Your SDR Team Is Too Big. Here's the Math.

Let me put the number on the table before I explain it: a fully-loaded SDR at a mid-market SaaS company costs between $150,000 and $200,000 per year when you account for everything.

Most leaders quote salary when they think about SDR cost. Salary is the smallest part. This is the first mistake.

Fully-loaded SDR cost breakdown: Base salary ($60–80K) + variable comp ($20–30K) + benefits (25–30% of base) + tools (Salesforce seat, Outreach, ZoomInfo, LinkedIn Sales Nav - $8–12K/year per rep) + management overhead (one SDR manager for every 6–8 reps, fully-loaded at $160K+, allocated per rep) + ramp cost (an SDR at 50% capacity for 3–4 months before they're fully productive). Add it up honestly and you're at $150–200K per fully-loaded SDR per year.

Now here's the cost of an AI agent doing the research, enrichment, first-draft, and scheduling coordination that SDRs spend 40–50% of their time on: roughly $500–2,000 per month depending on volume and tooling. Call it $15–25K per year at meaningful scale.

That's not a small gap. That's an order-of-magnitude gap for the mechanical portions of the role. I once gave a board presentation where I confidently committed to $5M in Q4 pipeline. We closed $1.8M. The board did not find my optimism charming. The SDR math is the thing I wish I'd had cold before that conversation.

What Agents Actually Do vs. What SDRs Should Be Doing

Before anyone accuses me of recommending you fire your SDR team on Monday morning, let me be specific about what the agent replaces and what it doesn't.

What an AI agent does well:

Account research and signal monitoring. Every morning, before your SDRs arrive, an agent can pull news, job postings, funding announcements, and competitive signals for every company on your target list. A skilled SDR doing this manually produces 12–15 quality account briefs per week. The agent produces 50 before 6 AM. At Prokeep, where I was doing pipeline expansion work across 3 new verticals simultaneously, this kind of coverage would have required 4–5 additional research-focused SDRs. The agent does it for $40/month.

To be fair: the first version of my AI agent once recommended we prospect a company that had gone bankrupt six months earlier. It cited the bankruptcy announcement as a "relevant business event." Technically accurate. Strategically disastrous. The technology has since improved.

Contact enrichment. Pulling accurate contact data, verifying emails, identifying the right title and reporting structure for a given ICP. This is almost entirely mechanical work. An agent does it faster, more consistently, and without the fatigue errors that show up when a human has been doing data enrichment for three hours.

Sequence management and first-draft personalization. Taking account research and generating a personalized first-touch email is now well within AI capability - not perfectly, but at a quality level that outperforms the template-based outreach most SDRs were doing anyway. One SDR managing 400–500 AI-generated and AI-personalized sequences versus one SDR manually writing 15 per week is a legitimate comparison.

Meeting coordination. Scheduling, follow-up reminders, calendar management. Fully automatable.

What an SDR should actually be doing:

Phone calls where reading tone and adjusting in real time matters. Warm outreach where the human element of relationship building is part of the value proposition. Complex multi-threaded outreach where navigating organizational dynamics requires judgment. Creative problem-solving when the standard approach isn't working. The follow-up call after a no-show where the right words at the right moment can save a meeting.

These are real skills. The best SDRs were always doing more of these and less of the mechanical work. AI gives them the infrastructure to focus there exclusively.

The Hybrid Model That Actually Works

I'm not recommending you eliminate SDRs. I'm recommending a different ratio.

Old model: 1 SDR manually researching, building lists, writing sequences, and making calls across 150–200 accounts per quarter.

New model: 1 SDR managing 400–500 AI-generated, AI-personalized sequences - reviewing, editing, approving, and handling all human responses and calls. The SDR's job becomes quality control, judgment calls, and live conversations. The agent handles the mechanical infrastructure underneath.

In the old model, you need 3 SDRs to cover what 1 SDR can now manage with AI infrastructure. That's the 3x ratio I opened with. Most companies are still running the old model, which means they're paying $450–600K for outcomes a leaner, better-equipped team could deliver for $200K. The gap is the laundry on the Peloton.

The SpringCM Calculation

At SpringCM, I founded the BD organization from scratch in 2018. In the first six months we generated $1.5M in new revenue. I was doing this with a lean team because we didn't have the budget for a large SDR function - we had to be surgical about who we hired and how they spent their time.

In retrospect, I was running a version of the hybrid model before I had the tools to run it well. I was forcing reps to be selective about account research because they physically couldn't cover everything. We prioritized by gut and by limited data.

If I had the AI infrastructure I'm describing now, I would have covered the same territory with 2–3 fewer SDRs, freed up significant budget for higher-caliber AE talent, and probably hit the $1.5M number in four months instead of six. The constraint wasn't effort. It was coverage and research capacity. Both are now solvable with $40/month and a weekend of setup (and the associated insomnia).

What This Means for SDR Career Paths

This is not a story about SDRs being replaced. It's a story about the SDR role evolving toward higher-value activities - which, honestly, is what good SDRs always wanted anyway. Nobody with talent goes into SDR work to spend half their day doing data enrichment.

The SDRs who adapt toward the conversation and judgment-intensive side of the role will have a stronger career trajectory in this environment, not a weaker one. Their skills become more valuable as the mechanical work disappears from the role definition.

The SDRs who are primarily valuable for their volume - high dial counts, high email counts, low genuine engagement per contact - are in a more difficult position. That specific value proposition doesn't hold up against an agent that can do the volume work faster and cheaper.

If you manage an SDR team, the most useful thing you can do is be honest with your people about this and help them develop the skills that differentiate humans from agents. That's a leadership responsibility. It's also a harder conversation than most managers are having right now - but significantly less hard than explaining to a board why your SDR cost per pipeline dollar hasn't moved in three years while the tools to fix it have been available since 2024.

The math doesn't lie. The question is what you do with it.

See the full architecture at joepeck.ai/projects/autonomous-sdr.