Testing

Testing isn’t just about catching typos — it’s about proving your AI Agent can think, qualify, and convert like your best salesperson.

Why Test Your Agent?

Most people build their AI Agent but never experience it from the perspective of a real lead. By asking the same questions your visitors would, you can check whether your Agent:

Understands your offerings clearly
Handles personalized scenarios naturally
Communicates your unique value
Responds confidently to pricing questions
Guides uncertain leads toward the next step
Recovers gracefully from random or unexpected inputs

Step 1: Open the Agent Test Mode

Log into your MagicBlocks workspace.

Go to Agents. Select the Agent you want to test. You can test your AI Agent right away under the Try My Agent tab.

While you’re making adjustments to your AI Agent, you can also test it right away by clicking Test in the top-right corner.

Watch the video tutorial here:

Step 2: Quick Test - Ask Key Questions

Chat with your Agent as though you are a potential customer.

Scenario	Sample Prompt	What You’re Testing
Product clarity	“Tell me more about your services.”	Tests understanding of your offerings.
Personalization	“Do you work with small businesses?”	Tests adaptability and segmentation.
Value differentiation	“What makes [your company] different?”	Ensures unique brand positioning.
Pricing	“What’s the cost?”	Tests accuracy and confidence.
Objection handling	“I’m not sure this is for me.”	Measures empathy and recovery.
CTA flow	“I’m ready to book.”	Ensures lead-to-conversion flow.
Curveball	“What’s your favorite movie?”	Tests how well your Agent redirects conversation.

Testing Your Agent for Accuracy (Knowledge Base QA)

Your AI’s intelligence depends on the accuracy and structure of its Brain (Knowledge Base).

Hallucination Prevention Guide

Follow these best practices to stop your AI from making things up:

Keep Knowledge entries short and factual preferably in Q&A format
Add a Priority Knowledge in the Brain to make sure you highlight the important information
Add a Guardrails rule to avoid AI to make things up

Test factual accuracy using questions like:

“What is your pricing?”
“Where is your company located?”
“Who is this product for?”
“What integrations do you support?”

Review Sessions to Improve Conversion (Testing CTA)

Once your Agent is functional, test how well it moves users to action.

Goals to check:

Does the Agent clearly present CTAs (“Book a Call”, “Get a Quote”)?
Are CTAs triggered in the right context (not too early or too late)?
Does it hand off smoothly to calendar or CRM?

Example Test Prompt:

“I’m interested — how can I get started?”

✅ Expected: The Agent either opens the booking calendar or routes to “Get in Touch.”
❌ Fails if it loops back to small talk or repeats info.

Debugging Logic: Missed Key Facts, Wrong Path, Stuck Loops

Logic testing ensures your Journeys, Conditions, and Actions work properly.

Common Issues

Problem	Cause	Fix
Agent stuck in one Block	Missing “Next Block” Action	Add conditional transitions.
Agent loops same question	Key Fact not recognized	Verify Key Fact mapping.
Wrong Journey triggered	Misconfigured Conditions	Check value-matching (“Equals”, “Contains”, etc.)
No response after user input	Action missing “Auto” fallback	Add default or fallback Action.

Key Facts Collection QA

Ask the Agent for the same fact in 3–4 phrasings:

“I want to refinance my home.”
“Looking for a refinance option.”
“Can you help me lower my loan?”

Confirm it correctly identifies Loan Purpose = Refinance.

QA Test Cases & Scorecard

Build a consistent testing habit by tracking results with a simple scorecard.

Test Type	Goal	Result (1–5)	Notes
Greeting & Hook	Natural, on-brand intro
Personalization	Adapts tone based on user input
Key Facts	Collects facts accurately
CTA Flow	Moves to next logical step
Knowledge Accuracy	Gives factual answers
CRM Handover	Sends correct lead info
Tone Consistency	Matches Persona voice
Recovery	Handles off-topic input

Persona Voice Consistency Test

Check whether your AI keeps a consistent tone across Journeys.

Checklist:

Does it always sound like the Persona you defined?
Does it aligned with your brand's voice?
Is the tone formal, friendly, or consultative — and consistent?
Does it avoid switching tone mid-conversation (e.g., polite → robotic)?
Use Try My Agent with these edge cases:
“That’s too expensive.” → Should respond empathetically.
“You sound like a bot.” → Should respond humanly and confidently.

If tone shifts unexpectedly, review your Persona in the Agent's General settings.

Booking / Calendar Flow QA

If your Agent integrates with Calendly, HubSpot Meetings, or MagicBlocks Bookings, test the flow end-to-end.

Checklist:

Calendar opens correctly after “Book a Call.”
Time zone displays accurately.
Form submissions sync with your calendar.
Agent confirms the booking politely (“Got it! You’re all set.”).
After booking, it does not continue pitching.

Testing CRM Handover & API Triggers

CRM testing ensures your automation fires correctly after lead qualification.

How to test:

In Test Mode, complete a full qualification journey.
Check your CRM (e.g., HubSpot, GoHighLevel) for the new record.
Verify fields (name, email, lead type, source) are correctly mapped.

Step 3: Review and Tweak

After testing, adjust your Agent to improve its performance:

Spot the leak using Robot Head Icon in your test chat.
Update knowledge gaps in Brain.
Refine answers inside Journey for better tone and clarity.
Add rules with Guardrails if you want to control certain responses.

Step 4: Publish and Go Live

Once you’re satisfied with all test results:

Save your changes.
Hit Publish.
Embed your Agent or connect it to your site, app, or CRM.

Troubleshooting & Refining Your AI Agent

Knowledge: Feed Information to Your Agent

Setting Smart Automation with Actions

Understanding Journey Blocks

Persona: Define Your AI’s Voice