Learn how to use Apollo.io AI Research to prioritize accounts at scale and personalize outreach without stitching together Clay, Zapier, or extra tools. We build a targeted list with lookalikes, craft a web-enabled prompt, create a Product Signal Score custom field, and turn that into a clean sentence you can drop into Apollo sequences. Then we automate it with weekly workflows that research companies, score fit, and queue verified contacts so your team spends less time digging and more time starting conversations. If you want help implementing this in your Apollo instance, reach out and I can set this up for you.
00:00 Intro and goals of the tutorial
00:12 Apollo.io AI Research inside Apollo (no Clay or Zapier)
00:19 Research goals: pick accounts, segment, lifecycle, personalize email
00:45 Client example and target list (user testing company)
01:02 Lookalikes for connected fitness (Whoop, Fitbit, Zwift)
01:20 Filters vs AI for hard-to-find signals
01:31 SDR manual research vs AI at scale in Apollo
01:58 Overview of the Apollo.io AI Research workflow
02:06 Stage 1 – build and run a custom AI prompt (use ChatGPT to design it)
03:02 Model choice in Apollo (Perplexity Sonar with web access)
03:16 Long-prompt details: scan site, collect signals, score 1–5
03:44 What to count as good vs bad signals (beta programs, roadmap, changelog; avoid Glassdoor)
04:09 Create the “Product Signal Score” custom field
04:34 Test on 10 rows and review outputs
05:07 Iterating and refining the prompt for better signals
05:29 Stage 2 – generate the explanation field
05:47 Stage 3 – generate a product-signal sentence for emails (style rules)
06:15 Example: Zwift scored 5 and referenced in the email
07:01 Using variables in the Apollo sequence
07:24 Contact example (Claire McGowan at Zwift) and FutureWorks reference
07:51 Tips before scaling: trial and error, tweak first
08:03 Automation overview
08:17 Workflow 1 – company-level weekly research run
09:09 Weekly cadence, 500 companies, credit usage
09:28 Workflow 2 – person-level sequencing after scoring
10:20 Verify emails and drop into the personalized sequence
10:41 Throughput pacing (about 300 per week)
10:49 Measure what works and iterate
11:10 Wrap-up and CTA to reach out for help
