How to Build a $10k/mo Outbound Pipeline with AI (2026)

Step-by-Step Workflow

7 steps · 6 hours to set up · 5 hrs/week ongoing

1. 1

   Define one ICP and one offer

   Pick a single buyer persona and a single offer. The biggest mistake first-time outbounders make is targeting three personas with three offers; reply rates drop below 1% because nothing resonates. Examples that work: 'series A SaaS founders with 10-50 employees, offer free 30-min teardown of their pricing page.'

   30 minOutput: One ICP doc, one offer one-liner

   Tip: Write your offer as 'I help [persona] [outcome] without [common pain]'.

2. 2

   Pull 1,000 prospects from Apollo

   Use Apollo.io with strict filters: company size, industry, tech stack, hiring signals. Export to CSV. At reply rates of 3-5%, 1,000 prospects per month generates ~30-50 booked calls.

   45 minOutput: CSV of 1,000 prospects with verified emailsTools: Apollo.io

   Tip: Filter for companies that posted job openings in your category within last 30 days - they're actively buying.

3. 3

   Enrich and score in Clay

   Import the CSV into Clay. Add enrichments: recent funding, headcount trend, tech stack changes, LinkedIn activity. Use Claude (via Clay's AI column) to score 1-10 on 'is this prospect a hot fit'. Filter to top 300.

   1.5 hrsOutput: 300 high-fit, scored prospectsTools: Clay, Claude

   Tip: The Clay 'AI Research Agent' column reading their LinkedIn last 30 days posts is your highest-signal data point.

4. 4

   Generate first-line personalization with Claude

   Use Claude with a prompt that pulls each prospect's LinkedIn headline + most recent post + company funding date. Output: one personalized first sentence per prospect. Length cap: 18 words. Don't let Claude write the whole email - first line only.

   30 min for setup, automated thereafterOutput: 300 personalized first lines, one per prospectTools: Claude

   Tip: Prompt: 'In one sentence under 18 words, reference [LinkedIn post snippet] without praising it. No greeting. Plain text.'

5. 5

   Set up Smartlead with warmed inboxes

   Buy 3-5 domain variations (e.g. tryyourbrand.com, getyourbrand.com). Set up 2-3 inboxes per domain. Warm for 14 days via Smartlead's warming pool before sending real outbound. This is non-negotiable; cold inboxes get spam-foldered.

   1 hr setup, 14 day warmupOutput: 10+ warmed inboxes sending 30 emails/day eachTools: Smartlead
6. 6

   Send 4-step sequence over 12 days

   Sequence: Day 1 personalized first line + 2-sentence offer, Day 4 short bump, Day 8 case study or proof, Day 12 'closing the loop'. Total emails: 4 per prospect. Send 300 prospects across 10 inboxes = 30 prospects per inbox per cycle.

   Automated; 30 min/week monitoringOutput: ~15-25 replies, ~5-10 booked calls per 300-prospect cycleTools: Smartlead

   Tip: Reply rates above 5% mean your first line is doing the work; under 2% means the offer is wrong, not the copy.

7. 7

   Iterate weekly using reply data

   Each week, pull replies into a spreadsheet. Tag each as positive/neutral/negative. Look for: which first-line variants got the most positive replies, which ICP segments converted, which days had highest open rates. Adjust prospect filters in Apollo for next cycle.

   1 hr/weekOutput: Weekly iteration log, improving reply rate 0.5-1% per cycle