B2B Prospecting on Autopilot: How Claude + Apify Build a Scored List of 100 Targets for $2.50

The hardest step in B2B is knowing who to email. In B2C a pixel can find a person; in B2B the target is known, the problem is listing the right 100 accounts. The video shows an end-to-end shortcut: hand over your own website, give a single prompt, let the agent do the rest through Apify, and get back both a visual HTML and an Excel file.

What Is Apify and Why “Epify”?

The tool called 'Epify' in the video is Apify — an automation marketplace that pulls Google Maps, Instagram, marketplaces and Shopify data through one pipe. Ready-made scrapers called Actors do the heavy lifting: the Compass Google Maps Extractor starts around $2.10 per 1,000 places, has over 91,000 users, a 98.8% success rate and a 4.85 rating. It can collect address, phone, website, review count and hours from Maps cards; profiles and links from Instagram; seller details from marketplace pages. You can start on the free tier and pay as you go as you scale, with community coupons to cut costs further.

Claude Desktop and the Single Prompt

The other half is Claude’s desktop app. The browser version is limited on file-system and automation permissions; once you download the desktop app it can save files, create Excel sheets and open HTML in your browser in one run — switch to terminal mode and it gets even more autonomous. In Settings you copy your Apify key from the API section and paste it into Claude. Then comes the single prompt: 'I will give you my company; find the best customer segment, pull up to 100 companies, estimate the cost, return names, sites and emails scored in detail, open the result as HTML and save to Excel; use Apify — pull from Maps, Instagram or marketplaces like Trendyol and Hepsiburada as you see fit, you define the target.' As the demo, the speaker hands over Dekor Vizyon, his own platform that generates product images for furniture — no need to describe the buyer, the agent reads the site and builds the niche 'furniture makers and sellers' on its own.

Once the prompt is sent the agent runs in auto mode and finishes setup without asking. The raw pool arrives at 307 records, 207 remain after filtering; among them 62 have an e-commerce site and 96 have an email, and the system scores and picks the top 100. Total spend is calculated at $2.50 with about 17,000 tokens — the author notes that one million tokens is the real large threshold. Opus is used as the model, with a note that Sonnet would handle the job as well; the work is in orchestration, not raw model power. The cost is Apify scraping plus Claude tokens, symbolic next to an ad budget.

The output is two files: a scored HTML table opened in the browser and the same data as Excel. Each row shows company name, website, email, Instagram, review count and e-commerce presence. The pro tip: info@ addresses convert at lower rates, so adding 'also find the general manager at each company' to the same flow lets the agent enrich the list with decision-makers through the same Apify pipe. You end up with a double-layer, directly callable list — work that would take hours manually is packed in minutes with one prompt.

The scale promise is blunt: 100 today, 1,000 or 10,000 tomorrow with the same prompt. Change the site and the agent rebuilds the niche; whether furniture or another vertical, the same line retargets. For B2B SaaS, emailing 100 well-chosen accounts directly can beat chasing pixels and broadcasting ads to everyone — as long as you verify the list and respect permission law. Ads speak to all; this method speaks only to those who can actually use your product, so the odds start higher.