logo
Case study

The AI team member that runs our own client acquisition

Before we install agents inside anyone else's firm, we run them inside our own. This is the live system - not a demo. Everything below is observable on a screen-share, and we're happy to show it.

Live & demoableClient: DimsouWorkflow: outreach ops
agent - activity log (excerpt)
09:41:07researched 15 candidate firms · 5 verified, 10 discarded as unverifiable
09:44:32crm check - 1 lead already in an active sequence → excluded from campaign
10:02:18drafted 5 firm-specific emails → queued for human approval
10:15:55escalated 1 ambiguous contact history → flagged for review
81Firms researched
4Campaigns launched
59Emails sent
25+Hours saved weekly
The problem

Client acquisition is classic operational drag.

Finding and contacting the right firms means researching companies, verifying who the decision-maker is, finding a real email address, checking whether we'd already contacted them, drafting outreach specific to each firm's situation, and running disciplined follow-up without dropping threads.

Done properly, it's hours of repetitive work every week. Done improperly, it's embarrassing - duplicate emails, wrong names, forgotten follow-ups.

That's exactly the category of work firms usually solve by hiring: an intake coordinator, a researcher, an admin. We solved it by deploying an AI agent instead.

What the drag looks like

  • Researching firms & verifying decision-makers
  • Hunting real, deliverable email addresses
  • Checking prior contact before every send
  • Drafting firm-specific outreach, not blasts
  • Follow-up cadence that never drops a thread
What the agent does

One workflow, four checkpoints, zero guessing.

The agent runs the full pipeline end-to-end - but every step has a rule about what it may do, and what it must hand to a human.

01 / Research

Research & qualification

Given a target profile - industry, size, location, signals like active hiring - it sources candidate firms, identifies the named decision-maker, and verifies a real email address.

rulecan't verify it? Discard the lead - never guess.
02 / Check

CRM cross-referencing

Before any firm enters a campaign, the agent checks our outreach system for prior contact - existing conversations, active sequences, past replies - and routes accordingly.

ruleno duplicate threads. No cold email mid-conversation.
03 / Draft

Drafts, human approves

For each qualified lead it drafts outreach tailored to that firm's specific operational pain. The agent drafts and queues; a human reviews every batch before anything sends.

rulenothing sends without sign-off.
04 / Run

Follow-up on autopilot

Approved sequences run automatically on a fixed cadence - and stop the moment someone replies. No thread is dropped, and no one gets chased after answering.

rulea reply halts the sequence, instantly.
[!]

Escalates when unsure. Ambiguous cases - a bounced address, a lead with messy contact history, anything it isn't certain about - get flagged to a human rather than guessed at.

The results

In its first week of operation:

8 days live - 30 June to 7 July 2026.

26 firms

Qualified & verified

From 81 researched across 4 market segments - each with a named decision-maker contact. Unverifiable leads were discarded, not guessed.

4 campaigns

Drafted & launched

Firm-specific messaging tied to each firm's actual operational pain - no generic blasts.

59 emails

Sent & managed

Every message human-approved before sending, with automated follow-up that stops the moment someone replies.

Reliability record
  • 0missed follow-ups - every thread runs to reply or completion
  • 0duplicate sends - every lead cross-referenced before contact
  • 100%of emails human-approved before sending
Cost comparison
Part-time researcher / SDR$2,000-3,000+ / mo
This agent, running the same rolea fraction of that
Vacation days, forgotten follow-upszero

The 25+ hours are an estimate for that first week - roughly 81 firms researched and verified at about 15 minutes each, plus campaign drafting and follow-up management, conservatively rounded. The equivalent human role - a part-time researcher or SDR - runs from about $2,000-3,000+ a month (a full-time SDR in the US averages ~$4,600). The agent runs for a fraction of that, doesn't take vacation, and never forgets a follow-up.

The boundaries

The part that matters most.

The agent only does what it's allowed to do. This is how we build every agent we deploy: scoped permissions, human checkpoints where they matter, and full visibility into what it's doing and why. Your client data would be treated the same way - under your control, inside boundaries you set.

CANNOT

Send email without approval. Every message a human signs off on, batch by batch.

CANNOT

Contact anyone mid-conversation. Active threads are off-limits to campaigns.

CANNOT

Invent contact information. Unverifiable leads get discarded, not guessed.

MUST

Ask when uncertain. Ambiguity gets escalated to a human, every time.

What this means for your firm

The same pattern applies to your operational drag.

If your team spends hours on document collection, client intake, status updates, or any repetitive coordination work - an agent that works like an extra team member, scoped to one workflow, with your people only stepping in when judgment is actually needed.

Currently running a small number of pilots with professional services firms.

  • [✓]Chasing clients for documents week after week
  • [✓]Partners pulled into routine intake and coordination
  • [✓]Status updates assembled by hand for every client
  • [✓]Follow-ups that live in someone's head - until they don't
See it for yourself

Watch the live system on a screen-share.

Thirty minutes. We'll walk you through the agent running our own outreach - the real system, real data, real checkpoints - and map which of your workflows the same pattern fits.

No deck. No demo environment. The live system, on screen.

Case Study: The AI team member that runs our own client acquisition | Dimsou