AI Lead Enrichment & Outreach Generator

🎯 AI Enrichment & Outreach Generator

Turn a flat list of leads into ready-to-send outreach in one run. The AI Enrichment & Outreach Generator takes a CSV or public Google Sheet, scrapes each company's website, detects business signals, and generates grounded, personalized outreach copy using OpenAI — all tied back to your own value proposition so every message actually sells something.

✨ Features

• 📥 Flexible input — upload a CSV or paste a public Google Sheets URL. No column mapping required if you use standard headers (first_name, company_name, website, title).
• 🌐 Live website enrichment — fetches each company's homepage and relevant internal pages (about, product, customers, blog) with Cheerio and pulls out a compact company summary.
• 🧭 Business signal detection — rule-based detection of hiring, dev/API focus, ecommerce, B2B SaaS positioning, recent content, agency positioning, and more — each with a confidence score and supporting evidence.
• 💬 Grounded outreach generation — generates icebreaker, outreach_angle, value_hook, email_subject, and linkedin_opener with OpenAI, strictly grounded in the scraped evidence and your seller offer. No hallucinated customers, no generic flattery.
• 🎯 Fit scoring & charge gating — every row gets a confidence_score (0–1) measuring fit between your offer and the lead. Set minFitScoreToCharge to automatically skip and not bill rows below the threshold — you only pay for real fits.
• 🧠 Domain deduplication — multiple leads at the same company trigger a single scrape, then generate per-lead copy from the shared context. Saves time and money.
• 📊 Dual output — writes every row to the Apify dataset and produces a downloadable OUTPUT.csv in the key-value store.
• 🚀 One-click demo — leave the prefilled inputs as-is and hit Start to see the actor enrich 3 real B2B SaaS companies (Linear, Notion, PostHog) against a sample value proposition.

🛠️ How It Works

1. Pick your lead source — upload a CSV or paste a public Google Sheets URL.
2. Describe what you sell — drop 1–3 sentences into the sellerValueProposition field. Optionally add proof points (customer names, metrics).
3. Pick an OpenAI model — gpt-4.1-mini is the sensible default.
4. Run the Actor — for each unique company domain: the Actor scrapes relevant pages, detects signals, runs OpenAI generation, and writes the enriched row.
5. Download the CSV — open the run's Storage tab, grab OUTPUT.csv from the key-value store, and drop it into your sequencer.

📥 Input

The Actor opens pre-populated with a working demo (CSV SAMPLE sentinel + Flowprint value prop) so you can click Start and see results without any setup. The fields:

Field	Type	Required	Description
inputSourceType	enum	yes	csv_upload or google_sheet.
csvFile	file	when csv_upload	Upload a CSV. Use the special value SAMPLE to load a bundled 3-row fixture (Linear, Notion, PostHog).
googleSheetUrl	string	when google_sheet	Publicly accessible Google Sheets URL. The Actor converts it to a CSV export automatically.
maxRows	integer	—	Max rows to process. Default 100, demo prefill 3.
websiteColumn	string	—	Column name holding the company website. Default website.
companyNameColumn	string	—	Default company_name.
leadNameColumn	string	—	Default first_name.
jobTitleColumn	string	—	Default title.
companyDescriptionColumn	string	—	Optional column used as extra context for the AI.
additionalContextColumns	string list	—	Any extra columns you want passed verbatim to the AI prompt.
sellerValueProposition	textarea	recommended	1–3 sentences describing what you sell, for whom, and the outcome. This is what makes the generated copy specific instead of generic.
sellerCompanyName	string	—	Your company name, used in sign-offs.
sellerProofPoints	string list	—	Facts the AI is allowed to cite: customer names, metrics, case studies.
minFitScoreToCharge	number (0–1)	—	If > 0, rows scoring below this fit threshold are marked skipped and not billed. Great for loose lead lists.
openAiModel	enum	yes	One of gpt-4.1-mini, gpt-4.1, gpt-4o-mini, gpt-4o. Pricing depends on this choice (see below).
outputTypes	string list	—	Subset of icebreaker, outreach_angle, value_hook, email_subject, linkedin_opener.
scrapeMode	enum	—	homepage_only, homepage_plus_about, or smart_multi_page. Default homepage_plus_about.
maxPagesPerCompany	integer	—	Cap on pages fetched per domain. Default 3.
tone	enum	—	professional, friendly, curious, direct, casual.
lengthPreference	enum	—	very_short, short, medium.
detectSignals	boolean	—	Run rule-based signal detection. Default true.
includeSourceUrls	boolean	—	Include the list of scraped URLs in each output row.
saveIntermediateFields	boolean	—	Include company summary and signal fields in the output row.
debugMode	boolean	—	Verbose logging.

📊 Sample Output Data

Each dataset row preserves the original CSV columns and adds enrichment + AI fields. Example:

[
    {
        "first_name": "Sarah",
        "company_name": "Linear",
        "title": "Head of Growth",
        "website": "https://linear.app",
        "processing_status": "success",
        "normalized_website": "https://linear.app",
        "scraped_urls": ["https://linear.app", "https://linear.app/about", "https://linear.app/customers"],
        "company_summary": "Linear – The system for product development. AI workflows at its core, built for modern teams.",
        "detected_signals": ["b2b_saas", "developer_or_api_focus"],
        "signal_evidence": "b2b_saas: platform | developer_or_api_focus: API",
        "icebreaker": "I noticed Linear is purpose-built for modern teams with AI workflows at its core, setting a new standard for product development.",
        "outreach_angle": "Linear's focus on self-driving product operations aligns with Flowprint's AI-guided onboarding tours to accelerate user activation.",
        "value_hook": "Flowprint can cut your new-user onboarding time by up to 40% by auto-generating interactive walkthroughs from your existing product docs.",
        "email_subject": "Faster onboarding for Linear's product users",
        "linkedin_opener": "Hi Sarah, I'm impressed by how Linear is redefining product development with AI workflows — would love to discuss growth strategies.",
        "personalization_reason": "Linear's AI-centric product development system aligns well with Flowprint's AI-guided onboarding solution.",
        "confidence_score": 0.8,
        "model_used": "gpt-4.1-mini",
        "charged_event": "result_gpt41_mini",
        "generated_at": "2026-04-09T12:14:22.108Z"
    }
]

Every run also produces OUTPUT.csv in the default key-value store — ready to drop into Instantly, Smartlead, Lemlist, Apollo, or any sequencer.

💰 Pricing

This Actor uses Pay Per Event monetization: you are charged per successfully enriched lead, and the per-result price depends on the OpenAI model you pick. Failed rows, skipped rows, and rows below your minFitScoreToCharge threshold are never billed.

OpenAI model	Tier	Price per 1,000 successful results
gpt-4o-mini	Cheapest	$3.99
gpt-4.1-mini	Lower-mid (default)	$6.99
gpt-4.1	Premium	$16.99
gpt-4o	Premium+	$19.99

Tips to control cost:

• Start with gpt-4.1-mini — it's the best quality/price tradeoff for outbound copy.
• Set minFitScoreToCharge to 0.5 once you've dialed in your value proposition so off-ICP leads don't get billed.
• Domain deduplication is automatic: 100 leads across 10 companies scrape 10 times, not 100.
• Use scrapeMode: "homepage_only" for very large lists if you don't need deep company context.

---

Go from a flat CSV to grounded, high-conversion outreach copy in minutes with the AI Enrichment & Outreach Generator. 🚀