“How to find and reach new clients without searching all day?”

See what we built
companies screened in 2 weeks
~509
first emails
~173
replies
9
reply rate on cold email
~11%
the agent hands it to a salesperson in the CRM when it’s time to call
Hot lead

The starting point

There are plenty of lead and cold email tools on the market, such as Apollo, Lemlist or Instantly. They have the infrastructure, but they lack what matters most: knowledge of the market and a message that catches a company at the right time. There is no one-size-fits-all solution.

Goal: cold email built on market signals that finds companies, vets them, reaches out and hands warm contacts to salespeople.

Why our own agent

Three ways to find new clients.

Manual researchThird-party toolAgent in Claude
Timehours a daylessreviewing results
Lead qualitygood, but slowgenericsignals from multiple sources
Emailswritten by handtemplatesbased on the signal and the service
Follow-upsforgottenautomaticlearns from replies

What we built

  1. Market signals

    Locations, FinStat and company registers, job boards, social media comments and a website scan. Every company gets a score.

  2. Research in Sweden

    allabolag.se and partner agencies.

  3. Creative by service

    Each service has its own angle. The message builds on a specific signal from the company and goes into Pipedrive as a draft.

  4. Sales agent

    Sorts replies, books meetings, sends follow-ups and hands hot contacts to salespeople in the CRM.

  5. The agent learns

    It measures what works for each group of companies and carries that into the next messages. Each one is more precise.

Timeline

September 2026today

  1. Manual research
  2. Research: Slovakia
  3. Research: Sweden
  4. Sales agent
  5. Follow-ups

“The system learns. Fewer emails, more replies and calls.”

  • Companies vettedAt the start · first 2 weeks: 509After tuning · 2 weeks: 509
  • First emailsAt the start · first 2 weeks: 173After tuning · 2 weeks: 140
  • RepliesAt the start · first 2 weeks: 9After tuning · 2 weeks: 15
  • Calls bookedAt the start · first 2 weeks: 1After tuning · 2 weeks: 3

What we tuned

  • Strong signalsthe agent better spots a company that has the problem right now
  • Personal messagesevery email builds on a specific signal from the company
  • Reviewopens and replies shape the next week

5% → 11%reply rate

Under the hood

For those who want to know how it’s built.

Scheduling
Claude scheduled tasks, results tracked by company group
CRM
Pipedrive MCP, hot contacts go straight to salespeople
Email
Gmail MCP with a brake on bounced emails
Calendar
Calendar MCP and Cal.com
Sources
Google Maps, FinStat, the RPO register, allabolag.se, Leadfeeder, job boards, social media
Browser
Claude in Chrome
Code
no custom server, only Claude’s own tools

“In the morning I check who I should talk to. The agent does the searching and the writing.”

“Want something similar?”

Let’s start with your problem.

The itashu assistant, a robot with a red head, waving

“Have a question? Just ask.”

Want to know more about our services? Interested in concrete use cases and references, whether we’ve already solved a problem like yours, or what it might all cost? Talk it through with our AI agent. It knows a lot more than you’d think.

Cookie settings

Choose which cookies we may use. You can change your choice at any time in the website footer.

Speaking

itashu assistant

Hello, I’m the itashu digital assistant. How can I help you?