Teams with repetitive tasks
They copy data between systems, sort emails or documents, or prepare reports by hand.
Automation and AI agents
We start with the process, not the tool: where time is lost, which data is involved and which decisions must stay with a person.

The problem
Many companies try AI tools and get impressive demos that never reach daily work: they do not connect to their systems, nobody reviews what they produce or the data is not ready.
Our approach is the same as any serious development: understand the process, measure how it works today, automate the part that makes sense and keep human oversight where there is risk. Sometimes the best solution is an integration with no AI at all.
Who it is for
They copy data between systems, sort emails or documents, or prepare reports by hand.
Catalogues, procedures or documentation their team constantly looks up.
Shop, ERP, CRM and forms that do not talk to each other.
Solutions and deliverables
The examples below describe technical capabilities we can implement. They are not client cases unless stated.
Capability example: reading invoices, delivery notes or requests, extracting fields and proposing a classification for a person to confirm.
Capability example: an internal assistant that answers from the company’s procedures or catalogue, citing the source and not making things up when it cannot find it.
Delivered project: Ferretería Dols’ shop automatically synchronises products, prices and stock with its ERP and the COINFER API.
Capability example: forms that create CRM records, notify the team and create follow-up tasks with no manual work.
Capability example: an agent that drafts replies or tasks and leaves them pending approval.
Grounded
Works with sample data to check whether the idea is viable. It helps decide; it is not for production.
Connected to your real systems, with error handling, activity logs, permissions, defined human oversight, documentation and maintenance.
We do not promise time savings or percentages before measuring the process. If a target is set, we measure before and after.
Process
Which steps exist, who does them, with which data and at what volume. We identify what to automate and what not to.
A pilot with anonymised or sample real data to measure quality and errors.
Integration with your systems, human reviews, logs and alerts.
Reviewing results, adjusting and maintaining. Models and APIs change and need monitoring.
Explaining the process with real examples, providing test data, defining who supervises and validating the pilot results.
Related work
Real cases where we have done this type of work. Each case explains our exact involvement.
Scope and exclusions
Engagement
Assessment, scoped pilot and production rollout. An assessment with its own quote, a scoped pilot and, if it works, phased implementation. Cost depends on the volume and variety of cases, the systems to connect, data quality and the level of oversight.
Frequently asked questions
Not exclusively. We are a design and technology studio that integrates AI where it adds value to a specific process, alongside development, integrations and maintenance.
We assess it before starting: which data is sent, to which provider, under what terms and whether it should be anonymised or handled with alternatives. It is documented in the proposal.
We do not know until we measure the process. If we set a target, we measure before and after the pilot.
Low-risk tasks can be automated further. Where there is an impact on clients, money or personal data, we keep human review.
Related services
Next step
Describe it with a real example. We will tell you whether automating it makes sense and how we would test it.