How to Bill Voice AI Clients: Retainers, Included Minutes, Overage and Stripe for Retell Agencies (2026)

TL;DR: Most voice AI agencies get the pricing right and the billing wrong. They agree a monthly retainer, then chase invoices by email, eat the call minutes when a client has a busy month, and find out a card failed three weeks after it happened. The fix is boring and it works: put every client on a Stripe subscription, charge a monthly fee that includes a block of minutes, add an overage price per minute above that block, take the setup fee up front, and let the client see their own usage before the invoice arrives. I build production voice agents for US clients on Retell, n8n, GoHighLevel and Twilio, and we run VoiceDash, the white-label client portal agencies hand their clients, which now bills clients through the agency's own Stripe. Here is how I set up billing for a voice AI client, and what I stopped doing.

This post is about the mechanics of getting paid. If you are still deciding what number to charge, start with voice AI agency pricing and Retell AI pricing explained, then come back here.

Why billing breaks before pricing does

Invoices by email get paid late. A local business owner opens your invoice on their phone between jobs, means to pay it, and forgets. You send a reminder, then another. By month three you are a collections department.

Minutes are a variable cost on a fixed price. Retell bills you per minute. If you charge a flat retainer with no ceiling, the client who doubles their call volume in peak season doubles your cost while your revenue stays flat. I covered that margin risk in the pricing post, but the billing setup is where you actually protect it.

Overage surprises cause churn. If you do charge for extra minutes, and the first time the client hears about it is a larger invoice, that invoice becomes a renewal conversation you did not want to have.

Failed cards go unnoticed. Cards expire. Banks decline. If nobody tells the client to update their card, you keep paying Retell for an agent that stopped earning.

None of these are pricing problems. They are plumbing problems, and plumbing is fixable.

The billing model I use for most clients

For a typical small business client running an inbound AI receptionist, I use four pieces.

1. A one-time setup fee, paid before the build

The setup fee covers discovery, prompt writing, integrations, testing and launch. I take it before I start building, not after launch. Two reasons. It filters out people who are not serious, and it means the build is funded even if the client stalls on giving you calendar access for two weeks. Scope for the setup fee belongs in the agreement, which I walk through in the voice AI agency contract post.

2. A monthly fee with included minutes

The core of the bill is a monthly subscription that includes a block of call minutes. The client buys an outcome, an agent that answers every call, at a price they can budget for. You size the included block from the call volume you saw in discovery or the pilot, with headroom.

This keeps the conversation about value, not about minutes. Owners do not want to think in minutes. They want to know the phone gets answered.

3. An overage price per minute above the block

The overage price is what protects your margin. If a client blows past their included minutes, each extra minute is billed at a rate that covers your Retell, telephony and LLM cost with room left over. Set it in the agreement on day one so it is never a surprise.

The overage rate also does quiet sales work. A client who goes over the block every month is a client who should move to a bigger plan, and that is an easy upsell conversation, covered in how to upsell voice AI clients.

4. Optional per-call or per-minute plans for odd shapes

Some clients do not fit a retainer. An outbound campaign that runs for six weeks, a seasonal business that is quiet half the year, or a pilot where nobody knows the volume yet. For those I sometimes bill purely per minute or per call. It is less predictable for both sides, so I use it as the exception, not the default.

Why Stripe subscriptions beat sending invoices

I bill every client through Stripe on a subscription with a card on file. The client pays once at signup and is charged automatically after that. Compared to emailed invoices, this changes a few things:

• You stop chasing. The charge happens on the billing date whether or not the owner remembers.
• Failed payments get handled. When a card fails, the client is asked to update it, instead of you noticing weeks later.
• The client manages their own card and invoices. No more "can you resend last month's invoice for my accountant" emails.
• Promotions are clean. Coupons, promo codes and a free trial period for a pilot are settings, not spreadsheet notes.

The objection I hear from agency owners is that asking a local business for a card feels pushy. In practice, the clients who refuse a card on file are the same clients who pay invoices 45 days late. A card on file is normal for any software they already use, and the more your service looks like software, the more normal it feels.

Show usage before you bill it

This is the part most agencies skip, and it is the one that prevents the angry overage call.

If the client can see their minutes used, calls handled and cost for the period while the month is still running, the overage line on the invoice is old news by the time it arrives. They watched it climb. Often they are pleased about it, because more minutes usually means more calls answered that would have gone to voicemail.

The opposite is a client who only sees a number on a PDF. Then every overage charge reads as "the agency found a way to charge me more."

So I give every client a place to see this themselves, next to the calls, recordings and transcripts that prove what the minutes bought. Which numbers to put in front of them is in voice AI client reporting.

How we do this inside VoiceDash

When I ran voice agents for clients, billing lived in one tool, the agent in Retell, and the reporting in screenshots. We built client billing into VoiceDash so the agency can run all of it from the same place the client portal lives. Here is what it does, based on what is live on the product today:

• Connect your own Stripe account. Payments go straight to your Stripe. VoiceDash takes no cut of what your clients pay you.
• Create the products you actually sell. A monthly fee with included minutes and an overage price, pay per minute, pay per call, or a one-time setup fee, in any currency.
• Share a payment link or start a subscription. When a client pays through a payment link, their branded portal and login are created automatically. Onboarding and payment become one step.
• Retell minutes are billed automatically. Call minutes flow from Retell into billing, and clients see their usage and cost in their own portal.
• Coupons, promo codes, free trials, installment plans and automatic tax are available for the deals that need them.
• Clients update their own card and download invoices. If a payment fails, they are asked to update their card.

All of this sits behind your logo and your domain, like the rest of the portal. Your client never sees Retell or VoiceDash, which is the whole point of white-labeling Retell AI. Client billing is included on every plan, starting with Starter at $19 a month, and there is a free trial.

If you would rather build this yourself, it is doable, but metering minutes from Retell into Stripe usage records, handling failed payments and scoping each client's data is a real project. I compared the two paths honestly in build vs buy a voice AI client portal.

A billing setup you can copy

Here is the order I set things up for a new client:

1. Agree the numbers in the contract. Setup fee, monthly fee, included minutes, overage rate per minute, billing date, and what happens if a payment fails.
2. Create the products once. One setup fee product, one monthly plan per tier with its included minutes and overage price. Reuse them across clients in the same niche.
3. Send one payment link at signing. The setup fee and first month together, so the build is funded and the subscription is live from day one.
4. Use a trial for pilots. If you run a voice AI pilot program, put it on a free trial or a discounted first period, with the card collected up front so conversion is automatic.
5. Give the portal on day one. Usage, calls and recordings visible before the first renewal.
6. Review overage every month. Any client consistently above the block gets a plan upgrade conversation, not a bigger invoice.

Mistakes I made so you do not have to

Flat retainer, unlimited minutes. One busy season wiped out the margin on a client. Every plan I sell now has an included block and an overage rate.

Billing after launch. I finished a build, launched it, then sent the setup invoice. It took two months to get paid for work that was already live. Setup is now paid before the build starts.

Invoices with no context. A bigger invoice with no visible usage behind it triggered a "why did my bill go up" call that nearly became a cancellation. Usage has been visible in the client portal ever since.

Too many custom deals. Every client on a slightly different arrangement made month-end a puzzle. Three tiers and one overage rate per niche is enough for almost everyone.

The bottom line

Billing voice AI clients is not about finding a clever pricing model. It is about a setup fee paid before the build, a monthly subscription with included minutes, a clear overage rate, a card on file, and usage the client can see before the invoice lands. Get that plumbing right and you stop chasing money and start spending that time on the next client.

Want to bill clients through your own Stripe from the same branded portal they log into? Start free on VoiceDash or book a demo and I will show you the billing setup I would use for your clients.