A real estate agency in Málaga: coaching on Claude, AI agents and automation

The agency we have in mind

Picture an agency of five people on the Costa del Sol: two agents, one person for rentals, an office manager and the owner. Buyers write in English, Spanish, French or German. Leads arrive from property portals, the website, Instagram and WhatsApp.

A normal week looks like this:

• the same lead answered twice, or not at all, because it arrived on a weekend;
• listing descriptions rewritten by hand for every language and every portal;
• visits booked by message, then confirmed again by phone;
• owners asking "any news?" and someone building a report from memory;
• documents for the notary chased one email at a time.

Nobody in this team lacks AI tools. Most of them already have ChatGPT or Claude on their phone. What they lack is a shared way to use them, and the connections that remove the copy-paste.

Step 0: choose the KPIs before anything else

Before the first session, we agree with the owner on three or four measures and we record where they stand today. For an agency, the useful ones are usually:

• lead response time: how long between a portal enquiry and the first real reply;
• visits booked per week from new leads;
• hours spent on admin per person: listings, reports, reminders, documents;
• leads without a follow-up after seven days.

We measure them again after eight weeks. That is the only honest way to talk about results: the figure belongs to the agency, and it depends on its volume and its market.

Step 1: coaching the team on Claude

The first sessions are about the writing the team does every day. Each person brings real examples.

• Listings in several languages. From the agent's notes and the property sheet, Claude drafts the description in English, Spanish and French, in the agency's tone. The agent checks surface areas, prices and legal mentions before publishing.
• Replies to enquiries. A short prompt with the role, the context (the property, the buyer's question), the task, the format and a check: "list anything you are unsure about".
• A shared Claude project. The agency's style guide, its areas, its fees policy and approved templates live in one place, so everyone gets the same quality.
• Data rules. What can go into an AI tool (property details, anonymised questions) and what never does (passport or NIE numbers, bank details, signed contracts).

At the end of this step, the team has a prompt playbook built on its own work, not a generic course.

Step 2: one AI agent, with clear limits

Next, we build a single agent for incoming leads. Its job is narrow:

• It can: read the enquiry, find the property in the listings sheet, detect the language, draft a reply and suggest two visit slots from the agent's calendar.
• It cannot: send anything on its own, discuss the price, promise availability, or share documents.

Every draft goes to the agent on duty, who approves, edits or rejects it. Every action is logged. If the agent is unsure about the property or the request, it says so in the draft instead of guessing.

Step 3: automations for the follow-ups

The rest is plain automation in n8n, with no AI needed:

1. Lead intake: enquiry emails from the portals and the website form go into one sheet or CRM, with the source and the property.
2. Alert: the agent on duty gets a WhatsApp or Teams message with the lead and the AI draft.
3. Reminder: if a lead has no reply after the agreed delay, the owner is alerted.
4. Visit confirmation: once a visit is booked, the buyer receives a confirmation and a reminder the day before.
5. Owner report: every Friday, a summary per property: enquiries, visits, feedback.

For the step after the offer, our deal to notary workflow tracks the documents until the file is ready.

Step 4: measure and adjust

After eight weeks, we compare the KPIs with the starting point. Where the figure moved, we keep going. Where it did not, we look at why: a prompt the team does not use, a reminder sent too late, a step that still depends on someone's memory.

We do not promise a percentage in advance. We promise that the agency will know exactly what changed, measured on its own figures.

Is your agency in Málaga or on the Costa del Sol?

We coach teams in English, Spanish and French, on the tools they already use: Claude, ChatGPT, Gmail or Outlook, Sheets or Excel, WhatsApp and their CRM.

Tell us how your agency works, or see our practical AI training.