How to Set Up an AI SDR in 5 Steps

Start with 2 documents: a one-page product cheat sheet and a 1,000 to 1,500 word ideal customer profile (ICP) brief. The cheat sheet lists problem statements, 5 value propositions, 3 case study bullets with numbers, and top objections. The ICP brief covers company size, titles, buying triggers, and 5 common signals that indicate fit. Keep both documents editable and dated.

Create a persona matrix with 3 personas per ICP and 3 prioritized pain points per persona. For each pain point add exact proof points and a recommended CTA. This gives the AI SDR clear context to choose the right message, not guess. Save all sources - product docs, case studies, battlecards - as indexed files the model can reference at runtime.

Quality of leads beats quantity. Define fit rules that yield 200 to 1,000 true prospects per campaign. Example rules: company size 50-500, tech stack includes Segment, revenue between 10M and 200M, and titles include Head of Revenue or VP Growth. Use boolean queries and firmographic enrichment to automate the filter.

Set contact acceptance rules: 95% direct email confidence, role match within top 2 titles, and recent activity signal within 90 days. Export a small seed list of 200 prospects for the first run. This lets you validate messaging and deliverability before scaling to several thousand.

Design 3 message paths: cold intro, follow-up, and break-up. For cold intro aim for 50 to 90 words. Include a one-line value prop, one personalized hook based on the persona matrix, and a simple CTA like Can we test a 15 minute call next Tuesday or Wednesday? Build 3 variants per message to A/B test subject lines and opening hooks.

Set a 5-touch sequence spread over 21 days: Day 0 email, Day 2 short LinkedIn connection or note, Day 5 email follow-up, Day 12 value-add email with a relevant stat or doc, Day 21 break-up. Limit first-touch sends to 25 emails per new domain per day while you validate. Log which template the AI picks and why so you can iterate fast.

Warm up sending domains for at least 14 days before any high-volume sends. Start at 10 emails per day, increase by 20 to 30 percent daily, and reach a steady-state of 50 to 100 sends per day per domain before scaling. Use dedicated subdomains per campaign segment and set up SPF, DKIM, and DMARC records with company IT.

Monitor these KPIs daily during warm-up: bounce rate under 2 percent, spam complaints under 0.02 percent, and open rate between 15 and 35 percent depending on industry. Pause and remediate if bounce or complaint thresholds are exceeded. Deliverability is a gating factor; even great AI messaging fails if domains are blacklisted.

Launch to the 200-prospect seed list. Track reply rate, positive reply rate, meetings booked, and lead quality. Use concrete thresholds: aim for a 6 to 10 percent reply rate, 1 to 2 percent meeting rate, and less than 20 percent low-quality leads in first full cycle. If reply rate is below 4 percent, re-evaluate personalization hooks and list fit.

Use an experimentation cadence: change only one variable per week - subject line, CTA, or persona filter - then measure. Implement feedback loops: route human-verified replies into the knowledge base, note new objections, and retrain the AI prompts weekly. Only scale volume after two successful cycles that meet your metrics. Where helpful, platforms like Outpace can centralize knowledge indexing and automate sequence rules, but keep control of final decisioning.