← InsightsCase study
Case study

The human–AI agent partnership that booked 48 demos in 3 months

The starting point

Our client's target customers are local businesses. They had early signs of fit across several markets:

Private clubsIndependent restaurants & barsLuxury salons & spasSpecialty cafés

Every one of those customers had found them through word of mouth.

The client lacked a proactive outreach system, which made their growth depend on referrals. This hindered the speed required to quickly test out different markets and scale with confidence.

The goal

They wanted to validate those markets and expand into similar ones, proactively. We designed an email program to book meetings every week, then used the live feedback from each week's conversations to sharpen the next week's targeting.

How we solved it

We took a human+agent approach. Outreach, client relations and reading market feedback were still driven by human judgement whereas, finding the right prospects, verifying the owners, sourcing accurate contact, scheduling emails was all agentic.

Human

Judgement and relationships
  • Outreach decisions
  • Client relations
  • Reading market feedback

Agent

The legwork, on autopilot
  • Finding the right prospects
  • Verifying each owner
  • Sourcing accurate contact details
  • Scheduling the emails
The weekly loop
AgentAutomated legwork
Find prospects
Verify owners
Source contacts
Schedule emails
HumanJudgement & relationships
Refine targeting
Read market feedback
Client relations
Meetings & replies
This project has shown us the true power of AI: used judiciously, it can enhance our businesses multi-fold.The Kunsoth AI team

Inside the agent

How we built it step by step, the regression tests that keep it from breaking, how we brought costs down strategically, and how it learns from every week of feedback.

Coming soon

Results vs industry benchmarks

3 months
1,699Unique businesses reached
10.4%
Reply rate
27%
Positive replies
48Meetings booked

Over the course of 3 months, we reached 1,699 unique businesses in 8 markets, averaging about 4 meetings per week.

The sheer scale of this level of iteration was possible due to the time and effort saved by delegating playbook-driven work to AI.

To see for ourselves how much this agentic AI system helped us, we first did the same work manually: to learn how to do it, and to test whether AI could do it at least as well, how long it would take and what it would cost. Here are the results:

Manual vs Agent

Manual
2,791 min
Time to source 500 contacts
Manual
$232.56
Cost for 500 contacts
Manual
10.8 / hr
Contacts sourced per hour

Want a pipeline like this?

Book a demo →