AI Voice Agent Client Onboarding Checklist (First 7 Days)
A systemized AI voice agent client onboarding process gets a new client's agent live in under one hour and sets clear expectations for the first seven days. The checklist covers five phases: build the agent from the client's website URL (5 minutes), configure call forwarding on their existing number (30 seconds), connect their calendar, send the welcome email, then review transcripts daily through Day 7 until you send the first performance snapshot. Agencies that follow this cadence report fewer Day 3 cancellations and more referrals at the Day 7 check-in.
This guide breaks down the full voice agent client onboarding process into a day-by-day checklist you can hand to any team member. Every step includes what to do, what to look for, and when to involve the client versus handling it yourself.
Day 0: Signing Day
Day 0 is everything between the signed contract and the client's first real call. The goal is to get the agent answering calls before the client closes their laptop for the day.
Build the Agent from the Client's Website URL
Paste the client's website URL into your white-label platform. The system scrapes their site, Google reviews, and social media to auto-generate a knowledge base covering services, pricing, hours, location, and common questions. Review the generated knowledge base before going live: check that business hours are correct, service descriptions are accurate, and pricing matches what the client actually charges. Add any information the website is missing, such as emergency service policies, seasonal offerings, or services they want to promote over others.
Configure Call Forwarding
The client keeps their existing business number. Call forwarding routes unanswered, declined, or busy calls to the AI agent. Setup is fast using carrier-tested step-by-step guides (available for Telstra, Optus, Vodafone, AT&T, Verizon, T-Mobile, and all major UK, NZ, and Canadian carriers). Walk the client through it on a screen share or send the guide for their carrier. Confirm forwarding is active by calling their number, letting it ring to the AI, and verifying the agent picks up.
Connect the Calendar
If the client wants the AI to book appointments, connect their calendar (Google Calendar, Cal.com, or Outlook) during this session. Test it: have the agent book a dummy appointment and verify it appears on the client's calendar with the correct time, date, and service type. Delete the test appointment afterward.
Send the Welcome Email
Before you end the Day 0 session, send the welcome email. This sets expectations for the next seven days and gives the client a single reference point for how the service works.
The Welcome Email Template
A strong welcome email covers four things: what is live right now, what the client should expect this week, how to reach you, and one clear instruction for what they should do if the AI gets something wrong.
Subject line: Your AI receptionist is live. Here is what to expect this week.
Body structure:
Confirmation. "Your AI receptionist is now answering calls on [their business number]. It activates when you miss a call, decline a call, or are busy on another line. You always have the option to pick up first."
What it knows. "The AI has been trained on your website, reviews, and the details we discussed. It can answer questions about your services, hours, and pricing, and it books appointments directly into your calendar."
This week's plan. "Over the next 7 days, I will be reviewing every call transcript daily and fine-tuning the AI's responses. By Day 7, I will send you a performance snapshot showing how many calls were handled, appointments booked, and any adjustments I made."
How to flag issues. "If a caller tells you the AI said something incorrect, forward me the details and I will fix it within 24 hours. You do not need to troubleshoot anything yourself."
Contact info. Your direct email or messaging channel.
Keep it under 200 words. The client should be able to read it in under a minute.
Day 1: First Call Review
Pull every transcript from the first 24 hours. Read each one completely. You are looking for three things: did the agent answer the caller's question correctly, did it collect the right information, and did it handle the call ending naturally. Flag any response where the AI gave incorrect information, failed to answer a question it should know, or sounded unnatural.
Fix knowledge gaps immediately. If the AI quoted the wrong price, update the knowledge base entry. If it could not answer a question about a service the client offers, add that service to the knowledge base. If a caller asked about something outside the business scope (like a service the client does not offer), verify the AI handled the deflection correctly by taking a message and offering a callback.
Call the agent yourself after making fixes. Test the specific scenarios that failed. Confirm the corrections are working before moving on.
Days 2 and 3: Monitor, Adjust, Test Edge Cases
Continue reading every transcript daily. By Day 2 and 3, you have enough call volume to spot patterns rather than isolated issues.
Common fixes during Days 2 and 3:
Wrong business hours. The website says "Mon-Fri 9-5" but the client actually works Saturdays or has different summer hours. Update the knowledge base with the actual schedule.
Missing services. The website lists general categories ("plumbing services") but does not mention specific offerings the client wants to promote (drain cleaning, water heater installation, sewer line repair). Add these with descriptions and approximate pricing.
Incorrect pricing. The website shows outdated pricing or does not list pricing at all. Get the current price list from the client and add it to the knowledge base. This is one of the few times you need the client's input directly.
Appointment type confusion. The AI books a 30-minute consultation when the client needs 60-minute slots for new patients, or books a "general inquiry" when the caller needs a specific service appointment. Adjust the calendar integration settings and appointment types.
After-hours handling. The AI should acknowledge it is after hours, take detailed information, and promise a callback during business hours. If it is not doing this, configure after-hours messaging.
Test edge cases yourself:
Call and ask about a service the business does not offer
Call and try to book an appointment at a time the calendar shows as unavailable
Call with a vague request ("I need help with my thing") and see how the AI qualifies it
Call and ask for pricing on a service with variable pricing ("How much does it cost to fix a leak?")
Call and pretend to be a spam caller or telemarketer
Each failed edge case is a knowledge base update. Make the fix, test it, move on.
Days 4 and 5: Verify Booking Accuracy and SMS Follow-Ups
By Day 4, shift focus from general call quality to the two features clients notice most: appointment booking and follow-up messages.
Appointment booking checks:
Pull every booked appointment from the last 4 days. Compare each one against the call transcript.
Does the appointment type match what the caller requested?
Is the appointment at a time the client is actually available?
Did the AI confirm the date, time, and service with the caller before booking?
Did the calendar receive the correct contact details (name, phone number, reason for visit)?
If you find booking errors, the most common causes are misconfigured appointment durations, missing buffer times between appointments, or calendar sync delays. Fix these in the calendar integration settings.
SMS follow-up checks:
Did confirmation texts go out after each booked appointment?
Did the SMS contain the correct date, time, and business name?
For callers who did not book but left a message, did the AI send a follow-up text?
Are the texts coming from the client's branded number (not a generic platform number)?
Document every fix you make during Days 4 and 5. You will include this in the Day 7 snapshot.
Day 6: Compile First-Week Data
Pull the numbers for the performance snapshot you will send on Day 7. The five metrics that matter most to clients:
Total calls handled by the AI (not forwarded, not missed, actually handled)
Calls that would have been missed without the AI (after-hours calls, busy-line calls, calls the client declined)
Appointments booked by the AI, with a count of confirmed versus rescheduled
Knowledge base updates made (this shows the client you are actively improving their agent)
Issues identified and resolved (with a one-line description of each fix)
Format these into a short, scannable summary. Do not send a raw data dump. The client wants to see: "Your AI handled 47 calls this week. 31 of those came in when you could not answer. It booked 12 appointments and I made 8 improvements to how it handles your callers."
If your agency manages 10 or more clients, build this reporting into a weekly workflow so it scales without eating your schedule.
Day 7: Performance Snapshot, Feedback, Referral Ask
Day 7 is the most important day in the onboarding cycle. It is the difference between a client who stays for 12 months and one who cancels after 30 days.
Send the Performance Snapshot
Email or message the client with the data you compiled on Day 6. Lead with the number they care about most: calls that would have been missed. A client who sees "31 calls handled that would have gone to voicemail" understands the value immediately without needing a revenue calculation.
If you want to include revenue impact, use simple math: multiply missed calls by the client's average job value and a 30% conversion rate. For a plumber averaging $350 per job, 31 missed calls recovered equals roughly $3,255 in potential revenue for the week. This math belongs in the snapshot, not in a separate ROI report, which you will send at the 30-day mark.
Ask for Feedback
Ask two questions, not ten:
"Has any caller mentioned something the AI got wrong that I have not already fixed?"
"Is there anything you want the AI to handle differently?"
These surface problems you cannot see in transcripts (callers who complain to the business owner directly, or preferences the client has not mentioned).
Ask for a Referral
If the client is satisfied, ask now. Not at Day 30. Not after the first invoice. Day 7, while the result is fresh: "Who else do you know that is dealing with missed calls?" One referral per client in the first two weeks compounds faster than any ad campaign.
When to Involve the Client Versus Handle It Yourself
The default rule: handle everything yourself unless you need information only the client has. Most onboarding friction comes from agencies asking clients to do things the agency should do.
Handle yourself (do not bother the client):
Knowledge base updates based on publicly available information (website, reviews, social media)
Call forwarding troubleshooting (use the carrier guides)
Appointment type and duration adjustments (based on transcript patterns)
After-hours messaging configuration
SMS follow-up wording and timing
Edge case testing and fixes
Involve the client (you need their input):
Pricing that is not listed on their website or has changed
Services they want to promote or hide that are not obvious from their website
Specific call handling preferences ("always ask for insurance info," "never quote a price for custom work")
Calendar availability exceptions (vacations, holidays, reduced hours)
Complaints from callers that did not show up in transcripts
When you do need client input, batch your questions into a single message. Do not send five separate emails across three days. Compile everything and send once.
Frequently Asked Questions
How long does it take to get a client's AI voice agent live?
Under one hour from signed contract to first call answered. Agent creation from a website URL takes about 5 minutes, call forwarding setup takes 30 seconds, and calendar connection takes another 5 to 10 minutes. The rest of the hour covers testing, knowledge base review, and sending the welcome email.
What if the client's website has outdated or missing information?
Build the agent from the website first (it still captures business name, location, general services, and reviews), then manually add or correct the gaps. Ask the client for a current service list and price sheet during the Day 0 session. Most corrections take 10 to 15 minutes.
Should I give the client access to the agent dashboard during onboarding?
Not during the first 7 days. You want to control the experience while you are still tuning the agent. After Day 7, once the agent is performing consistently, give the client dashboard access so they can view transcripts and call summaries. This is when a branded client dashboard adds value without creating confusion.
How many clients can one person onboard per week using this checklist?
With this process systemized, one person can onboard 3 to 5 new clients per week while maintaining daily transcript review for each. Beyond 5 simultaneous onboardings, transcript review becomes the bottleneck and you need to hire or automate parts of the review process.
What is the most common reason clients cancel in the first 30 days?
The AI giving incorrect information that the agency did not catch and fix. Wrong hours, wrong pricing, or missing services erode trust fast. Daily transcript review during Days 1 through 7 prevents the most common cancellation triggers by catching errors before the client hears about them from a customer.
Related Resources
Why Website Scraping Beats Manual FAQ Entry for Voice AI Agency Onboarding
Voice Agent Client Churn Reduction: How Agencies Keep Clients on Retainer in 2026
Voice AI Client Success Metrics: What Agencies Should Track in 2026
Build and resell voice AI agents under your own brand at trillet.ai/whitelabel.
