AI projects vary because the surrounding software, data and operational requirements vary. Our proposal separates the implementation work from the services and support needed after launch.
This page explains the buying options and cost drivers. It is not a fixed-price tariff or a promise that every project fits the same package.
Four ways to engage
Project blueprint.
A defined discovery engagement covering the problem, workflow, architecture, data, acceptance criteria and next-stage options. Useful when material questions remain unresolved.
Working pilot.
One bounded use case, a stated dataset or integration scope, a review process and a clear decision about continuation. A pilot is not automatically a fully supported production application.
Production application or integration.
The agreed software, AI behaviour, interfaces, deployment work, evaluation and handover, delivered through milestones.
Operate and improve.
A defined agreement for monitoring, maintenance, regression checks, incident handling and planned changes. Coverage, response arrangements and exclusions are agreed explicitly.
Implementation · one-off
What changes implementation cost?
The principal questions are how much software must be built, how many integrations are involved, how usable the data is, how permissions work, what must be evaluated and what the release must support.
A narrow feature in an established application differs from a new product with accounts, multiple roles, document processing and an administration interface. Poorly documented systems or difficult source data add uncertainty that should be resolved or priced openly.
Operation · ongoing
What continues after launch?
Compare model usage, hosting, storage, search/indexing, document processing, monitoring, third-party licences, human review and maintenance. Some items vary with usage; others are fixed or depend on service coverage.
A low token bill does not represent the full cost of operating a business application. Conversely, not every project needs a large recurring support arrangement. The scope should match the system and your team's responsibilities.
How we prepare a proposal
We identify the agreed result, required work, dependencies and uncertainty. The proposal explains the fee structure, milestone amounts, payment terms and assumptions. If an early estimate is conditional, those conditions remain visible.
For a fixed scope, the change process distinguishes defects from new requirements. Where discovery is still needed, we propose a bounded stage instead of presenting an unsupported final price.
How to compare suppliers
Ask whether frontend, backend, data preparation, integration, testing, deployment and handover are included. Compare the same deliverables and acceptance criteria. Check who pays third-party charges and who owns operational support.
Source-code delivery, rights to reuse components and a runnable deployment are separate questions. A proposal should answer all relevant questions rather than leave them behind a single “AI development” line item.