Why Proper Onboarding Determines Your ROI
There are two ways to go live with an AI receptionist. The first is to spend an afternoon configuring it properly — building a thorough knowledge base, testing edge cases, connecting your CRM, and launching with deliberate controls. The second is to rush through setup in twenty minutes, skip the testing phase, and just point your live number at the new system.
The outcome gap between these two approaches is enormous. Businesses that follow a structured onboarding process see their AI receptionist handling 80 percent of routine enquiries correctly within the first week. Businesses that skip onboarding often spend the first month firefighting — caller complaints, missed bookings, awkward escalations, and eroded trust in the technology.
The good news is that proper onboarding does not require days of effort. The full process — from blank account to live, tested AI receptionist — takes most businesses a single afternoon. The investment is front-loaded, but the payoff starts from the very first call.
Businesses that complete all seven onboarding steps before going live report three times faster time-to-positive-ROI compared to those who configure in stages after launch. The knowledge base is the single highest-leverage step — businesses that invest 60 or more minutes building it report 40 percent fewer escalation calls in the first month.
This guide walks you through every step in order. Each section includes exactly what to do, why it matters, and what to watch for. By the end, you will have a fully configured AI receptionist that handles calls professionally, books appointments accurately, feeds your CRM automatically, and improves with every conversation.
Pre-Setup Checklist: What to Prepare Before You Configure
Before you open the dashboard and start configuring, gather the following information. Having it all in one place means you can move through setup without stopping to look things up. Most businesses find it useful to create a simple document — even a notes app entry — with the answers to each item before starting.
Business information
- Your business trading name (as you would say it on a phone call, not the ABN name)
- Your physical address and service area, if relevant
- Your primary phone number and any secondary lines
- Your website URL
- Your business ABN and industry category
Operating hours and coverage
- Your regular trading hours, day by day
- Public holiday hours or closures
- Whether you want the AI to handle calls 24/7 or only outside business hours
- What should happen to calls that come in at 2am on a Sunday (take a message? book a callback? provide emergency contact?)
Services and pricing policy
- A list of every service you offer, with a one-sentence description of each
- Whether your pricing is fixed, quoted on inspection, or varies by job
- Whether you want the AI to provide pricing on call or direct callers to your website
- Any services you do not offer (so the AI can clearly redirect enquiries)
- Your service area or delivery regions if applicable
Booking and scheduling rules
- How you prefer bookings to be made (calendar link, phone confirmation, email?)
- Minimum notice required for appointments
- How far in advance you accept bookings
- Buffer time between appointments
- Whether you offer same-day or emergency bookings and the surcharge if applicable
Common questions callers ask
- Your top 10 to 15 most frequently asked questions, with your preferred answers
- Any misconceptions callers commonly have about your services
- Warranty, guarantee, or refund policy questions
- Qualification and licensing questions (if relevant to your industry)
Escalation and emergency contacts
- The phone number of a human to transfer urgent calls to
- What qualifies as urgent enough to escalate immediately
- The email address for after-hours message delivery
- An SMS number for immediate lead alerts
Pull your Google Business Profile, your website FAQ page, and any quote templates you use. These three sources contain 80 percent of the information your AI needs. Copying and pasting from them is much faster than writing from scratch.
Account Setup and Initial Configuration
Navigate to talkingwidget.ai/onboarding and sign up with your business email. You will be prompted to select a plan — if you are unsure, start with the Starter plan and upgrade once you have validated call volume. Your plan can be changed at any time from your account dashboard without affecting your configuration.
Enter your business name, industry, trading hours, and service area. This information is used to configure the AI's baseline understanding of your context and to generate your default greeting. Be precise about your industry — a dental practice and a general health clinic have different call patterns, and the industry selection shapes the AI's initial response styles.
You have two options: provision a new Australian local number through the platform, or port your existing business number. Porting an existing number is recommended if it is printed on vehicles, signage, or directories — porting takes one to three business days and your current line continues to function throughout. For immediate testing, provision a temporary number while the port processes.
When you submit a port request, keep your existing line active until port confirmation is received. Do not cancel the source service beforehand — this can invalidate the port. Porting typically completes within two business days for Australian mobile numbers and three to five business days for landlines.
Building Your Knowledge Base — Teaching Your AI About Your Business
The knowledge base is the most important configuration step. It is the AI's reference document — the source it consults to answer caller questions, handle objections, and manage the scope of what it will and will not commit to on your behalf. A thorough knowledge base means fewer escalations, more accurate responses, and callers who feel genuinely helped.
Think of this step as writing the perfect briefing notes for a new receptionist on their first day. You want them to know everything they need to represent your business correctly — without needing to call you every five minutes for clarification.
Core knowledge sections to complete
Services catalogue
List every service with a short description. Include what the service involves, typical duration, and any prerequisites or exclusions.
Pricing and quoting policy
Define what you will say about price on a call. Even "we provide a personalised quote after a site visit" is a complete answer — the AI needs to know this.
Booking process
Step-by-step: what happens when someone wants to book? Do they get a calendar link? Do you confirm by SMS? Is a deposit required?
FAQ answers
Your top 15 questions with your preferred verbatim answers. Include questions you wish callers would ask, not just the ones they do.
Out-of-scope handling
What you do NOT do — and how you would prefer the AI to handle those enquiries. A polite redirect is far better than a confused response.
Emergency protocol
What constitutes an emergency for your business? Who should the AI escalate to, and what information must it collect before doing so?
Writing effective FAQ entries
Each FAQ entry in your knowledge base should follow this structure: question as the caller would ask it, followed by your preferred response in natural, conversational language. Do not write formal policy documents — write how you would actually speak on a call.
Include variations of common questions. Callers do not always phrase things the same way. "How much does it cost?" and "What are your rates?" are the same question — include both in your FAQ to ensure the AI recognises both phrasings.
Do not copy and paste large blocks of terms and conditions or formal policy text into your knowledge base. The AI will struggle to extract clean answers from dense legalese. Summarise policies into plain-language FAQs instead. If legal detail is required, have the AI direct callers to your website or offer to email them the relevant document.
Tone and brand guidelines
In the tone settings section, describe how you want the AI to come across. Is your business warm and conversational, or precise and professional? Do you use industry terms with callers or deliberately avoid jargon? Add two to three example phrases that represent your preferred communication style. These examples shape the AI's word choices throughout every conversation.
A well-built knowledge base for a small business typically has 15 to 25 FAQ entries, a complete services catalogue with 5 to 15 services, a clear pricing policy, a defined booking process, and an emergency escalation rule. If your knowledge base is shorter than this, there are likely gaps that will surface as unexpected escalations in the first week.
Voice Persona Customisation — Choosing the Right Voice, Name, and Greeting Style
Your AI receptionist's voice persona is what callers experience directly. It encompasses the voice itself, the name the AI uses, the opening greeting, the conversational style, and the way it closes calls. Getting this right matters — callers form an impression of your business within the first eight seconds of a call.
Selecting a voice
The platform offers a range of Australian English voice profiles. Listen to each option in the preview player — do not just read the label. The voice that sounds best in isolation may not be the one that sounds most natural for your specific business context. A relaxed, warm voice may suit a wellness practice; a clear, efficient voice suits a legal firm.
Key criteria when choosing a voice:
- Clarity: Is every word audible and clear, even at normal conversational speed?
- Tone: Does the emotional register match your brand — warm, professional, calm, energetic?
- Naturalness: Does it sound like a real conversation or a recorded announcement?
- Pacing: Does the speaking pace feel comfortable, or does it rush or drag?
Naming your AI receptionist
You can give your AI receptionist a name — most businesses do. Choose a name that fits your brand personality and that callers will find easy to remember. A few considerations: avoid names that are identical to your own staff members (this creates confusion when callers call back asking for a specific person), and choose something that sounds natural when said alongside "speaking with [name] at [business name]."
Configuring the opening greeting
The opening greeting is the first thing callers hear. It should accomplish three things in under ten seconds: identify the business, identify the AI, and invite the caller to state their reason for calling.
A well-structured opening greeting follows this pattern:
"Thanks for calling [Business Name] — you're speaking with [AI Name]. How can I help you today?"
Warm, immediate, invites response. No lengthy descriptions of services. No "your call is important to us." Just a natural, professional open.
After-hours greeting
Configure a separate greeting for after-hours calls. Callers who ring outside business hours have different needs — they need to know you are not available right now and they need to know what will happen next. A good after-hours greeting acknowledges the time, confirms business hours, and either offers to take a message, book a callback, or both.
Closing and confirmation messages
At the end of a call, the AI should confirm what has been agreed or recorded — appointment time, message taken, callback booked — and give the caller a next step. "I have booked you in for Thursday at 2pm — you will receive a confirmation SMS within a few minutes. Is there anything else I can help with?" is a complete, professional close.
Integration Setup — Connecting Your Calendar, CRM, and Communication Tools
Integrations transform your AI receptionist from a call-answering system into the front end of your entire business workflow. When a caller books an appointment, it appears in your calendar immediately. When a new lead rings, their details flow to your CRM automatically. When a job enquiry comes through after hours, your team receives an SMS within 60 seconds.
All of this happens without any manual data entry — and it is configured in the integrations dashboard in 15 to 30 minutes.
Calendar integration
| Calendar System | Connection Method | Setup Time |
|---|---|---|
| Google Calendar | OAuth — sign in with Google | 2 minutes |
| Microsoft Outlook / 365 | OAuth — sign in with Microsoft | 2 minutes |
| Calendly | API key from Calendly account settings | 5 minutes |
| Acuity Scheduling | API key from Acuity account | 5 minutes |
| Cal.com | API key from Cal.com dashboard | 5 minutes |
| Cliniko | API key + clinic ID | 10 minutes |
After connecting your calendar, configure the booking rules: minimum notice period, maximum advance booking window, buffer time between appointments, and which calendar the AI should book into if you have multiple calendars on the same account.
CRM integration
Your CRM integration determines where lead and call data flows automatically. The platform connects natively to GoHighLevel, HubSpot, Salesforce, Zoho CRM, Pipedrive, and ServiceM8. For businesses using other CRM platforms, the Composio universal connector provides access to 980+ additional apps through a single authenticated connection — see the full integration guide for details.
When configuring your CRM integration, define the field mapping: what data captured during the call maps to which fields in your CRM. At minimum, map caller name, phone number, enquiry type, and call summary. If your CRM supports pipeline stages, map new enquiries to a specific stage so they appear in the right column of your sales view immediately.
Communication tools
Configure notification routing so your team knows about new leads and bookings in real time:
- SMS alerts: Immediate notification to a mobile number when a new lead is captured or an appointment is booked
- Email summaries: Call transcripts and lead details delivered to your inbox after each call
- Slack notifications: Instant Slack message to a designated channel when calls of specific types come in
- Microsoft Teams: Equivalent Teams integration for businesses on the Microsoft stack
If you are short on time, connect in this order: (1) calendar integration first — missed bookings are the costliest gap, (2) SMS lead alerts second — ensures no lead sits unactioned overnight, (3) CRM integration third — cleans up your lead pipeline. Do not skip step one waiting to configure steps two and three perfectly.
Testing and Refinement — How to Test Before Going Live
Testing is not optional. Every AI receptionist behaves slightly differently with real spoken language versus the written text you configured it with. Testing reveals edge cases — ambiguous questions, unusual accents, interrupted sentences, background noise — that configuration alone cannot predict.
The goal of this step is to feel confident that your AI handles the five most common call types correctly before a single paying customer reaches it.
The five core test scenarios
-
Standard booking request — Call in as a new customer wanting to book the most common service you offer. Go through the booking process end to end. Confirm the appointment appears in your calendar within 60 seconds.
-
Pricing enquiry — Ask for a price on one of your services. Verify the AI gives the answer you configured, not a fabricated number or an unhelpful non-answer.
-
Out-of-hours call — Call at a time outside your configured business hours. Verify the after-hours greeting plays correctly and the message is delivered to the right place.
-
Unclear or rambling caller — Call in and be deliberately vague. Say something like "I am not sure exactly what I need, I have been having a problem with..." and let the AI ask clarifying questions. Does it guide the caller towards a resolution or get confused?
-
Request to speak to a human — Say "Can I speak to someone please?" or "I need to talk to a real person." Verify the AI's escalation response matches your configured protocol — it should not simply hang up or loop indefinitely.
Extended test scenarios for thorough coverage
For businesses where call quality is especially critical — healthcare, legal, financial services — extend your testing to cover:
- An enquiry about a service you do not offer
- A caller who speaks very quickly or has a strong accent
- A caller who provides information in an unexpected order (e.g. gives their phone number before being asked)
- A complaint or upset caller
- A caller who asks the same question multiple times in different ways
- A call that drops mid-conversation (redial and check what the AI does with a second call from the same number)
Reviewing and acting on test results
After each test call, review the transcript in your dashboard. For any response that was not quite right, locate the relevant FAQ or knowledge section and refine the wording. Then re-test that specific scenario. Do not proceed to go-live until all five core scenarios pass cleanly.
The in-dashboard preview uses text input, which behaves differently to live speech. Always test your final configuration by calling the actual assigned phone number from at least two different mobile phones. Network compression, background noise, and speaking pace all affect real-world call quality in ways the preview does not capture.
Soft Launch — Going Live with a Percentage of Calls First
A soft launch means routing a controlled portion of your live call traffic to the AI receptionist while the remainder continues to reach your existing answering arrangement. This gives you live performance data with real callers and real conversations — without the risk of every call being handled by a system you have not yet fully validated.
Most businesses run a soft launch for five to ten business days before moving to full deployment. The exact duration depends on your call volume. You want at least 30 to 50 real calls through the system before drawing conclusions about performance.
How to configure a soft launch
In the routing settings, set the AI to handle calls in one of these configurations:
- After-hours only: The AI takes all calls outside your business hours. Your team handles everything during the day. Low risk, immediate value.
- Overflow routing: The AI picks up only when your existing line rings out. Catches missed calls without touching answered calls.
- Full routing at reduced hours: The AI handles all calls, but only during a two-hour window each day. Expands as confidence grows.
- 20/80 split: One in five calls routes to the AI; four in five go to your existing arrangement. Good for high-volume businesses.
The after-hours-only configuration is the recommended starting point for most businesses. It captures real leads that would otherwise have gone to voicemail, generates genuine call transcripts for review, and carries zero risk because your team is not available anyway.
What to monitor during soft launch
Check your dashboard daily during the soft launch period. Look for:
- Calls that ended without a clear resolution — review the transcript and add knowledge to fill the gap
- Escalation requests — are callers asking for humans more often than expected?
- Booking accuracy — do appointments in your calendar match what callers expected?
- CRM data quality — are the fields populated with useful, accurate information?
- Call duration — are calls running significantly longer than expected? (May indicate the AI is struggling to reach a resolution)
If your AI is handling some calls and your team is handling others, make sure your team knows. Nothing undermines confidence in an AI rollout faster than a team member who is unaware of the system and contradicts the AI's responses when callers follow up. Brief your team on what the AI will and will not commit to on your behalf.
Monitoring Your First 100 Calls — What to Watch For
The first 100 live calls are your richest source of optimisation data. Real callers ask questions in ways you did not predict, raise objections you did not prepare for, and follow conversational paths that no test scenario anticipated. Treating this period as active learning — not passive observation — is what separates businesses that extract strong ROI from those that plateau.
You do not need to listen to every call. Review the transcripts instead — they are available within minutes of each call ending in your dashboard. Aim to review a sample of 10 percent of calls daily, plus every call that ended in an escalation or left the caller without a resolution.
The four call outcomes to categorise
Resolved cleanly
Caller got what they needed — booking made, question answered, message taken. No human follow-up required.
Escalated intentionally
AI transferred to a human as configured. Escalation was appropriate to the call type. Follow-up handled correctly.
Knowledge gap
Caller asked something the AI could not answer well. The gap is identifiable and fixable with a knowledge base update.
Configuration issue
Booking was made incorrectly, escalation did not trigger when it should have, or data did not flow to CRM. Requires a settings fix, not a knowledge update.
For every call in categories three and four, identify the specific fix required and implement it within 24 hours. The first 100 calls typically surface five to ten distinct fixable issues. After those are addressed, the system stabilises and performance improves significantly.
Key metrics to track
| Metric | What it tells you | Target by call 100 |
|---|---|---|
| Resolution rate | Percentage of calls resolved without human intervention | 70% or above |
| Booking accuracy | Percentage of bookings that match caller expectations when followed up | 95% or above |
| Average call duration | How long calls are taking to reach resolution | Under 3.5 minutes for standard enquiries |
| Escalation rate | Percentage of calls transferred to a human | Under 15% for most business types |
| CRM capture rate | Percentage of calls where lead data flowed to CRM correctly | 99% or above |
A 70 percent resolution rate by call 100 is a solid baseline. Most businesses reach 80 to 85 percent by call 200 as knowledge gaps are filled. The 15 to 20 percent that escalates typically involves complex pricing negotiations, complaints, or genuinely unusual enquiries — these are the calls that genuinely benefit from a human touch and should escalate.
Week 1 Optimisation Playbook: The 5 Tweaks Every Business Makes
After analysing thousands of onboarded businesses, there are five consistent adjustments that come up in the first week. Not every business needs all five, but most need at least three. Read through each one and check whether it applies to your configuration.
Expanding the FAQ library with real caller questions
No matter how thorough your pre-launch FAQ library was, real callers will ask questions you did not anticipate. In week one, every call that ends with "let me get someone to call you back about that" is a signal that a FAQ entry is missing. Review these transcripts each morning and add the missing question with a clear answer before the business day starts.
Adjusting the greeting tone
The greeting you configured often sounds slightly different spoken aloud versus when you wrote it. Some businesses find their greeting is too formal, or too casual, or slightly too long. If callers are frequently interrupting the greeting before it finishes, shorten it. If you are getting feedback that the AI sounds stiff, soften the tone setting in your persona configuration.
Tightening booking confirmation messages
A common early issue is callers who are not sure what happens after they book — they did not hear the confirmation clearly, or the confirmation was phrased in a way that did not feel definitive. Review your booking confirmation script and make sure it explicitly states: the appointment date and time, the address or location if applicable, and what the caller should expect next (confirmation SMS, email, or both).
Setting up after-hours messaging properly
Businesses that launch without a properly configured after-hours protocol often find that late-night callers get an awkward experience — the AI tries to book an appointment for a time that has already passed, or provides business hours that conflict with what the caller heard. Audit your after-hours configuration specifically: does it acknowledge the time? Does it offer the right options? Does it deliver messages to the right place?
Mapping a missed-call SMS follow-up
For after-hours leads captured outside business hours, configure an automated SMS follow-up that sends the next morning at 8:00am. Something like: "Hi [name], thanks for your enquiry last night — we would love to help. A team member will call you shortly, or you can book online at [link]." This closes the loop on overnight leads before your competitors open their doors.
Month 1 Mastery: Graduating from Setup to Scaling
By the end of the first month, your AI receptionist should feel like a settled, reliable part of your business operations. Call quality should be consistent, your team should trust the data flowing into the CRM, and you should be starting to see measurable effects on your lead capture rate and booking volume.
Month one is also when the strategic work begins. Now that you have validated the core configuration, you can start expanding what the AI does.
Month 1 expansion opportunities
- Add a second language: If you serve a community where callers frequently speak a language other than English, enable multilingual support for that language. The AI will detect the caller's language automatically and respond accordingly.
- Enable proactive call-back for missed calls: Configure the system to automatically send an SMS to any caller who did not reach the AI — this catches the fraction of calls that drop before the AI answers.
- Activate conversation analytics: Review the most common enquiry types, most requested services, and peak call times from your month-one data. This information directly informs your marketing calendar and staffing decisions.
- Set up industry-specific integration templates: If you are in healthcare, legal, trades, or another industry with specialised software, configure the purpose-built integration templates for platforms like Cliniko, LEAP, simPRO, or Fergus.
- Create a review request workflow: For calls that end in a booked appointment, trigger an automated SMS seven days after the appointment asking for a Google review. This compounds the value of every successful booking.
Signs that your AI receptionist is performing well
- Your team stops asking "did anyone call about X?" — they already know, because the CRM is up to date
- After-hours bookings appear in your calendar every morning without any overnight effort
- Your lead-to-contact time drops — leads captured at 11pm are followed up by 8am the next day
- Callers who called after hours say they got through rather than going to voicemail
- You start seeing ROI within 30 to 45 days of go-live
A well-configured AI receptionist at 30 days should have: an 80%+ resolution rate, a booking accuracy rate above 95%, all CRM and calendar integrations functioning without manual corrections, and a measurable reduction in the percentage of calls going unanswered. If you are below any of these benchmarks, the troubleshooting section below covers the most likely causes.
Troubleshooting Common Issues
The following issues come up regularly across new AI receptionist deployments. Each one has a specific, actionable resolution. If you encounter a problem not listed here, your Talking Widget dashboard includes a support chat where the team can diagnose issues from your call transcripts directly.
Issue: The AI gives incorrect or outdated information
Cause: The knowledge base contains conflicting entries, or a service, price, or policy has changed since setup.
Fix: Search your knowledge base for the relevant topic. Remove any entries that contradict each other or are no longer accurate. The AI will prefer the most recently updated entry — so rewriting the correct entry rather than just adding a new one is important.
Issue: Bookings are appearing in the wrong calendar
Cause: Multiple calendars exist on the connected account and the wrong one was selected during setup.
Fix: Go to Integrations > Calendar > Calendar Selection. Confirm the correct calendar is selected. If you have changed your primary calendar since setup, the connection will need to be re-authenticated.
Issue: CRM is not receiving lead data after calls
Cause: Most commonly a token expiry on the CRM OAuth connection, or a field mapping referencing a CRM field that was renamed or deleted.
Fix: Go to Integrations > CRM > Test Connection. If the connection has lapsed, re-authenticate. If the connection is active, check the field mapping — remove any mappings pointing to fields that no longer exist and remap them to the correct fields.
Issue: The AI escalates too many calls to a human
Cause: Either the escalation trigger threshold is set too low, or large categories of questions are not covered in the knowledge base, causing the AI to escalate as a default.
Fix: Review the transcripts of escalated calls. If the questions are reasonable and answerable, add them to the knowledge base. If escalation is being triggered by the wrong keywords, adjust the escalation trigger settings under Call Routing.
Issue: Callers say the AI sounds robotic or unnatural
Cause: Voice persona settings are not matched to your business context, or the response templates in your knowledge base are written in formal, stiff language.
Fix: First, change the voice profile under Persona > Voice Selection and listen to alternatives. Second, review your FAQ answers and knowledge base entries — rewrite any that read like policy documents rather than conversational responses.
Issue: Caller says they booked but the appointment is not in the calendar
Cause: The booking confirmation was given verbally but the calendar write failed — most commonly due to a write permission issue on the connected calendar account.
Fix: Check the calendar connection under Integrations and verify the account has write access (not read-only). If your calendar is a shared or team calendar, ensure the authenticated account has edit permissions on it.
Issue: SMS notifications are not being received by the team
Cause: The notification phone number was entered incorrectly, or the number requires a country code that was not included.
Fix: Go to Notifications > SMS Alerts and verify the number. Australian mobile numbers should be entered in the format 04XX XXX XXX or +61 4XX XXX XXX. Re-save the setting and use the Test Notification button to send a test SMS.
Issue: After-hours calls are not triggering the after-hours greeting
Cause: Business hours are configured in a timezone that does not match your actual location, or the hours were saved incorrectly (e.g. 24-hour clock error).
Fix: Go to Settings > Business Hours and verify the timezone is set to your state (e.g. AEST for Queensland, AEDT for NSW/VIC during daylight saving). Double-check the start and end times for each day.
Issue: The AI keeps asking the caller for information they already provided
Cause: A data collection flow is configured to ask for specific fields regardless of whether the caller has already volunteered that information in the opening of the call.
Fix: In your call flow settings, enable the "caller context awareness" option. This instructs the AI to listen for volunteered information at the start of a call and skip the corresponding collection steps if the data has already been provided.
Issue: Callers are hanging up before the AI finishes the greeting
Cause: The opening greeting is too long. Most callers will hang up if they are not invited to speak within eight to ten seconds.
Fix: Shorten the greeting. The maximum effective length is two sentences — business name, invite to speak. Everything else can be said after the caller has stated their purpose.
Frequently Asked Questions
Quick answers to the most common onboarding questions.
Initial account setup takes around five minutes. Building a solid knowledge base — the part that most determines call quality — takes one to two hours and is done in a guided wizard. Connecting your calendar and CRM adds another 15 to 30 minutes. Most businesses are live within a single afternoon, and refining the AI through the first 100 calls takes another two to three weeks of light monitoring.
The AI needs to know your business name, the services you offer, your pricing policy (whether to quote on call or send callers to your website), your trading hours, your booking process, your escalation contacts for emergencies, and the top 10 to 15 questions callers ask most often. The more complete this knowledge base, the fewer calls will need a human to follow up.
Call your assigned test number from at least three different phones. Run through five core scenarios: a standard booking request, a pricing question, an out-of-hours call, an unclear or rambling caller, and a caller wanting to speak to a human. Review the transcript for each call, check that calendar bookings appear correctly, and verify that lead data has flowed to your CRM. Fix any gaps before enabling the live number.
A soft launch means routing a controlled percentage of real calls — typically 20 to 30 percent — to your AI receptionist while the rest continue to ring your existing line. This lets you observe how the AI handles genuine callers, catch any unexpected edge cases, and build confidence before transferring 100 percent of call volume. Most businesses maintain a soft launch period of five to ten business days.
The five most common week-one adjustments are: (1) adding more service-specific FAQs after hearing real caller questions, (2) adjusting the greeting to match the business's natural tone, (3) tightening booking confirmation messages so callers are clearer about what happens next, (4) setting up after-hours messaging so late-night callers receive an appropriate response rather than a booking attempt, and (5) mapping a missed-call SMS follow-up so leads captured outside hours are contacted promptly the next morning.
Yes. Talking Widget supports multilingual AI receptionists. You configure the primary language during setup, and can enable secondary languages so the AI detects the caller's language automatically and responds accordingly. This is particularly useful for businesses serving diverse communities in major Australian cities.
When the AI encounters a question outside its knowledge base, it has three configurable fallback behaviours: (1) politely acknowledge it does not have that specific information and offer to take a message for a human to follow up, (2) escalate the call to a live team member if one is available, or (3) book a callback appointment. You decide which fallback applies during setup, and can set different behaviours for business hours versus after hours.
Start your AI receptionist setup now — it takes 5 minutes
Every missed call is a missed opportunity. Set up your AI receptionist today and let it start capturing leads, booking appointments, and answering questions around the clock — while you focus on the work only you can do.
Start free setup See a live demo first