Home/AI App Development/AI App Cost
AI App Development Cost
What an AI app costs,
including the bit others hide.
An AI app has two costs: the fixed price to build it, and the ongoing inference, the per-call cost of the model, that scales with usage. Most quotes mention only the first. Here is the honest breakdown of both, what drives them, and how a well-architected build keeps the running cost from eating your margin.
The two costs
Build cost is fixed.
Running cost is not.
The build cost is a one-time fixed price, what you pay us to build the app. The inference cost is ongoing and variable, what you pay the model provider every time your app uses the AI. The first is predictable; the second scales with your users and has to be designed for. Confusing the two is how AI budgets go wrong.
The build cost
What you pay to build the app
The build cost is the straightforward part, and we keep it fixed and transparent like every Wall & Fifth build. A focused AI app, a grounded assistant, a RAG product, a single-purpose agent, is from 16,000 GBP, delivered in eight weeks. A larger AI build, with agentic workflows, multiple user types, an admin panel, and integrations, is 30,000 GBP. Fixed price, no hourly creep, full code ownership.
What moves a build between those tiers is the same as any product, plus the AI-specific complexity: how much grounding and retrieval is involved, whether it is a single AI feature or true agentic behaviour across systems, how many data sources, how much evaluation the use case demands. We scope it precisely before committing the number, so the build price is known up front.
That fixed build cost is comparable to a normal app, because under the AI layer it is a normal production app, the AI adds engineering but a well-scoped build does not balloon into a different order of magnitude. The thing that makes AI costs feel unpredictable is not the build. It is the second cost, the one most quotes stay quiet about.
The inference cost
The running cost most quotes hide
Every time your app uses the AI, it calls the model provider, and that call costs money. This is inference, and it is the cost that surprises founders, because unlike the build it is ongoing and it scales with usage. A handful of test users cost almost nothing. Thousands of active users each making many AI calls a day is a real monthly bill, and it grows exactly as your product succeeds.
We handle this honestly: inference is passed through to you directly, at cost, never marked up. It is your usage and your cost, and dressing it up in a markup would be both dishonest and a misaligned incentive. You pay the provider's price for what your app actually uses, and we make that visible rather than burying it.
Crucially, the inference cost is not fixed by nature, it is a design outcome. A naive build sends everything to the most expensive model on every call and runs up a bill many times larger than necessary. A well-architected one caches repeated work, uses retrieval so the model reads less, and right-sizes the model to each task. We build for low inference from the start, because we pay our own on sellyourboat.io and know exactly how much architecture moves that number.
The honest comparison
Reading an AI quote properly
When you compare AI build quotes, the build price is rarely the real difference, the inference design is. Two studios can quote a similar build cost, and one delivers an app that costs five times as much to run because it was architected without a thought for inference. That running-cost gap dwarfs any difference in the build price over the life of the product, and it is invisible at quote time.
So the questions to ask an AI builder are not only what does it cost to build, but how will you keep the inference cost down, do you pass it through at cost or mark it up, and what happens to my running cost as I scale. A builder who cannot answer those clearly has not thought about the cost that will actually dominate your AI budget.
Our answer is consistent: fixed build price from 16,000 GBP, inference passed through at cost, and an architecture designed to keep that inference low because we run a product on the same principles. The total cost of an AI app is the build plus the running cost, and we are honest about both, because the running cost is the one that decides whether the product is viable at scale.
FAQ
Questions, answered straight.
How much does an AI app cost to build?
What is inference cost?
Do you mark up the AI API cost?
Why do AI build quotes vary so much?
How do you keep the running cost down?
What questions should I ask an AI builder about cost?
Keep reading
Ready
Know both costs
before you build.
Tell us what you want to build and we will give you a fixed build price and an honest view of the inference cost, before you commit.