HomeBuilt with AIMy outreach system

Something I built · AI and automation

My outreach system

From finding the right companies to a tailored email draft and a tidy sales pipeline, with a person approving every email.

Build2026
A diagram of the outreach system in five steps: find companies, check the email, write a draft, I read and send it, HubSpot updates itself.

How it works, drawn for this page. No real data.

The company

I run my own sales, and I wanted outreach that is personal without taking hours a day. I built a system with Claude, n8n, Hunter, Gmail and HubSpot. Each step is automated, except the one that matters most: I read and send every email myself.

The brief

  • Find companies that match a clear profile, and rank them.
  • Write a first email that reads as if I wrote it for that company.
  • Keep the pipeline up to date without typing anything into the CRM.

Worked in

Clauden8nHunterOpenAIGmailHubSpotSlack

What I did

Finding the right companies

I describe the companies I want in plain words. Claude researches them, scores each one A, B or C on how well it fits, and adds them to a table. Only the best fits move on.

The right email address, checked

Hunter finds the founder's email address and checks that it works. Addresses it is not sure about are held back, so I do not email the wrong person or damage my sending domain.

A tailored draft, never an automatic send

AI reads the company's website and writes the personal part of the first email. The rest follows rules I wrote, in my own tone. The email lands in Gmail as a draft. I read it, change it if I want, and send it myself.

A pipeline that updates itself

When I send an email, a contact and a deal appear in HubSpot. When someone replies, the deal moves on. When a call lands in my calendar, it moves again. Every weekday at 8:30 a short email tells me who to write to, who to follow up and which calls are coming.

A sales assistant in Slack

Claudio is an AI agent I built in one day. I message it from my phone: “where is my pipeline?”, or “draft a follow-up to this client”. It reads HubSpot and my email and writes drafts. It never sends, and changes to a deal wait for my approval. It costs about $10 to $15 a month to run.

Inside it

What it looks like in use.

n8n · first email drafts
An n8n workflow: get founders ready for a first email, fetch and read their homepage, an AI step writes a personal line, then a check decides between saving a Gmail draft and flagging the row.
The workflow behind first emails. It reads the company’s homepage, AI writes one personal line, and the email is saved as a Gmail draft. If there is not enough to go on, it flags the company for me instead.
n8n · Gmail to HubSpot
An n8n workflow that searches Gmail, works out what changed, then adds contacts, creates deals or moves them to Replied in HubSpot, or marks a bounce.
Every evening this reads Gmail for emails I sent and replies I received, then creates or moves the deal in HubSpot. Nobody updates the CRM by hand.
Slack · Claudio
A Slack thread: Giovanni asks for the deals in his HubSpot pipeline and the sales agent replies with each deal grouped by stage.
Claudio, my sales assistant in Slack, answering “tell me the deals in my pipeline”. Every company and person name is made up for this page.

What came of it

Results.

Research, email finding and first drafts done by the system, with every email read and sent by me
Sent emails, replies and booked calls reach HubSpot without any typing
A to-do email every weekday at 8:30, and a sales assistant in Slack

Figures belong to this engagement and this client. Nothing here is an average or a benchmark, and no number appears without the work behind it described above.

What I took from it

Automation is good at the slow parts: research, checking and record keeping. The part that wins the reply, deciding what to say to one person, I keep for myself.

Giovanni Perri

Want one like this, for your team?

Tell me the job you want off your team’s plate. Most of these would be a Build, at a fixed price agreed before we start.