What Is an AI SDR and How Does It Work?

A human SDR’s week is a mix of list building, research, writing first emails, following up, answering routine replies, making calls, asking a few qualifying questions and booking meetings for account executives. Most of that is repeatable and specifiable, which is why it is the part of sales that AI agents took on first. An AI SDR is one or more agents given that job, measured the way an SDR is measured: qualified meetings held, positive reply rate, show rate, opt-out and complaint rate.

What it is not: a closer, a negotiator or a relationship. An AI SDR opens conversations and qualifies them. The moment a conversation turns to price, terms or a serious objection, it should hand off — which is exactly what a good human SDR does too.

The phrase “AI SDR” usually covers three distinct roles, and it is worth separating them because they need different levels of trust. A prospecting agent finds accounts and contacts that match the ideal customer profile, researches them, enriches and verifies details, de-duplicates against the CRM and suppression lists, and scores fit, intent and urgency. It never contacts anyone, so it can run fully unattended. An outreach agent personalises and sends an approved sequence from a real mailbox, reads replies, records opt-outs and answers routine questions; because it reaches customers, enrolment is the step that needs approval. An appointment setter — often the same agent as the outreach role — asks qualification questions by email or on the phone, offers a slot and books it, then writes a brief for whoever attends.

Splitting the roles is what makes the whole thing safe. The finder can be wrong about a contact and nothing happens; the contactor can only send what a person approved; the setter can only book what the calendar allows.

The hand-off is where AI SDR projects succeed or fail. A booked meeting with no context is barely better than a cold call for the AE. The agent should deliver three things: a qualification record with the actual answers, not a score; a brief that reads like a colleague wrote it — who, what they said, what they want, what to avoid; and a clean CRM record with every touch logged so the AE can see the history. After the meeting, the AE’s notes and next step should flow back so the agent can run the agreed follow-up.

The other hand-off is the escalation. When a prospect asks about price, wants a discount, raises a contractual point or complains, the agent should say a colleague will follow up and flag the conversation for a person immediately. That flag is the AE’s cue, and it should arrive with the whole thread attached.

In Eligoo the AI SDR is two employees. Radar is the prospecting agent: Apollo people search, web research, enrichment, verification, de-duplication, fit/intent/urgency scoring and segmentation, with suppression lists honoured — and it never contacts prospects. Hook is the outreach agent and appointment setter: it personalises and sends approved sequences from your own SMTP/IMAP mailbox, polls for replies with de-duplication, records opt-outs, makes and takes AI phone calls on your own SIP trunk, asks your qualification questions and books meetings as Google Calendar events with a Meet link, plus a meeting brief. Ledger keeps the CRM clean and reports funnel metrics and attribution; Atlas sees it all in the weekly plan.

Hook’s KPI is qualified meetings held. Sequence enrolment and starting a calling campaign are approvals; failed executions return to pending; every decision is logged. Actions that reach the outside world — sending an outreach sequence, launching or changing ad spend, publishing outside an approved calendar, starting a calling campaign — wait in an approvals queue until a person approves them. Pricing, discounts, promises and complaints are outside Hook’s boundary and are flagged to a person.