Why we don't list prices
Two businesses asking for "an AI agent for customer support" can need completely different builds. One connects to a single help desk and answers order-status questions. The other touches five systems, handles refunds, and has to satisfy a compliance team. Same request, different project. A published price list would either be so vague it's useless or so specific it's wrong for most people who read it.
So we scope first. That costs you nothing and usually takes one call plus a short follow-up look at the workflow. You get a number, and the reasoning behind it, before you owe us anything. The full sequence is on our process page — discover, design, build, run.
What actually drives the cost
Every quote we give comes down to a handful of variables. In rough order of how much they move the number:
- How many systems it touches. A build that reads from one CRM and writes back to it is simpler than one that has to reconcile data across a CRM, a billing platform, and a scheduling tool. Integration work — especially with older systems that don't have a clean API — is usually the biggest line item.
- How much judgement the system needs. A workflow that follows the same steps every time is cheaper to build and cheaper to trust than one that has to weigh context and make a call. The more a system has to decide rather than execute, the more testing and guardrail work it needs before it's safe to run unattended.
- Volume. A tool that processes a handful of cases a week can tolerate a simpler design than one processing hundreds a day, where edge cases show up constantly and performance under load starts to matter.
- Regulatory exposure. Handling health records, financial data, or anything else that sits under a compliance regime adds review, logging, and access-control work that a purely internal tool doesn't need.
- The state of your data. Clean, structured, well-labelled data is faster to build against than records that are inconsistent or scattered across systems that don't agree with each other. If your data needs work first, we'll scope that separately rather than quietly folding it into the build price.
Our guide to what a custom AI agent costs walks through how these factors interact with real examples, if you want more detail before a call.
What's included, and what's billed separately
Every scope we send breaks the project into what's inside the build price and what sits outside it. As a rule:
Included in the build: discovery and mapping the workflow, the design of the system and its guardrails, development, testing against your real cases, and getting it live.
Usually billed separately: data cleanup if your records aren't in a fit state to build against, new integrations added after the scope is agreed, and ongoing running costs once the system is live — because those depend on volume and are genuinely hard to predict up front.
We flag anything that looks likely to fall into the second category during scoping, not after the invoice arrives. If something changes mid-build that would move the price, you hear about it before we do the work, not after.
How the running and maintenance side works
An AI system is not a one-off purchase. Models get updated, the tools it connects to change their APIs, and your business evolves — new products, new policies, new edge cases the system has never seen. Running costs typically cover model and infrastructure usage that scales with volume, monitoring so we can see what the system is doing and catch drift early, and periodic tuning as your workflows shift.
We size this with you before launch so it's a known ongoing figure, not a surprise. Our guide to AI maintenance costs breaks down what typically drives that number up or down. If you'd rather build something narrower and keep running costs lower, we'll tell you that too — a smaller system that actually gets used beats a large one that gets switched off.
What you get in writing before you commit
Before any build work starts, you receive a written scope covering what the system will do, what it will explicitly not do, which systems it touches, how we'll measure whether it's working, the build price, and a running-cost estimate. Nothing starts until you've agreed that document. If your needs change once we're underway, we scope the change and confirm the new price before touching it — we don't expand a project quietly and bill for it later.
This applies across everything we build, whether it's a custom AI agent handling a specific workflow or a broader automation project. See the full range of what we build if you're still working out which shape fits your problem.