How to Run a Voice AI Discovery Call (Script + Framework)

How to Run a Voice AI Discovery Call (Script + Framework)

Editor's note (September 2026): converted universal timing, close-rate, and pricing claims into a testable discovery-call framework and aligned the Trillet offer with current entitlements.

Phase 1 (Minutes 1 to 2): Understand Their Business

The first two minutes are not rapport building. They are data collection. You need three numbers from the prospect before anything else: how many calls they get per day, what happens when they miss one, and what their average job or appointment is worth. Without these, you cannot run the missed call math later, and without the math, you have no close.

Open the call with something like: "Thanks for booking this, I know you're busy. I want to make sure this is actually a fit for your business before I show you anything. Can I ask you three quick questions?"

The Three Opening Questions

1. "How many calls does your business get per day?" Most owners underestimate this. If they say "a few," press: "Would you say closer to 5 or closer to 15?" You need a number, not a feeling.

2. "What happens right now when you miss a call?" The answer might be voicemail, an answering service, a callback queue, or nothing. Ask how many callers leave a message and how many are recovered; the prospect's own call logs are more useful than a generic voicemail statistic.

3. "What's your average job or appointment worth?" For a plumber, it might be $350. For a roofer, $8,000. For a dentist, $3,000 in lifetime patient value. This number is the foundation of every ROI calculation you will use for the rest of the call.

Write these numbers down visibly (or repeat them back) so the prospect knows you are paying attention. You will use them in Phase 3.

Phase 2 (Minutes 3 to 7): The Live Demo

Before the call, build a bespoke voice AI agent for the prospect's business using your white-label platform. Trillet can generate a starting agent from a website, but you still need to review the imported information, remove unsupported claims, and test realistic calls before showing it. The prospect does not need to watch the setup process.

On the call, frame it like this: "I went ahead and built a custom AI receptionist for your business. Let me show you what it sounds like." Then call the agent live. Let the prospect listen.

Demo Rules

You run the demo. They listen. Do not hand over control. Do not let them call the agent themselves. Do not give them the phone number to "play with later." A prototype is not a finished product, and prospects who test edge cases on an unrefined agent will find flaws and judge the entire concept by them. You control the narrative. You choose which questions to ask. You show the agent at its best.

Ask it questions the prospect's actual callers would ask. If it is a plumbing business, ask about emergency drain service at 9 PM. If it is a dental office, ask about new patient appointments and insurance. Make it feel real, because it is real.

Keep it to two or three calls. More than that and you lose momentum. The demo should take three to five minutes, not fifteen.

The Closing Line After the Demo

"This is the starting point. We have not connected it to your calendar or CRM, completed the service-specific workflow, or run production acceptance tests yet. The next step is configuring and measuring it against your real calls."

This line works because it reframes the demo as a starting point, not a ceiling. Whatever impressed them about the prototype, the finished product will be better. For more on structuring demos that close, see Voice Agent Sales Demo Best Practices.

Phase 3 (Minutes 8 to 15): Pricing, Objections, and Close

Phase 3 is where most agencies lose deals, not because their price is wrong, but because they present it without context. Before you say a number, do the missed call math live on the call using the data the prospect gave you in Phase 1.

The Missed Call Math (Do This Live)

Take the numbers they gave you and multiply:

"You said you miss about 8 calls a week. Your average job is $350. Even if only 30% of those would have converted, that is $350 times 2.4 jobs per week, times 52 weeks. That is $43,680 a year in revenue walking out the door. I am charging you $400 a month, which is $4,800 a year, to catch those calls. That is a 9x return."

Do the math out loud. Let them hear it. The price you quote after this calculation feels small by comparison. For the full breakdown of industry-specific missed call calculations and scripts, see The Missed Call Math.

Present One Number

Do not present tiers. Do not break out per-minute costs. Do not show a Bronze/Silver/Gold pricing table. SMBs want one monthly number that covers everything they need. A confused mind says no.

"Based on what you need, this is $450 a month. That covers 24/7 answering, calendar booking, and SMS confirmations. Month-to-month, cancel anytime." A $300-$600 monthly range is a useful planning scenario for some small-business offers, not an industry benchmark. Set the actual price from usage, telephony, support, implementation, and client value. If your agency offers a guarantee, define its conditions in your own agreement rather than implying it comes from the platform.

One number. Clean deliverables. If you need guidance on setting that number for different industries, How to Price AI Voice Agents covers the one-number rule in detail.

Handle Objections by Resistance Level

Not every prospect objects the same way, and not every objection requires the same response. Match your response to the resistance level:

The full framework with word-for-word scripts for each level is in 5 Levels of Client Resistance (And Exactly What to Say at Each One).

The 5 Qualifying Questions to Ask Before You Quote a Price

Five questions separate a qualified prospect from a time-waster. Ask all five before Phase 3. If you cannot get clear answers to at least four of them, the prospect is not ready to buy.

1. "How many calls does your business get per day?" This determines call volume and whether an AI agent makes financial sense. Below three calls per day, the ROI math gets weak.

2. "What happens when you miss a call right now?" If the answer is "my office manager gets it" or "we have an answering service," you are displacing a human, which is a harder sell than replacing voicemail. Adjust your pitch accordingly.

3. "What is your average job or appointment worth?" This is the multiplier for missed call math. Higher values make the ROI undeniable. Lower values (under $100) mean you need higher call volume to justify the price.

4. "Have you tried anything else to solve this problem?" If they have tried and failed with answering services, voicemail systems, or hiring staff, they already know the problem is real. You are presenting a better solution, not educating them on the problem. If they have never tried anything, you may need to spend more time building urgency.

5. "If this worked as shown, when would you want to start?" This separates buyers from browsers. "Immediately" or "this week" means they are ready. "Maybe next quarter" or "I'll think about it" means they are either not in pain or not the decision-maker.

Red Flags: When to Walk Away

Not every prospect is a fit. Spending 30 minutes on a call with someone who will never buy is 30 minutes you could have spent closing a real deal. Watch for these signals:

Low call volume (fewer than 3 calls per day). The math does not work. If they get 2 calls a day and miss one, the annual cost of missed calls is too small to justify a $300 to $500 monthly fee. These businesses need more marketing before they need an AI agent.

No budget and no urgency. If they say "I need to check with my partner" and cannot give you a timeline, they are not the decision-maker. Ask: "Is there someone else who should be on this call?" If the answer is yes, reschedule with that person present.

"Just curious" energy. Some people book discovery calls because they saw an ad and thought it was interesting. They have no active problem, no missed calls, and no intention of spending money. The giveaway is vague answers to your qualifying questions and zero emotional reaction to the missed call math. Thank them for their time and move on.

They want to build it themselves. If the prospect is technically savvy and starts asking about APIs, which LLM you use, or whether they can self-host, they are not your customer. They want developer infrastructure, not a done-for-you service. This is fine. Point them in the right direction and protect your time.

They are already happy with their current solution. If their answering service or receptionist is working well and they are just price-shopping, you are fighting an uphill battle. The best prospects are the ones whose current system is failing them.

Putting the Full Script Together

Here is the complete 15-minute structure in sequence, with approximate timing for each section:

Minutes 0 to 2 (Qualify): "Thanks for booking this. I want to make sure this is actually a fit before I show you anything. Quick three questions: How many calls does your business get per day? What happens right now when you miss one? And what's your average job worth?"

Minutes 2 to 3 (Transition to demo): "Based on what you've told me, this is exactly the kind of business where this works best. I actually went ahead and built a custom AI receptionist for [business name]. Let me show you what it sounds like."

Minutes 3 to 7 (Live demo): Call the bespoke agent two to three times. Ask realistic questions. Let the prospect listen. After the last call: "This is what it sounds like out of the box, and we haven't even connected it to your calendar or trained it on your specific services yet."

Minutes 7 to 10 (Missed call math): "You said you miss about [X] calls a week. At $[Y] per job and a 30% conversion rate, that's $[Z] per year in lost revenue. This costs $[price] a month, which is $[annual] per year. That's a [multiplier]x return."

Minutes 10 to 12 (Pricing and terms): "Based on what you need, this is $[single number] a month. That covers 24/7 answering, calendar booking, and SMS follow-up. These are the usage limits, overage terms, and cancellation terms."

Minutes 12 to 15 (Close or handle objections): If yes: "I can have this live and answering your phones by tomorrow. Want to get started?" If objection: match to resistance level and respond accordingly.

Once they say yes, the speed of your delivery becomes the next selling point. Walk through how to get the agent live fast in the voice agent client onboarding process. And if you are still filling your pipeline, how to get your first 5 AI agency clients covers where these discovery calls come from in the first place.

Frequently Asked Questions

How long should a voice AI discovery call take?

Fifteen minutes is a useful first agenda: a short qualification, a focused demo, then pricing and next steps. Complex workflows or regulated deployments may need a longer technical follow-up. Measure which duration produces enough discovery without wasting the prospect's time.

Should I build a custom demo agent for every prospect?

Usually, if the account is qualified. A website-generated agent can give you a fast starting point, but review and test it before the call. A relevant demo makes the workflow concrete; track its effect on your own conversion rate rather than assuming a universal lift.

What if the prospect asks what platform I use?

You designed and support the client solution under your brand, but answer architecture and subprocessor questions accurately. A safe explanation is: "We operate this service on Trillet, a first-party voice AI application platform. Trillet manages the model, telephony, and other subprocessors through one platform relationship; we remain responsible for your configuration and service." For how white-labeling positions your brand on top of the platform, see the white-label voice AI platform guide.

What close rate should I expect from discovery calls?

There is no reliable universal close-rate benchmark for this framework. Track qualified opportunities separately from all booked calls, then segment results by niche, source, offer, and salesperson. If conversion is weak, review demo quality, discovery depth, the missed-call assumptions, pricing clarity, and follow-up before changing the offer.

What if the prospect says they need to think about it?

Ask what specifically they are unsure about and address it in the moment. "Is it the money, the technology, or whether it'll work for your type of business?" Most "I need to think about it" responses are fear, not logic. If they genuinely need time, follow up within 48 hours. Prospects who "think about it" for more than a week have mentally moved on.

Honest caveat: This framework works for prospects who have a real missed-call problem and enough call volume to justify the investment. It will not save a deal where the prospect has no budget, no urgency, or no authority to make a decision. Qualifying hard in Phase 1 is how you avoid wasting time on calls that were never going to close.

Run this framework on a platform built for agencies: explore Trillet White-Label at $99/month for Studio (three workspaces and 1,000 included AI minutes) or $299/month for Agency (unlimited workspaces and 3,000 included AI minutes). AI overage is $0.12/minute after the allowance and telephony is separate. The card-required white-label trial lasts seven days and includes usage credit ($12 on Studio, $24 on Agency); it is distinct from Trillet's 28-day D2C money-back guarantee. Compare tiers on the white-label pricing page, and read the full white-label voice AI platform guide.

Updated for September 2026: aligned the trial, workspace, minute, and telephony terms; removed unsupported close-rate promises; and clarified how agencies should describe the platform relationship. Also corrected the White-Label trial credit to $12 (Studio) / $24 (Agency).

Related Resources

• The 60-Day AI Voice Agency Launch Plan (Complete Playbook)
• Voice Agent Client Onboarding Process: How to Get Clients Live in Under 24 Hours
• How to Price AI Chatbot Services
• Voice AI for HVAC Companies Reseller: How Agencies Can Capture the Home Services Market