The challenge
The client had released their four-person SDR team in Q4, partly for cost reasons, partly for performance. They needed to keep pipeline coming in but could not afford to rebuild the team. Pipeline targets were $1.8M qualified for the quarter.
The approach
We built a signal-driven outbound stack, not a spray-and-pray sequence engine. We started from the ICP and the buying signals (funding round, headcount change, specific tech adoption) and worked backward into prospecting and messaging.
What we built
We connected Apollo and Clay to the prospect pipeline, watching three buying signals weekly. A first-line generation agent running Claude Sonnet 4.6 writes genuinely personalized openers, not the “I noticed your company recently” pattern every other outbound tool produces. Smartlead handles the sending.
The most valuable component is a reply triage agent that reads every response and classifies it into book-meeting, not-now, not-interested, or hostile. Book-meeting cases receive a calendar link inserted into the reply automatically. Not-now cases are filed for re-engagement at 90 days. A human reviews the hostile queue before any action is taken.
Total monthly cost: $200 in tools, plus sender domain warm-up and email sending volume costs. Build cost: $11K one-time.
Results
- 47 qualified meetings booked in month 1, vs the previous SDR team’s 38/month average
- $24K/month in salary saved, redeployed into product and marketing
- +38% pipeline vs prior quarter, exceeded target
- Reply rate 4.8%, about 2× industry baseline
- Founder spends 30 min/day reviewing the queue, vs 8 hours/week with the previous team
Tech used
Apollo · Clay · Smartlead · Claude Sonnet 4.6 · GPT-5 · custom Python orchestration · HubSpot CRM integration
Takeaway
The agent stack is not strictly better than a great human SDR team. It is substantially better than a mediocre one, and at this price point, it can be wrong half the time and still win on unit economics. The unlock was the reply triage agent. That is what made it a system instead of a sending bot.