How much does a custom AI agent cost?
A custom AI agent is priced per project between $300 and $9,000, then a monthly maintenance retainer if you want it kept running. The number moves with how many systems the agent touches, how many people use it, and how much of your data has to be cleaned before it can be trusted.
On this page
What sits at the bottom of the range and what at the top?
The bottom is one agent, one channel, one job, running inside a tool you already pay for. The top is a product with its own screens, its own sign in and its own users.
Most owners are surprised by how far down the range their first build sits. If the job is “answer every enquiry within minutes and hand the live ones to a named person”, and the enquiries already arrive in one place, that is a small build. The price climbs when the agent has to read three systems that disagree with each other, when several teams need different views of the same record, or when the data in the system today is too messy to act on without a cleanup first.
| Project shape | Where it sits in the range | What ships |
|---|---|---|
| A pilot on one channel | Bottom of $300 to $9,000 [1] | One working piece running on your real data |
| An agent inside tools you already run | Lower middle of the range [1] | Follow up, booking or reporting done without a person |
| A first version product with screens | Top of the range [1] | Something you can put real users into [2] |
What does it cost to run each month?
Two things, and only one of them is ours. The tools you already pay for keep their own subscriptions, and our monthly maintenance retainer is optional on top.
The retainer covers monitoring, fixes and small changes. It exists because the systems on the other end change without warning: a vendor renames a field or changes a login rule, and a connection that ran for six months stops. Somebody has to notice that, and the retainer is us being that somebody. If you stop it, the system keeps running on your accounts. You simply stop having us watch it.
We do not publish a retainer figure, because it follows the size of the system it looks after. It appears in the proposal next to the project price, before anything is signed. There is no hourly rate and no per seat charge anywhere in our pricing.
Why is a pilot the cheapest way to find out?
Because it costs the bottom of the range and it ends with something that actually runs on your data. You get a result to judge rather than a promise to trust.
This matters most to owners who paid an agency once and got a slide deck. A pilot inverts the risk: the smallest thing we will sell is a working piece, tested on your real leads, real orders or real bookings, with you watching. If it is wrong, you found out for the price of the smallest project rather than the price of the whole one.
It is also the honest way to discover the real scope. Half of what an owner describes in the first message turns out to be two problems wearing one coat, and a pilot separates them before either of us has committed to a number for the wrong build. What a pilot should prove, and the signs to walk away, are in how to test a provider with a pilot.
What does the proposal actually commit to?
One fixed number and one delivery date, written before any deal is signed. Not a range, not an estimate and not a rate card.
The full list of what moves the price is on the pricing page, what an agent actually is sits in what is an AI agent for a business, and the five steps of a project are on how we work.
Tell us the problem on the contact page, and you get a proposal and a price before any deal is signed.