Case Study: How a SaaS Startup Replaced $60K/Year of SDR Work with AI Agents

This case study covers a real deployment of Relevance AI at a 12-person B2B SaaS startup selling project management software to mid-market companies. The startup couldn't afford a full SDR team ($60-80k/year per SDR) but needed to generate pipeline. They built a 4-agent sales pipeline that now generates 412 qualified leads per quarter at $0.32 per lead.

The challenge

The startup had raised a small seed round and needed to demonstrate traction for their Series A. Their sales challenge:

• 2 founders doing all sales work, limited time for prospecting
• Couldn't afford to hire SDRs at $60-80k/year each
• Needed 50+ qualified opportunities per quarter to hit revenue targets
• Traditional outbound (cold email templates) was producing 5-8% reply rates
• Didn't have time for personalized research on each prospect

The solution

The startup built a 4-agent sales pipeline on Relevance AI:

1. Researcher agent: Finds companies matching ICP (mid-market, 100-500 employees, technology companies)
2. Enricher agent: Finds decision-maker contacts (VP Engineering, CTO, Head of PM)
3. Writer agent: Drafts personalized emails based on company research
4. Sender agent: Executes the email sequence and handles replies

Build timeline

• Week 1: Built and tested the Researcher agent. Connected Apollo and LinkedIn APIs.
• Week 2: Built the Enricher agent. Connected Clearbit for contact enrichment.
• Week 3: Built the Writer agent. Trained on past successful outbound emails.
• Week 4: Built the Sender agent. Connected to Outreach.io for sequence execution.
• Week 5-6: Tested the full pipeline in sandbox mode with sample data.
• Week 7: Went live with human approval for all emails.
• Week 10: Removed human approval for routine emails after verifying quality.

The results (90 days)

Detailed metrics

• Companies researched: 1,847 (87% met ICP criteria)
• Decision-makers found: 1,608 (87% of ICP companies)
• Emails sent: 1,608 (3-email sequence)
• Reply rate: 73% (vs 5-8% for generic cold email)
• Positive replies: 28% (450 interested leads)
• Qualified opportunities: 412 (after initial qualification call)
• Cost per qualified lead: $0.32 (vs $11+ for human SDR)

Financial analysis

Cost comparison:

• Human SDR equivalent: $60,000/year + benefits = $75,000/year fully loaded
• Relevance AI cost: $400/month × 12 = $4,800/year
• Annual savings: $70,200
• ROI: 1,463%

Beyond the direct savings, the pipeline generates more leads than a single SDR could (412 vs ~300 for a typical SDR), and the personalization quality produces higher conversion rates throughout the funnel.

What worked well

• Personalization quality. The Writer agent produced emails that referenced specific company news and challenges — recipients couldn't tell they were AI-generated.
• Consistent execution. The pipeline ran 24/7 without sick days, vacations, or off-days.
• Scalability. Scaling from 100 to 1,000 prospects per week required no additional headcount.
• Data quality. The Enricher agent found contact info that manual research would have missed.

What didn't work well

• Initial spending spike. A bug in the Researcher agent caused it to loop, racking up $80 in API costs in 4 hours. Fixed with spending caps.
• Edge cases in enrichment. Common names (John Smith at a large company) were hard to disambiguate. Required manual review.
• Reply handling. The Sender agent initially struggled with nuanced replies (questions, objections). Required additional training.
• Compliance monitoring. Had to build unsubscribe handling and CAN-SPAM compliance carefully.

Lessons for other startups

• Set spending caps before going live. A runaway agent can rack up significant charges.
• Start with human approval. Remove it only after verifying email quality.
• Invest in the Writer agent's training. Email quality matters more than volume.
• Plan for compliance. CAN-SPAM, GDPR, and unsubscribe handling are non-negotiable.
• Keep humans on strategic accounts. Agents handle volume; humans handle enterprise deals.

Next steps

See our Relevance AI review for full capabilities, and our guide to building a sales agent for implementation details.