Title: AI system maintenance and optimization | ARKEA IA
Description: We maintain and optimize AI systems after launch: token usage, behavior, integrations, models and user experience.

# Implementation does not end on launch day.

Models, APIs, tools and processes change. Maintenance lets us observe the system, resolve issues and optimize it as the business evolves.

[Discuss maintenance](contacto.html#proyecto)
## Token usage optimization

We review how context is built and used to reduce unnecessary information, redundant calls and consumption that does not improve the result.

Depending on architecture, this can include context and prompt design, model selection and routing, caching when viable, retrieval precision, result reuse and reducing unnecessary calls. We do not promise a universal savings percentage.

## What maintenance can include

### Monitoring

Review of relevant operational signals.

### Incidents

Diagnosis and correction of system issues.

### Integrations

Adaptations when APIs or tools change.

### Tools & skills

Adjustments as requirements evolve.

### Models

Evaluation or changes when they create a real improvement.

### Usage

Optimization of context, calls and resource consumption.

### UX

Improvements based on real use and feedback.

### Evolution

Evaluation of new needs after launch.

## Operate and evolve

[See control dashboard](dashboard-ia.html)
[Understand the implementation cycle](como-trabajamos.html)
## Frequently asked questions

### How do you control token usage?

We review context, retrieval, model selection, calls and reuse to avoid unnecessary consumption without degrading the required result.

### What happens when an API or model changes?

Maintenance can include adjustments to integrations, tools, skills, models and user experience so the system remains operational as its environment changes.