Short answer: BYOK ("bring your own keys") platforms advertise a low base fee (~£0.05/min) but you pay separately for speech-to-text, the language model and text-to-speech, so the real cost is usually £0.11–£0.33/min. All-inclusive platforms bundle everything at a higher headline per-minute or flat rate. For a non-technical business taking a few hundred calls a month, all-inclusive (ideally a flat plan) is usually cheaper once you count engineering time.
The AI voice agent market has quietly split into two pricing philosophies, and comparing the headline numbers without understanding them will lead you to the wrong choice. Here's how each model works, and how to tell which one wins for your front desk.
The two pricing models
BYOK (bring your own keys)
Developer-first platforms like Vapi and Retell charge a low base orchestration fee and let you plug in your own components. Vapi's base is around £0.05/min; Retell is similar. But you then pay for three separate providers on top: speech-to-text (e.g. Deepgram), the language model (e.g. GPT-4o or Claude), and text-to-speech (e.g. ElevenLabs). A basic build lands near £0.11–£0.15/min; switch to premium voices and models and you're at £0.25–£0.33/min. You also own the wiring, monitoring and maintenance.
All-inclusive
Platforms like Synthflow and Bland — and most done-for-you receptionist products — bundle every component into one price. Synthflow moved to a fully pay-as-you-go model in 2026, metered per second on successful calls only, landing most setups at roughly £0.15–£0.24/min. The per-minute number looks higher, but there are no separate provider bills and nothing to build.
A worked example: a 400-call-a-month front desk
Assume 400 calls a month averaging 3 minutes = 1,200 minutes.
| Model | Effective rate | Monthly call cost | Plus |
|---|---|---|---|
| BYOK, basic components | ~£0.13/min | ~£156 | Your setup + maintenance time |
| BYOK, premium voice/model | ~£0.29/min | ~£348 | Your setup + maintenance time |
| All-inclusive PAYG | ~£0.20/min | ~£240 | Nothing to build |
| All-inclusive flat plan | bundled | ~£150–£300 fixed | Predictable; best at volume |
The headline "£0.05/min" of BYOK never actually happens — that's just the orchestration slice. And the numbers above ignore the biggest hidden cost of BYOK for a non-technical business: your time, or a developer's, to build and keep it running.
The decision rule
- Choose BYOK if you have an engineering team, want full control of the voice/model stack, and are running high, predictable volume where squeezing per-minute cost matters.
- Choose all-inclusive if you're a clinic, trades business, MOT centre or care home that wants it to just work, with one bill and someone else owning reliability.
- Choose a flat all-inclusive plan if you take 200+ calls a month and want a predictable number you can forecast against.
For the full budgeting picture — including setup fees and the year-two renewal hike — see what an AI receptionist actually costs in 2026.
Frequently asked questions
What is BYOK pricing?
BYOK ("bring your own keys") means the platform charges a low base fee (~£0.05/min) and you connect and pay separately for speech-to-text, the language model and text-to-speech. Real all-in cost is usually £0.11–£0.33/min.
Is BYOK cheaper than all-inclusive?
Not usually for a non-technical business. Once you add component costs and the time to build and maintain it, an all-inclusive flat plan is typically cheaper and much simpler.
Hear it answer your phone
Book a 20-minute demo and hear an AI receptionist take a real call for a business like yours.
Book a demo →