AI Voice Agent Build Guide
Published 15 June 2026
Purpose
Building an AI Voice Agent is similar to training a new receptionist. The AI doesn't automatically know how your business operates, what services you provide, or how you would like calls handled.
This guide explains the information we need to configure your AI Voice Agent and why each requirement is important.
The more detail you provide, the more accurately your AI Agent can answer questions, guide callers, and automate common tasks.
Our goal is to build an AI Voice Agent that sounds natural, represents your business professionally, and follows your preferred call handling process.
Your Checklist and Guide
7. Detailed Workflow for Each Intent
10. Escalation Rules
12. Compliance, Privacy, and Guardrails
1. Tell us about your business
Why do we ask for this?
Before your AI Agent can answer questions, it needs to understand your business.
Think of this as introducing a new employee to your company.
A good business overview helps the AI understand:
- what your business does
- the products or services you provide
- who your customers are
- the types of enquiries you receive
- what information is considered relevant
Without this context, the AI may struggle to determine whether a caller is asking about one of your products, another service, or something outside your business.
Example:
Instead of saying:
"We sell telecommunications."
Provide something like:
"Untangled provides business telecommunications solutions including AI Voice Agents, cloud phone systems, contact centres, business internet, mobile plans and managed communications."
The more context you provide, the better the AI can stay on topic.
2. Let's create your AI Agent profile
Why do we ask for this?
Your AI Agent represents your business.
Just like hiring a receptionist, you can choose how your AI sounds and communicates.
This section defines:
- the AI's name
- voice (male or female)
- accent
- personality
- communication style
These settings help create a consistent experience for every caller. So although they may be speaking to an AI Agent it sounds like they are speaking to someone from your team.
Every industry has a different communication style, and your AI should reflect your brand.
3. When should your AI Agent answer or make calls
Why do we ask for this?
Your AI needs to know when your business is open.
This determines how calls are handled during business hours and after hours.
Example:
During business hours the AI might:
- transfer calls
- book appointments
- connect callers with staff
After hours it might:
- take a message
- schedule a callback
- send an SMS
- book an appointment for the next business day
This ensures callers always receive the appropriate experience.
4. What would you like your AI Agent to do?
Why do we ask for this?
Not every AI Voice Agent has the same purpose.
Some businesses simply want an answering service.
Others want a virtual receptionist that answers questions, books appointments, qualifies leads, or provides technical support.
By understanding your goals, we can design conversations that achieve the outcomes you want.
Examples include:
- Answer common questions
- Capture new leads
- Book appointments
- Provide technical support
- Transfer cals
- Process callback requests
- Create support tickets
- Route callers to the correct depoartment
The clearer your objectives, the more effective your AI Agent will be.
5. What should your AI Agent say when opening a phone call?
Why do we ask for this?
The first few seconds of every call create the first impression of your business.
The greeting sets the tone and helps callers understand they've reached the right place.
Some businesses prefer a simple greeting, others prefer the AI to immediately ask how it can help.
Example:
'Hello, thanks for calling {Company Name}, this is {Agent Name}, your AI Agent. How can I help you today?'
6. Caller Intent - Why do people usually call your business?
Why do we ask for this?
One of the AI's first jobs is to understand why someone is calling.
We call these Caller Intents.
Think of these as your main menu options, or different departments/ reasons people contact your business.
Examples:
-
Sales
- Support
- Billing
- Appointments
Once the AI Agent identifies the caller's intent, it can guide the conversation down the correct workflow.
Without clearly defined intents, conversations become less accurate and may require unnecessary transfers.
7. Workflow for each type of call intent
Why do we ask for this?
After the AI knows why someone is calling, it needs to know what should happen next.
This is called a workflow.
A workflow is simply the step-by-step conversation the AI follows.
For EACH intent, please provide:
-
what customers usually ask
- what questions the AI should ask
- what information should be collected
- what actions should be performed
- when the AI Agent should transfer or escalate
Example:
For a Sales workflow
Customer asks about pricing
↓
AI explains products
↓
Collects name and phone number
↓
Books appointment
↓
Emails sales team
Every workflow becomes the "playbook" the AI follows during calls.
8. What actions should your AI Agent be able to perform?
What actions are allowed to be performed by your AI Agent during and after a phone call?
Why do we ask for this?
Knowing what to say is only part of the conversation.
Examples include:
☐ Transfer calls
☐ Book appointments
☐ Send emails
☐ Send SMS
☐ Create CRM records
☐ Create support tickets
☐ Log notes
☐ Arrange callbacks
If yes, please tell us:
Who should receive the email?
Who should receive transfers?
Which calendars should be used?
Which phone numbers should be called?
Every action must be clearly defined so the AI behaves consistently.
9. Knowledge Base in FAQ format
Think of the Knowledge Base as everything you would teach a new receptionist on their first day.
It contains the information your AI Voice Agent uses to answer questions accurately and consistently.
The more information you provide, the smarter and more helpful your AI Agent becomes.
Why do we recommend using a Question & Answer (FAQ) format?
Although your AI Agent can learn from web pages, PDFs, brochures, and other documents, we recommend providing your key information in a simple Question & Answer (FAQ) format wherever possible.
This makes it easier for the AI to understand exactly what customers are asking and provides more accurate, natural, and consistent responses. It also helps the AI Agent scrape data from prompts faster.
Example FAQs:
Question:
Do you service rural areas?
Answer:
Yes. We service metropolitan and regional areas across New South Wales.
Question:
Do you install solar batteries?
Answer:
Yes. We supply and install a range of residential and commercial battery storage solutions across Australia.
FAQs generally produce the most natural and accurate responses because they closely match how customers ask questions during a phone call.
10. Escalation Rules
When should the AI Agent stop trying to help and involve a human person.
Why do we ask for this?
Not every call should be handled by an AI Agent. Some situations require a human.
Escalation Rules tell the AI exactly when to stop assisting and hand the conversation to your team.
Examples:
- Customer requests a human
- Issue cannot be resolved
- Legal or compliance issue
- Billing dispute
- Technical issue beyond basic troubleshooting
For each scenario, provide information on:
Who should receive the escalation, and as how?
- Warm transfer with context to:
- Cold transfer to:
- Send a Request - Callback email to:
Clear escalation rules help avoid frustration and ensure callers receive the right level of support.
11. Data Capture Requirements
The information that the AI Agent should collect and capture for each workflow, specify:
- Mandatory information
- Optional information
Why do we ask for this?
Often the AI needs to collect information before transferring a call or creating a callback request. Tell us exactly what information should be collected for each workflow.
Examples:
- For a Sales Workflow
- Firstname (mandatory)
- Company name (mandatory)
- Mobile number (mandatory)
- Email address (mandatory)
- Product interest (mandatory)
- Summary of enquiry (mandatory)
- For a Support Workflow
- Firstname (mandatory)
- Company name (optional)
- Mobile (mandatory)
- Product with problem: (optional)
- Issue summary (mandatory)
Only collect information that is genuinely useful. Asking unnecessary questions can make conversations longer than they need to be.
12. Compliance & Guardrails
Why do we ask for this?
Some businesses have legal or privacy requirements. Ensures that the AI Agent works within what it can support and how to control the conversation to ensure callers do not use the AI Agent for gaining or providing incorrect information during the call.
Examples:
-
Call recording notice
-
Privacy statement
-
Identity verification
-
Information that must never be discussed
-
Anything that requires a staff member
Providing these requirements ensures your AI handles customer information appropriately and complies with your business obligations.
13. CRM Integrations
CRM or ERP tools that the AI Agent can access as part of an action/ outcome in their workflow, or to update during a phone call.
Examples:
-
CRM (Salesforce, Hubspot, ZenDesk, etc.)
- Calendar (Google, Outlook)
14. How should your AI Agent close the call?
Why do we ask for this?
The final impression is just as important as the first. This is what the agent should do and say before closing the call.
Some businesses prefer a simple farewell. Others like to remind callers about next steps before ending the call.
Examples:
-
Ask if there is anything else that the caller needs help with?
- "Thanks for calling {Company Name}. Have a wonderful day."
A professional closing ensures every conversation ends naturally and consistently.
🚨Final Tip
There is no such thing as providing too much information.
When in doubt, include it.
If your staff regularly answer a question, follow a particular process, or refer to a document, there's a good chance your AI Voice Agent should know about it too.
The more knowledge, workflows, and business context you provide, the more accurately your AI Voice Agent can represent your business and deliver an exceptional experience for your customers.
Have other questions?
Call or text our support team on 0480 006 410, alternatively you can raise a ticket to support@untangled.com.au