Pricing
We build each agent around your workflow, so the price depends on the workflow. Here is exactly how it is put together and what moves it — so you can judge whether this is worth a conversation before you have one.
One-off
Scoping the agent, writing it around your process, connecting it to your systems, testing it on real conversations, and launching it. Quoted as a fixed fee once we understand the job, so there is no meter running while we work.
A single well-scoped use case typically takes two to four weeks.
Ongoing
You pay for the time the agent spends actually talking to people. Industry rates generally fall between roughly $0.10 and $0.25 per minute depending on language, voice quality and volume.
No minimum seats, no per-user licence. An agent that handles fewer calls costs less.
Worth reading before any vendor conversation, ours included. These are the variables that separate a two-week project from a two-month one.
An agent that only talks is cheap. One that reads your calendar, writes to your CRM and files tickets in your helpdesk is three separate pieces of work.
A second language is not double the work, but it is not free either — it needs its own voice, its own testing, and its own knowledge base review.
If your answers are already written down, the agent has something to work from. If they live in two people’s heads, writing them down is part of the project.
Routing everything to one queue is simple. Routing by intent, language, region and working hours takes longer to get right.
Volume does not change the build much, but it changes your per-minute rate and how much infrastructure work is worth doing up front.
Some clients launch and leave it. Others tune weekly for months. Both are fine, but they are different commitments.
Per-minute cost is the easy figure to compare. The one that actually determines return is how many calls the agent resolves without a person.
Published enterprise benchmarks put median containment near 41%, with the strongest quartile around 59%. Vendors routinely advertise 70 to 80%. A cheap agent that resolves nothing is not cheap, and an expensive one that takes half your repetitive calls usually pays for itself quickly.
We would rather quote the number you are likely to hit and then beat it.
Custom builds are priced per project, because the work varies enormously: a single-purpose agent reading from an existing FAQ is a very different job from one integrated with three internal systems in two languages. We scope it on a discovery call and quote a fixed build fee, then per-minute usage on top.
Because a fixed price for custom work is either padded to cover the worst case or quietly re-quoted once the real scope appears. Neither serves you. We would rather spend thirty minutes understanding the job and give you a number we will stand behind.
Mainly four things: how many systems the agent must connect to, how many languages it handles, whether your knowledge base already exists in writing, and how complex the escalation rules are. A single use case with clear content is the cheapest possible start, which is why we recommend it.
You pay for the minutes the agent actually spends on calls. Industry pricing generally sits between roughly $0.10 and $0.25 a minute depending on language, voice quality and volume. Higher volumes reduce the rate; premium voices and harder languages increase it.
One clearly scoped use case — usually your most repetitive call type — built, tested on real conversations, and launched narrowly. It is the cheapest way to find out whether an agent works for your business before committing to anything wider.
Usage is the main one. Changes to the agent as your business changes are quoted separately or covered by a support arrangement, depending on how much you expect to iterate. We will be explicit about which applies before you sign anything.
Tell us the calls you want handled, the systems involved and the languages your customers use. We will come back with a fixed build fee, a per-minute rate, and an honest view of whether it is worth doing at all.