Most voice AI agencies are leaving money on the table. Not because they're bad at their craft — because they're pricing like software vendors when they should be pricing like operators.
Here's what that looks like in practice: an agency builds a fully functional AI receptionist for a dental practice, negotiates a $300/month retainer, and then wonders why the economics don't work when they're managing fifteen clients. The answer is usually that $300 was a number someone made up, not a number derived from what the service actually costs to deliver — or what it's actually worth to the client.
This post is about fixing that. Real numbers, real structure, no vagueness.
The three pricing models agencies use (and which one actually scales)
Model 1: Per-minute / usage-based
You charge based on call volume — typically somewhere between $0.08 and $0.25 per minute depending on your provider costs and margin. Simple to explain. Scales with usage.
The problem: your most valuable clients are high-volume clients, and high-volume clients are exactly who will eventually ask you why they're paying so much. Usage-based pricing commoditizes the relationship. The client sees a line item, not a service.
Model 2: Monthly retainer
A flat monthly fee that covers a defined scope of calls, management, and reporting. Most agencies end up here eventually.
This works — but only if you've priced the retainer correctly. Most agencies price their first retainer by guessing what a client might accept rather than calculating what it costs them to deliver the service. The result is retainers that are profitable at one client and loss-making at five.
Model 3: Value-based / outcome-linked
You tie pricing to business outcomes — booked appointments, qualified leads, calls handled without a human agent. Higher ceiling, harder to sell, requires you to have a clear picture of what the client was doing before your service.
This is the right model for mature agency relationships. It's not where you start.
The answer: most agencies should run a hybrid of Models 2 and 3. A predictable monthly retainer for the base service, with a performance component layered in once you have data to back it up. You're not guessing. You're charging for results.
What to actually charge
These numbers assume a mid-market client: a multi-location business with meaningful inbound call volume and a real before/after comparison you can point to.
Setup fee: $2,500–$5,000
This covers discovery, configuration, testing, and the first two weeks of monitoring. If someone pushes back on a setup fee, that's useful information — it tells you how much they value the service. The setup fee also creates a natural transition between the sales process and the ongoing relationship.
Monthly retainer: $1,200–$2,500 per client
The lower end of this range is appropriate for single-location clients with straightforward workflows. The upper end is appropriate for multi-location clients, clients with custom integrations, or clients where you're providing reporting and ongoing optimization as a named deliverable.
Below $1,000/month is where the economics stop working. At $800/month you're not running an agency — you're running a managed service that slowly eats your team's time.
Usage overage: charge for it
Build a usage tier into your retainer (say, 1,000 minutes/month) and define the overage rate in the contract. Don't eat overage. It's a legitimate cost, and clients who generate high volume are getting high value.
Platform access: separate line item if relevant
If your clients can log into a portal to see their own call data and reports, that's a product feature — and it's reasonable to include it as a named line item rather than hiding it in the retainer. It also makes the value tangible at renewal time.
The mistake that makes your pricing look cheap
Bundling everything into one number.
When you say "$1,500/month," the client hears a cost. When you say "$3,500 setup, then $1,500/month which includes 1,000 minutes of call handling, a dedicated client portal, and monthly reporting," the client hears a service.
The itemization isn't just for perception — it protects you commercially. If a client asks you to cut costs, you have something to actually negotiate: reduce the reporting cadence, move to a lower minute tier, remove the portal access. Without itemization, the only thing left to cut is your margin.
The counterintuitive thing about pricing at volume
Here's something most agencies figure out too late: your cost to serve a client goes down as you scale, but your pricing shouldn't reflect that.
When you have fifteen clients on similar configurations, you've built expertise, templates, and operational processes that make each new client cheaper to onboard. That efficiency belongs to you, not to the client. Your pricing to each individual client should be based on the value you're delivering to that client — not on how much cheaper your fifteenth client was to onboard than your first.
This is the argument for retainer pricing over usage pricing. Usage pricing passes scale efficiency to the client. Retainer pricing keeps it with you.
The operational implication: as you grow, the work is less in building and more in managing. Each additional client adds configuration management, per-client reporting, and the administrative overhead of keeping client data isolated and their workflows running. A platform like Voxfra handles that layer — provisioning client environments, keeping call data separated, giving each client their own reporting view — so the marginal cost of client fifteen doesn't creep back up as your client list grows.
How to raise prices on existing clients without losing them
You'll need to do this eventually. Here's how:
Don't apologize. "Given what we've built together and the results we've delivered, we're moving to $1,800/month effective next quarter" is a complete sentence.
Give 60 days notice. Less than that feels reactive. More than that creates an opening for the client to shop around during the notice period.
Tie it to something specific. A new feature, a significant result, an expanded scope. You're not raising prices because costs went up — you're raising prices because the value went up.
Expect 10–15% attrition. Some clients will leave. If 100% of your clients accept a price increase without any pushback, you didn't raise prices enough.
What your pricing structure should look like by client 10
By the time you have ten clients, you should have a standardized pricing sheet that you send at the start of every sales conversation. Not a menu of options — a clear default with named add-ons.
If you're still quoting custom prices for every client, you're spending sales cycles on pricing negotiations instead of onboarding. Standardization is the prerequisite for scale.
The agencies that grow past ten clients are the ones who figured out that their pricing is a product, not just a number. Standardization is the prerequisite for scale.
Voxfra handles the client management layer for voice AI agencies — portals, data separation, and reporting across all your clients. See how teams use it.