How I Signed 5 Clients Using Claude Agents

I signed five new real estate clients in 47 days using Claude agents. Not by working more hours — I actually worked fewer. If you’d told me that in early 2026 I would be running automated research, outreach, and follow-up sequences from a single Claude project while sipping espresso on my terrace in Madeira, I would have called you optimistic. But that’s exactly what happened, and I want to walk you through the exact process so you can replicate it — or at least avoid the mistakes I made getting there.

I run a solo real estate consultancy. No team, no virtual assistants, no junior agents. Just me, a CRM, and whatever tools I can stitch together to punch above my weight. Client acquisition has always been my biggest time drain — not because the work is complicated, but because it requires consistency I can’t always deliver when I’m juggling active transactions.

The Problem With Solo Client Acquisition in Real Estate

Here’s the honest picture of my pipeline before I set this up. I had a list of roughly 340 warm leads — people who had inquired about Madeira properties at some point in the past two years but never converted. Some filled out forms. Some emailed once. A few had calls with me and then went quiet. Every week I told myself I’d follow up with 20 of them. Most weeks I followed up with four or five, usually the same ones I remembered off the top of my head.

The leads weren’t dead. They were just neglected. And I was losing deals to other consultants not because I offered less, but because I was slower to stay visible.

I needed a system that could do three things without me babysitting it: research each lead before I contacted them, draft personalized outreach, and schedule follow-up reminders with context I could act on fast. That’s what I built with Claude agents — and it took about two weekends to get working properly.

What Claude Agents Actually Are (And How I Set Mine Up)

If you’re not familiar with the term: Claude agents are essentially Claude running inside a project with a persistent system prompt, access to uploaded documents, and — depending on your setup — the ability to call external tools via the API or through platforms like Make.com. They’re not magic. They’re structured workflows where Claude acts as the reasoning engine.

My setup used three separate Claude agents, each with a distinct job:

• Agent 1 — The Researcher: Given a lead’s name, email, and any notes from my CRM, this agent would draft a research brief. I fed it property market data for Madeira (uploaded as PDFs), recent listings, and a template for what “useful context” looks like for a real estate conversation. Output: a 150-200 word brief per lead.
• Agent 2 — The Outreach Writer: Took the research brief plus my personal voice guidelines (I spent about 90 minutes writing a detailed style document) and produced a first-contact email. Not a template — an actual personalized email referencing the lead’s original inquiry, current market conditions, and one specific property or neighborhood relevant to them.
• Agent 3 — The Follow-Up Sequencer: After I sent the initial email, I’d paste the response (or note that there was no response) into this agent, and it would generate the next follow-up message plus recommend a send timing. It also flagged leads it classified as “high intent” based on language patterns in any replies.

The whole stack runs on Claude Pro ($20/month) for the agents themselves, plus a Make.com scenario (I’m on the Core plan at $9/month) that handles the trigger from my CRM and formats the input before it reaches Claude. Total monthly cost: $29.

My Real-World Experience: 47 Days, 5 Signed Clients

I started running the full system on a Monday in February 2026. I pulled 80 leads from my “dormant” list — people who hadn’t responded to anything in at least 60 days — and fed them through Agent 1 and Agent 2 over three evenings. Each batch of 20 leads took me about 35 minutes to process: paste inputs, review outputs, make small edits, send. Before this system, writing 20 personalized emails from scratch took me closer to 3.5 hours.

That’s the number that still gets me: 3.5 hours down to 35 minutes for the same volume of outreach. I recovered roughly 2 hours and 55 minutes per batch. Over the 47-day test period, I ran six batches. That’s approximately 17.5 hours I got back — hours I used to take on two additional property viewings per week and do the market analysis reports I’d been putting off since November.

Of the 80 leads I contacted, 23 replied within two weeks. That’s a 28.75% reply rate on dormant leads, which is honestly better than I expected. From those 23 replies, 11 moved to calls. Of those 11 calls, 5 signed consulting agreements within the 47-day window. One more signed on day 51, so technically 6 — but I’m only counting the 47-day window to be precise.

The clients who signed weren’t the ones I expected either. Three of them had gone quiet for over six months. One had inquired about a property category I no longer focused on, but Agent 2 caught a detail in his original email about a secondary interest that I’d missed entirely. That deal alone was worth €4,200 in commission. I would have skipped that lead entirely in my old manual process because his primary request felt like a dead end.

The follow-up sequencer was the piece I underestimated. Three of the five clients required three or more touchpoints before they responded. Agent 3 kept those threads alive with context I would have lost track of manually. One client specifically mentioned in our call that she appreciated how I’d remembered a detail she mentioned months earlier. That detail came from Agent 1’s research brief — I hadn’t remembered it at all. The system had.

Exact Workflow: Step by Step

Here’s the process I actually followed, stripped of any theoretical padding:

1. Export leads from CRM — I use a basic HubSpot Free setup. I filter for leads tagged “dormant” with last contact date over 60 days ago. Export to CSV.
2. Trigger Make.com scenario — The scenario parses each row and formats it into a structured input block that Claude can read cleanly. This took me about 4 hours to build, mostly because I kept getting the JSON formatting wrong.
3. Run Agent 1 (Research Brief) — I paste the formatted input into the Claude project. The agent has my Madeira market data, my positioning documents, and a clear instruction to produce a brief with: original inquiry summary, current relevant listings, one local market insight, and one suggested conversation angle.
4. Run Agent 2 (Draft Email) — I paste the research brief. The agent produces the email. I read it, usually make one or two small tweaks, then send directly from Gmail.
5. Log send in CRM, set reminder for 5 days — I do this manually. Still haven’t automated this part and I should.
6. Run Agent 3 on replies or non-replies — For replies, I paste the full thread. For non-replies at day 5, I note “no response, day 5” and the agent generates a soft follow-up. The sequencer handles up to 4 touches before flagging a lead as “rest for 30 days.”

What the System Does Not Do Well

I want to be direct about the failures, because there were real ones.

Agent 1’s research is only as good as the documents I’ve fed it. When I was working with leads interested in rural quinta properties — a niche I don’t focus on often — the briefs were shallow because I hadn’t uploaded good reference material for that segment. Two emails went out that were technically fine but missed the specific concerns quintas buyers usually have around agricultural classifications and renovation permits. Neither lead converted. That’s on my setup, not Claude, but it’s a real limitation of the document-dependent approach.

Agent 2 occasionally produces emails that are slightly too polished. Real estate outreach in Portugal walks a cultural line — too formal feels cold, too casual feels unprofessional. About 15% of the drafts needed significant tone adjustments before I’d send them. I’ve since added more examples to my style document and the rate has dropped, but it hasn’t disappeared.

The Make.com trigger setup requires some technical patience. If you’ve never worked with APIs or structured data formatting, plan for a steep first weekend. The Claude part is easy. The plumbing between your CRM and Claude is where most people will get stuck.

And finally: this system does not replace good judgment on which leads to prioritize. I made the mistake in week two of running a batch of genuinely cold leads — people who had never really engaged — through the same process. The reply rate was 6%. The system performs best on warm-but-dormant contacts who had real intent at some point.

Results at a Glance

Metric	Before Claude Agents	After Claude Agents
Outreach emails per week	4–5	18–22
Time per 20 emails	3.5 hours	35 minutes
Reply rate (dormant leads)	~8%	28.75%
Clients signed (47 days)	1–2 typically	5
Monthly tool cost	$0 (manual)	$29
Hours recovered per month	—	~17.5 hours

What I’d Do Differently If Starting Over

Build the style document first. I spent more time fixing tone in Agent 2’s emails than I should have because I rushed the voice guidelines. A good style document — with 8 to 10 real email examples of your own writing, annotated with what you like about each one — will cut your editing time in half from day one.

Segment leads before running them through. Warm-but-dormant is the sweet spot. Genuinely cold leads need a different approach and probably a different agent prompt entirely.

Automate the CRM logging step. I’m still doing this manually and it probably costs me 20 minutes per batch. That’s the next thing I’m fixing.

Start with 20 leads, not 80. I ran too large a batch on the first try and got overwhelmed reviewing outputs. Twenty leads per session is the right pace for a solo operator who still needs to edit and think.

Practical Summary

If you run a solo service business and you have a dormant lead list sitting in a CRM doing nothing, a three-agent Claude setup is one of the highest-ROI things you can build in 2026. My total setup time was roughly 12 hours across two weekends. My monthly ongoing cost is $29. The return in the first 47 days was five signed clients and 17.5 hours recovered.

It’s not a set-and-forget system — you still read every email before it goes out, and you still make judgment calls about which leads to pursue. But the research, the drafting, and the follow-up sequencing are no longer eating your week. That shift alone changes what’s possible when you’re running everything alone.

The biggest mistake solopreneurs make with AI tools is treating them like magic wands. These agents are structured workflows. The quality of your inputs — your style document, your market data, your lead notes — determines the quality of the outputs. Garbage in, garbage out, no matter how good the model is.

If you want to build something similar, start with Agent 2 — the outreach writer. It’s the one with the most immediate, visible impact. Get that working well with 10 leads first. Then add the researcher. Then the sequencer. That order made the most sense for me, and I think it will for most solo operators.

Ready to build your own lead reactivation system? Start by writing your style document this week — 8 examples of emails you’ve actually sent that got responses. That’s the foundation everything else runs on. Drop a comment below if you want me to share the exact system prompt I use for Agent 2.