Skip to main content

How to Configure AI Capabilities

This guide walks you through enabling and configuring built-in capabilities for your AI Employees. Follow these steps to set up capabilities that handle common business tasks like lead capture and appointment booking.

Prerequisites

Before you begin:

  • You have an AI Employee created in your Business App account
  • You have appropriate permissions to configure AI settings
  • You understand which capabilities align with your business needs

Step 1: Access AI Employee Configuration

  1. Navigate to AIWorkforce in your Business App dashboard.
  2. Locate the AI Employee you want to configure.
  3. Click the Configure button next to the employee's name.
tip

If you don't see the Configure button, verify you have the necessary permissions to modify AI Employee settings.

Step 2: Open the Capabilities Section

  1. On the AI Employee's Configure page, scroll down to the Capabilities section.
  2. Click to expand the section if it's collapsed.
  3. You'll see a list of capabilities already added to the AI Employee.

Step 3: Add Desired Capabilities

Capabilities that appear in the Capabilities section are active. To add a new one:

  1. Click + Add a capability.
  2. Review the capability description to understand what it does.
  3. Add specific instructions (optional) to customize how the capability behaves.

To remove a capability, click the three-dot menu next to it and select Remove.

Common capability configurations

Lead Capture

  • What it does: Automatically collects contact information from potential customers
  • Recommended instructions:
    Always ask for name, email, and phone number. 
    Qualify leads by asking about budget and timeline.
    Be friendly but persistent in gathering complete information.

Appointment Booking

  • What it does: Schedules meetings using your connected calendar
  • Prerequisites: Ensure your calendar is connected in your Business App settings
  • Recommended instructions:
    Check availability for the next 2 weeks.
    Confirm all details before booking: date, time, service type, duration.
    Send confirmation details to the customer.

Communication Style

  • What it does: Maintains consistent tone and brand voice
  • Recommended instructions:
    Use a professional but friendly tone.
    Always end responses with "How else can I help you today?"
    Refer to our company as "we" and "our team."

Step 4: Add Goals and Instructions (Optional)

For more precise control over capability behavior:

  1. Click the "Add Instructions" link beneath each enabled capability.
  2. Write clear, specific instructions that define:
    • When to use the capability
    • What information to collect
    • How to respond in different scenarios
    • What to do if something goes wrong

Writing effective instructions

Good example:

Lead Capture: Only collect contact information after the customer shows interest in our services. Always ask for name, email, and phone. If they're hesitant to share information, explain that it helps us provide better service.

Poor example:

Get contact info when needed.
Best Practices for Instructions
  • Be specific about trigger conditions - Tell the AI exactly when to act
  • Include error handling guidance - What should happen if something goes wrong?
  • Use clear, actionable language - Write like you're training a new employee
  • Show examples, not just rules - "For example, say: 'I'd be happy to help with that.'"
  • Keep it concise - Shorter, specific instructions often work better than long paragraphs
  • Test and iterate - Try different phrasings and see what produces better results

Write instructions as directives, not descriptions

Write every instruction as an order addressed to the AI employee. An instruction phrased as a description of what the employee does reads as background information, and the AI may not connect the described behavior to itself.

AvoidUse instead
The receptionist says "I'll text you a booking link."Respond with "I'll text you a booking link."
The assistant will offer to take a message.Offer to take a message.
Bookings are handled by the call center.Tell the caller that bookings are handled by the call center.

This applies anywhere you write instructions: capability prompts, goals, and the instructions attached to individual capabilities.

Back up a prompt before you rewrite it

Capability prompts have no version history. Before making substantial edits, copy the existing prompt into a separate document so you can restore it. Experimenting is otherwise safe — you can adjust a prompt and re-test as many times as you need, and un-assign a capability from the AI employee if it is not behaving the way you want.

Tailor a capability to specific channels

Your AI Employee knows which channel it's responding on, so a capability's instructions can reference the channel by name. This is useful when a task should behave differently depending on where the customer reaches out — for example, collecting fewer details for lead capture over SMS than by email:

When capturing a lead on SMS, ask for name and phone number only, one question at a time, and keep each message short.
When capturing a lead by email, you can ask for name, email, phone, and preferred appointment time in a single reply.

For guidance on channel-specific behavior across the whole AI Employee, see Adjust responses by channel.

Why Examples Matter

Including examples in your instructions helps your AI understand exactly what you want:

Without examples:

Be friendly when greeting customers.

With examples:

Be friendly when greeting customers. For example: "Hi there! Thanks for reaching out. How can I help you today?"

The second version gives the AI a concrete pattern to follow, leading to more consistent results.

Instruction patterns that work

Use these proven patterns when writing capability instructions:

Conditional Logic: "If X, then Y"

If the customer asks about pricing: Share our standard rates and offer to send a detailed quote.
If they mention a competitor: Acknowledge their research and focus on our unique benefits.

Sequential Steps: "First... Then... Finally..."

First, greet the customer warmly.
Then, ask what brings them to our site today.
Finally, based on their answer, offer relevant help or information.

Boundary Setting: "Only... when..." / "Never... unless..."

Only ask for contact information when the customer shows clear buying interest.
Never transfer calls unless the customer specifically requests to speak with someone else.

Error Handling: "If... can't..."

If you can't find the answer in the knowledge base, say: "I don't have that specific information, but I'd be happy to connect you with someone who does."

Step 5: Save and Test Your Configuration

  1. Click Save Changes at the bottom of the Configure page.
  2. Test the capabilities by starting a conversation with your AI Employee.
  3. Try different scenarios to ensure capabilities activate correctly:
    • Ask questions that should trigger knowledge base access
    • Express interest in services to test lead capture
    • Request an appointment to test booking functionality

Testing checklist

  • Capabilities activate at appropriate times
  • Required information is collected before proceeding
  • AI follows your custom instructions
  • Error scenarios are handled gracefully
  • Customer experience feels natural and helpful

Step 6: Monitor and Refine

After initial setup:

  1. Review conversation logs to see how capabilities perform in real interactions.
  2. Collect customer feedback about their experience.
  3. Adjust instructions based on what you learn.
  4. Add or remove capabilities as your business needs evolve.

Troubleshooting common issues

Capability not activating

  • Check trigger conditions: Ensure customer requests match capability parameters
  • Review instructions: Make sure activation criteria are clear
  • Test with different phrasings: Try various ways customers might make requests

Information not being collected

  • Verify required fields: Ensure the capability knows what information to gather
  • Check instruction clarity: Make collection requirements explicit
  • Test conversation flow: Walk through the process as a customer would

Capabilities conflicting

  • Review capability priorities: Some capabilities may override others
  • Clarify instructions: Make trigger conditions more specific
  • Remove conflicting capabilities: Remove capabilities that interfere and re-add them once instructions are refined

The AI has the information but never acts on it

Knowledge and capabilities do different jobs. A Knowledge Base entry is reference material the AI retrieves when it judges a lookup would help — it does not guarantee the AI takes a particular action or uses particular wording. When a question should trigger the same process every time, that content belongs in a capability rather than in knowledge.

To turn a knowledge entry into a capability:

  1. Open the knowledge entry and copy its content.
  2. Go to the Capabilities section and click + Add Capability.
  3. Name the capability after the task it performs, such as Booking workflow.
  4. Paste the content into the prompt and restructure it as a workflow, using ## for each instruction and ### for the steps beneath it.
  5. Rewrite any descriptive lines as directives. See Write instructions as directives, not descriptions.
  6. Save, then test with the phrasings a customer would actually use.

Leave the original knowledge entry in place. Knowledge continues to answer open-ended questions; the capability governs the process.

Advanced configuration tips

Capability layering

Enable multiple complementary capabilities that work together:

  • Lead Capture + Appointment Booking for service businesses
  • Knowledge Base + Communication Style for support scenarios
  • Product Lookup + Lead Capture for sales interactions

Gradual rollout

Start with essential capabilities and add more over time:

  1. Week 1: Enable basic communication style and knowledge base
  2. Week 2: Add lead capture once AI is responding well
  3. Week 3: Include appointment booking after testing lead flow
  4. Week 4+: Add custom capabilities for specialized needs

Performance monitoring

Track key metrics to measure capability effectiveness:

  • Activation rate (how often capabilities trigger)
  • Completion rate (successful information collection)
  • Customer satisfaction scores
  • Conversion rates for lead capture

Next steps

  • Create Custom Capabilities: Learn how to build custom capabilities for specialized business needs
  • Advanced Prompting: Explore prompt engineering techniques for better capability performance
  • Integration Setup: Connect external systems to enhance capability functionality

Need help with specific capability configurations? Check our troubleshooting guide or contact support for personalized assistance.

New to how AI Employees work? Take the AI foundations course in Vendasta Learn — Beginner, 6 lessons.