How to use AI sales agents to book more B2B meetings
A step-by-step guide to deploying AI sales agents that actually fill the calendar — without burning your domain or your reputation.
The promise of AI sales agents is irresistible: a tireless SDR that runs thousands of personalised touches a week, books meetings while you sleep, and costs less than one junior hire. The reality, for most teams, is a quiet inbox and a damaged sender reputation.
The difference between AI outreach that works and AI outreach that gets filtered is rarely the model. It's almost always the operating model around it. This guide walks through the exact setup we use at Growth Broker to make AI agents reliable meeting machines.
Step one is domain hygiene. Never run AI outreach from your primary domain. Buy two or three look-alike sending domains, warm them for 21 days, and rotate them. Configure SPF, DKIM, and DMARC properly. Treat deliverability as a product, not an afterthought.
Step two is ICP narrowing. The agent is only as good as the list it works from. Define the segment in one sentence: 'Series B fintechs in the UK that just hired a Head of Compliance.' That's a list of maybe 80 companies. Perfect.
Step three is signal enrichment. Plug the agent into hiring data, funding announcements, product launches, and tech-stack changes. A cold email that opens with a real, recent signal converts 3–5x better than a generic 'I noticed your company...' opener.
Step four is the message itself. Short, specific, and ending in a low-friction ask. 'Open to a 15-minute call next Tuesday?' beats 'I'd love to learn more about your business' every single time. Let the AI generate ten variants and A/B test ruthlessly.
Step five is the handoff. The single biggest leak in AI outreach is the gap between 'interested reply' and 'meeting on calendar.' Build a workflow where a human responds within 20 minutes during business hours. That one rule doubles booked-meeting rates.
Step six is measurement. Track three numbers weekly: reply rate per segment, meeting rate per reply, and qualified-pipeline rate per meeting. Anything else is noise. If pipeline isn't moving, kill the segment, don't tweak the subject line.
Done well, this setup produces 8–15 qualified meetings per agent per month at a cost that makes a traditional SDR look expensive. Done badly, it produces spam complaints and a blacklisted domain. The difference is discipline, not technology.
Frequently asked questions
AI Outreach — answered
- How many emails per day should an AI agent send?
- Per mailbox: 30–50 once warmed. Scale by adding mailboxes, not by raising per-mailbox volume — that's what triggers spam filters.
- Should I disclose that the email was AI-generated?
- No legal requirement in most markets, but tone matters more than disclosure. If a recipient can't tell, they don't care; if they can, you've already lost.
- What reply rate is realistic for AI outbound?
- 3–8% positive reply rate on a well-targeted list with signal-driven openers. Below 2% means your list or message is wrong, not your tool.
- Can AI agents handle LinkedIn outreach safely?
- Yes, with sensible limits (15–25 connection requests/day, human-paced sends) and a dedicated account. Avoid third-party automation that violates platform terms.
Growth Broker editorial
Filed under ai outreach