An AI agent can involve a monthly subscription, usage charges, or an upfront investment in a custom build. Before comparing prices, I would establish exactly what the agent needs to do and what the supplier is responsible for delivering.
Bushe’s 2026 pricing guide provides useful estimates for that first budgeting conversation.
Typical costs
Bushe reports these estimates in US dollars:
- Website chat: tens to a few hundred dollars monthly.
- AI receptionist: roughly $29–$500 monthly.
- Voice platform you configure: about $0.05–$0.25 per minute, plus a phone line.
- Managed voice service: about $0.25–$0.50 per minute.
- Custom build: approximately $2,500–$45,000 upfront.
- Custom running costs: roughly $200–$1,000 monthly.
My perspective on these prices
These ranges align with what I’ve been seeing as the industry standard for this kind of work. I would use them as a starting benchmark, rather than assume every business should pay the same amount.
For me, the important distinction is between buying access to a tool and paying someone to deliver a working business process. When I assess a proposal, I want that distinction made clear: what am I receiving, who is responsible for it, and how will we know it works?
What I would expect in a professional quote
I would ask for a written proposal covering four points:
- A defined task. Specify what the agent will handle, what it must leave to a person, and which systems it needs to use.
- A clear cost breakdown. Separate the initial fee, recurring charges, included usage, and the price of exceeding those limits. Ask what maintenance and support cover.
- An escalation process. Agree how uncertain answers, failed bookings, and requests outside the agent’s scope reach a person. I would make this part of acceptance, rather than leave it for later.
- A measurable result. Choose a test that reflects the task. For appointment booking, I would check whether the right details reach the calendar and whether exceptions are handled correctly.
I would also ask the supplier to price a normal month and a busier month using the same assumptions. That gives me a more useful comparison than a headline subscription price alone.
Start with a focused pilot
My preference would be to begin with one clearly defined workflow, review its performance, and expand only when the result justifies it. For example, an appointment-booking pilot could have agreed test cases, a spending cap, and a review date before any wider rollout.
The decision I want to make is whether the quoted service can deliver a useful, dependable outcome at a cost the business can sustain. A professional proposal should give me enough detail to make that decision.
Discussion