Skip to content
See all services
Services

Automation and artificial intelligence

Typical timeline:4 to 10 weeks

The problem

Your team answers the same questions on WhatsApp every day, copies data from one place to another, and builds reports by hand. Those hours show up in no metric, but you pay for them all the same. And when people get tired, replies start arriving late or steps get skipped.

Who it is for

Operations with repetitive volume: customer service over WhatsApp, orders, booking, or teams spending hours moving data between systems and writing the same reports every week.

The official API, not a workaround

Chatbots are built on the **official WhatsApp Business API**. Solutions that wire up a personal number through unofficial channels work right up until Meta blocks them, usually at the worst possible moment. The official API requires verifying the business and respecting its message templates, and in exchange it does not vanish overnight.

What I do differently with AI

An assistant that invents answers is worse than no assistant: it destroys the customer's trust within a single conversation. So the assistant answers **from your documents and your data**, and when it does not know something it says so and hands over to a person. I would rather have an assistant that admits its limits than one that sounds convincing and is wrong.

I am also blunt about what is **not** worth automating. Some processes are still cheaper and more reliable with a person doing them, and saying so is part of the job.

How we work

  1. Discovery

    Which tasks repeat, how many hours they consume, and which are actually worth automating. Automating a broken process only breaks it faster.

  2. Proposal and scope

    What gets automated, with what technology, and what stays out. Including estimated running costs, because AI models are billed by usage.

  3. Architecture

    Where the automation lives, how it reaches your data, and what happens when it fails. An assistant with no failure plan ends up inventing answers.

  4. Build

    Short iterations with your real cases. An assistant is tuned with the questions your people actually ask, not with laboratory examples.

  5. Handover

    Documentation, a usage dashboard, and transfer. Your team must be able to adjust replies without depending on me for every wording change.