Portfolio Project · f1rstword

Know Who to Talk To, and When.

An account copilot I built for B2B sales. It finds, qualifies and briefs, so reps spend their time on people.

Last updated:

f1rstword logo
10 sec
Inbound lead to alert
40
Security and behaviour checks
GTM Systems

A GTM system that finds, qualifies and briefs

Sales reps lose hours on work that is not selling: researching companies, hunting contacts, updating the CRM and preparing for calls. I built the whole loop around one rule: outreach stays human, everything around it is automated and explainable.

TypePortfolio project
My roleDesigned and built end to end
StackSupabase, HubSpot, built with Claude Code
CodeOpen source on GitHub
1. The Problem

Good leads wait while the rep catches up.

Outbound at volume sends thousands of emails to the wrong people at the wrong time. What closes deals is a rep writing to the right person at the moment something changed. Meanwhile, the moments that matter get missed:

A company raises moneySales roles open upA new head of sales joinsA demo request goes cold
"Sales teams do not need more leads. They need the right lead, at the right time, with a real reason to reach out."
2. The Approach

Build the whole loop, keep the human in charge.

Find

Spots target companies, hiring posts and inbound leads.

Qualify

Checks fit and signals, then labels each Hot, Warm or Watching.

Act

Shows who to contact and why. The rep writes and sends.

Close the loop

A one tap debrief drafts the follow up and updates HubSpot.

3. What a Rep Sees

Six tabs, each answering one question.

Today

Who do I contact right now?

Accounts

Which companies fit, and why now?

Inbound

Is this new lead worth my time?

Meetings

What do I need before and after a call?

Research

Who is this company, and who works there?

Setup

Who am I selling to?

4. The Results

Measured on the live system.

Speed to lead
10 sec
Inbound form to alert

Enriched, scored and routed in about 10 to 12 seconds.

Research
10 sec
Company dossier

Any website becomes a dossier, with its people listed.

Discovery
9
New companies in the first run

It checked 15, added 9 and turned down 6 with a reason.

5. Decisions Worth Explaining

Why it works the way it does.

Outreach stays manual

The tool never contacts a prospect, which keeps it safe and trusted.

Fit and signal, not one score

One number mixed company fit with timing, and nobody trusted it.

Rules over AI

Fixed rules make every decision explainable. AI only writes short text.

HubSpot stays the record

Nobody has to learn a second tool.

6. Honest Limits

Built lean, and open about the gaps.

Headcount from the free data source is an estimate, phone numbers need a paid provider, and discovery leans on one public hiring source for now.

For technical readers: how it runs, stack and safety

Scheduled jobs

Every minuteMeeting briefs, "how did it go?" emails, cancel and move alerts, lead reminders
Every 30 minutesReads HubSpot back: completed to dos, deal changes, closed lost deals
Every morningReads sources again for followed companies, sends hot alerts, then emails the Today list and creates HubSpot to dos
Every MondayDiscovers new companies and runs the full weekly scan and scoring

Stack

Supabase PostgresEdge FunctionsScheduled jobsHubSpotClaude CodeHunterGoogle Calendar and GmailGitHub Pages

Safety

Public visitors only see masked names and emails. Every table has row level security, public endpoints are rate limited, and about 40 security and behaviour checks run against the live project. Details are in the README.

← Back to All Case Studies

Want a system like this?

I can design and automate the sales and RevOps setup that fits your team, your tools and your market.