verify that your field keys match. For example, if the agent extracts "company_name" but your field key is "company," the data will not map.

Agent does not ask for qualification data

Check that your qualification fields are set to Required under AI Agent > Qualification. If all fields are optional, the agent will prioritize answering questions over collecting data.

Knowledge base answers are wrong or outdated

Go to Knowledge Base > Documents and check the Last Updated date on relevant files. If a document is outdated, re-upload the current version. Use the Knowledge Base Test tool to verify the correct chunks are being retrieved.

Conversation does not appear in the dashboard

• Preview mode: Test conversations appear under /admin/conversations with a "Test" tag. Make sure you are not filtering them out.
• Live widget: Verify the data-tenant attribute in your script tag matches your active tenant ID. Check the browser console for errors.

Agent sounds too robotic or too casual

Adjust the tone under AI Agent > Personality. If the tone preset is not enough, use the Custom Instructions field to give specific guidance. For example: "Use contractions (you're, we'll, it's). Avoid corporate jargon. Keep sentences short."

Interpreting the Qualification Score

The score is not just a number — it maps to lead quality tiers that you can customize:

| Score Range | Default Tier | What It Means | Typical Action | |-------------|-------------|---------------|----------------| | 80-100 | Hot Lead | All required fields collected, strong buying signals (budget fits, short timeline) | Immediate sales follow-up or auto-book a meeting | | 50-79 | Warm Lead | Most fields collected, but missing budget or timeline | Add to nurture sequence, follow up within 48 hours | | 25-49 | Cool Lead | Basic contact info collected, no strong buying signals | Add to marketing list, send educational content | | 0-24 | Cold Lead | Minimal info collected, visitor disengaged early | Log for reference, no active follow-up |

You can customize these tiers and their associated automation triggers under AI Agent > Qualification > Scoring Rules.

Run a Full End-to-End Checklist

Before going live with real visitors, run through this complete checklist:

• [ ] Widget loads on your production site without console errors
• [ ] Greeting message displays correctly and matches your brand voice
• [ ] Agent answers product questions accurately using knowledge base content
• [ ] Agent collects all required qualification fields naturally
• [ ] Lead appears in the dashboard with correct data and score
• [ ] Conversation transcript is complete and readable
• [ ] Email notifications fire when a lead is qualified
• [ ] Slack notifications post to the correct channel (if configured)
• [ ] Webhook payloads arrive at your endpoint (if configured)
• [ ] Agent handles pushback gracefully without being aggressive
• [ ] Agent redirects off-topic questions back to relevant subjects
• [ ] Fallback behavior works when the agent does not know an answer
• [ ] Calendar booking flow works (if integrated)
• [ ] Widget respects your position, theme, and color settings
• [ ] Widget works on mobile devices (test on your phone)

Next Steps

Your Synaptiq setup is validated and ready for real visitors. Here is what to do next:

• Connect Your CRM — Automatically push qualified leads to HubSpot
• Set Up Calendar Booking — Let the agent book meetings directly into your team's calendar
• Configure Webhooks — Build custom workflows with Zapier, Make, or your own backend
• Explore the API — Programmatically manage leads, conversations, and agent configuration
• Review Analytics — Monitor conversation volume, qualification rates, and agent performance over time

You are live. Start driving traffic to your site and watch the leads come in.