How to Use AI for Sales Prospecting: A Step-by-Step Workflow
Part 2 of 6. A practical, repeatable workflow for using AI to source, score, research, and personalize prospecting without becoming a spam machine.
Part 2 of the AI Sales Outreach Series.
Most teams bolt AI onto a broken process and get faster bad outreach. This is the workflow that works: ingest, score, research, draft, decide, learn.
Step 1 — Ingest the right data
- Your LinkedIn network, connections, and profile views
- CRM history and past conversations
- Inbox replies and meeting history
- Public company signals: hiring, funding, launches, leadership changes
Step 2 — Score fit and timing separately
Fit answers "should we ever sell to them?" Timing answers "is something happening right now?" Combining them into one number hides the reason you should act. Keep them side by side.
Step 3 — Let AI do the research pass
For each high-scoring relationship, generate a short brief: what the company does, what changed recently, the likely pain, and the warmest path in.
Step 4 — Draft, don't send
Generate two or three variants per opportunity — an initial message and a follow-up. The human picks one and edits a line. This one rule preserves quality at scale.
Step 5 — Work a ranked queue, not a list
Your daily view should be five to fifteen named relationships with a reason attached, not 300 rows sorted alphabetically.
Step 6 — Feed outcomes back
Log replies, positives, and ignores. Scoring that never learns is just a static filter.
A realistic daily routine
- Open the ranked queue (2 min)
- Review reasons and drafts (10 min)
- Edit and send 8–12 messages (20 min)
- Handle replies (rest of the day)
Common mistakes
- Automating sends before quality is proven
- Scoring on job titles alone
- Ignoring second-degree warm paths
- Treating every trigger as urgent
Key takeaway
AI prospecting is not about sending more. It is about starting each day knowing exactly who to contact and why.
Next in the series: which AI sales outreach tools actually matter.