# AI automation: fewer repetitive tasks, with control.

> Plain-text version for machine reading (LLMs). Original page: https://damian.dearce.com.uy/en/ai-consulting/ai-process-automation/

Generative AI can automate tasks that used to need a person reading: sorting emails, extracting invoice data, answering WhatsApp inquiries or preparing reports. I design those automations with human review where it matters and just the permissions they need.

_Updated: 2026-10-26_

## At a glance

- **Tools:** n8n, Make, Zapier, Power Automate and AI APIs
- **Channels:** Email, WhatsApp, forms, spreadsheets and your ERP or CRM
- **Agents:** Only with least privilege and a log of what they do
- **Control:** Human review on decisions with impact
- **Result:** Hours freed up, measured before and after

## What can be automated

- Email: sorting, summarizing and drafting replies for review.
- Invoices and receipts: reading the data and loading it into a spreadsheet or accounting system.
- WhatsApp: answering FAQs, taking orders or booking, handing over to a person when needed.
- Reports: pulling data from several sources and producing a weekly summary in plain language.
- Documents: filling proposal, contract or report templates from structured data.
- Internal help: an assistant that answers from the company's manuals and procedures.


## AI agents: when to use them and when not to

An AI agent does not just answer: it decides which steps to take and uses tools (email, files, systems). It is useful for multi-step tasks with variable judgment, but every permission it gets is a new risk. An agent that reads external email can be manipulated by a message with hidden instructions (prompt injection).

- Start with fixed workflows and add agents where they genuinely help.
- Give them only the permissions they need and separate read from write.
- Require human confirmation before paying, sending, deleting or publishing.
- Log what they do so it can be reviewed and corrected.


## How to choose the tool

n8n is open source and can be self-hosted, which helps when data must not leave the company. Make and Zapier are simpler to start with in the cloud. If the company runs on Microsoft 365, Power Automate integrates natively. We choose based on your systems, your data and who will maintain it.

## How I help

- Mapping repetitive processes and estimating the possible savings.
- Designing and building workflows with n8n, Make or Power Automate.
- Chatbots and assistants for WhatsApp or internal help.
- Security review: permissions, credentials, prompt injection and logging.

## FAQ

### What is n8n and what is it for?

It is an open-source automation tool that connects applications (email, spreadsheets, CRM, WhatsApp, AI models) into workflows. It can run in the cloud or be self-hosted so data does not leave the company.

### Can I put an AI chatbot on my company's WhatsApp?

Yes, through the official WhatsApp Business API. What matters is defining what it answers on its own, when it hands over to a person and what customer data it stores.

### Are AI automations safe?

They are when designed with least privilege, properly stored credentials, human review on important decisions and a log of what they do. Without that, an automation can leak data or be manipulated.

[Let's discuss your case](https://wa.me/59899231327?text=Hi%20Dami%C3%A1n%2C%20I%27d%20like%20to%20ask%20you%20about%20AI%20process%20automation.)

## Contact

- WhatsApp: +598 99 231 327 (https://wa.me/59899231327)
- Email: damian.de.arce@gmail.com
- LinkedIn: https://www.linkedin.com/in/damiandearce/