AI Lead Qualifier — Score Leads & Write Openers

Turn a raw list of leads into a scored, prioritized outreach list — with a personalized opening line already written for each one. Point it at any batch of leads, describe your ideal customer and your offer, and get back every lead scored 1–10 with a one-sentence reason and a ready-to-send cold-open line.

Not a scraper — this is the qualification layer that goes after scraping. Feed it the output of any lead-scraper on Apify (including the Wild Apricot, MembershipWorks, and ChamberMaster directory scrapers) and it turns hundreds of raw rows into a ranked call list.

Bring your own key = you pay pennies here

The model runs on your OpenAI-compatible API key (OpenAI by default; works with Azure, OpenRouter, Together, local servers — anything OpenAI-compatible). Because the model cost is billed to your own account, this Actor charges only a few cents per lead for the orchestration, prompting, parsing, and delivery.

Input

{
  "leads": [
    { "name": "Bluewater Bistro", "category": "Restaurant", "city": "Seattle" }
  ],
  "inputDatasetId": "PASTE_A_SCRAPER_RUN_DATASET_ID_optional",
  "icpDescription": "Independent restaurants and bars, not chains.",
  "offerDescription": "Modern restaurant POS with free install.",
  "scoreThreshold": 7,
  "onlyReturnQualified": false,
  "model": "gpt-4o-mini",
  "apiKey": "sk-your-own-key"
}

• leads: array of objects (any shape) or strings. And/or…
• inputDatasetId: qualify another Actor's output directly — paste its dataset ID.
• icpDescription / offerDescription: the sharper and more specific, the better the scores.
• onlyReturnQualified: return just the leads that meet your threshold.
• model / apiBaseUrl / apiKey: your provider and key (key stored as a secret).

Output

Each input lead is returned unchanged, plus a _qualification block:

{
  "name": "Bluewater Bistro",
  "category": "Restaurant",
  "city": "Seattle",
  "_qualification": {
    "score": 9,
    "reason": "Independent food-service business, strong ICP fit.",
    "opener": "Saw your spot listed locally — most independents here still run legacy POS; ours installs free before your next weekend rush.",
    "qualified": true,
    "model": "gpt-4o-mini",
    "at": "2026-07-23T21:40:00.000Z"
  }
}

Typical workflow (the flywheel)

1. Run a directory/lead scraper → dataset of raw businesses.
2. Run AI Lead Qualifier with that inputDatasetId + your ICP + offer.
3. Export the qualified rows (with openers) to your CRM / Google Sheet / Slack.

Notes

• Robust to messy model output — a malformed response never kills the run; it fails safe per lead.
• Concurrency is configurable (1–10) for speed vs. rate limits.
• Your API key is stored as a secret and used only to call your chosen model.
• No personal data is sent anywhere except your own model provider.

---

🔗 Part of a lead-generation suite (scrape → enrich → qualify)

This Actor works best chained with its companions on the same account:

1. Directory scrapers — get businesses/members (name + website) from Wild Apricot, MembershipWorks, or ChamberMaster / GrowthZone directories.
2. Business Website Contact Extractor — add the emails/phones the directories hide.
3. AI Lead Qualifier — score each lead against your ICP and write a personalized opener.

Pass a run's dataset ID into the next step. NOTE: cross-actor chaining needs your apifyToken in the input (Apify runs actors with limited permissions, so reading another run's dataset requires your token).