Published price

AI integration. A model where it earns its place, wired into software you already run.

Document processing, assistants, extraction, classification, summarising and automated workflows, built into an existing product or a new one.

MVR 7,500 – 20,000. Prices are starting estimates. Final pricing depends on confirmed scope and requirements.

What it costs Ask about this

Prices in MVR

What it costs.

The range is the engineering, for a scope agreed in writing before anything starts. What decides where inside it a job lands is written out below.

Development MVR 7,500 – 20,000

Maintenance MVR 2,100 – 4,600/month

MVR 7,500

One well-defined task: summarise this, classify that, pull these fields out.

MVR 20,000

Several tasks chained into a workflow, with review steps, structured output and evaluation of whether it is actually right.

What moves it up:

How many distinct tasks the model has to perform
Whether the output has to be structured and validated
Whether a human needs to review or correct results
How much work goes into checking accuracy before it is trusted
Whether your own data has to be retrieved and fed in

Prices are starting estimates. Final pricing depends on confirmed scope and requirements. Third-party infrastructure and usage charges are billed to you by the provider, separately.

Scope

What is in it, and what is not.

Both halves are published, because a list of everything included says nothing about the things it quietly leaves out.

What's included

Provider and model selection for the job
Prompt and output design, with structured results where needed
Integration into your application and its data
Guardrails, error handling and fallbacks
Cost controls, so usage cannot run away quietly

What's not

AI provider usage charges
Provider accounts and approval
Model training or fine-tuning unless scoped separately

You pay the AI provider directly, on your own account, at their rates. Their usage charges are not our revenue and we do not mark them up. We will also tell you when a database query or a plain rule does the job better than a model.