Verdict: You can build a fully automated lead generation pipeline — one that discovers local businesses, analyzes their websites and social presence, scores them 0–100 on need, and writes a personalized pitch for each — using three tools that together cost nothing on their free plans. Hermes Agent (open-source AI agent by Nous Research) orchestrates the research, Apify (a web scraping marketplace) collects business data via ready-made "actors," and Airtable (a no-code database) stores your leads. Set it up once as a reusable skill, schedule it with a cron job, and it runs on its own — finding qualified prospects while you do other work. The entire pipeline works for any business type in any city, and the free tiers of all three platforms cover your first 200–500 leads.
Last verified: July 31, 2026 · Best free stack: Hermes Agent + Apify + Airtable · Best for: agencies, freelancers, SMBs selling local services · Cost: $0–$5/month on free plans
What Is Automated Lead Generation With AI Agents?
Automated lead generation with AI agents is the process of using an autonomous AI agent to discover, research, qualify, and pitch prospective customers without manual intervention at each step. The agent scrapes business directories (like Google Maps), analyzes each prospect's web presence to identify gaps, scores them on how likely they need your service, writes a personalized pitch, and saves everything to a database — all on a schedule you define once.
The key difference from traditional lead-gen tools: instead of paying per seat for a closed SaaS platform, you connect an open-source agent to marketplace scrapers and a spreadsheet-database. The agent does the analysis and writing — the creative judgment work — while the scrapers handle data collection.
How Does This Pipeline Actually Work? (The Architecture)
The pipeline has four stages that the agent runs sequentially for each batch of prospects:
- Discovery — A Google Maps scraper (Apify actor) finds businesses matching your niche and city, returning names, addresses, phones, websites, ratings, and review counts.
- Audit — For each business found, the agent takes a screenshot of their website (via a screenshot actor), reads the site content (via a website crawler actor), and checks whether they are running Facebook ads (via a Facebook Ads scraper actor).
- Scoring & Gap Analysis — The agent evaluates everything it collected and scores each business 0–100 on how much they need your service. It identifies specific gaps: outdated website, no social media presence, not running ads, low Google rating, incomplete Google listing.
- Pitch + Save — The agent writes a one-paragraph personalized opener referencing the specific problems it found, then saves the entire record — business details, audit notes, score, recommended services, and pitch — to your Airtable.
Once you turn this into a reusable skill and schedule it, the entire four-stage pipeline runs from a single line: "run lead-finder for plumbers in Chicago for 5 leads."
Why this beats manual prospecting or a static lead list
Manual prospecting means searching Google Maps, opening each business one at a time, checking their website, looking at their socials, and taking notes. For 20 prospects that is 2–3 hours of clicking. This pipeline does it in minutes and produces a scored, pitched table ready for outreach. Unlike buying a lead list, the data is fresh — scraped the day the pipeline runs — and each lead comes with an evidence-based pitch rather than a generic template.
What Tools Do You Need? (Costs and Free Plans)
Three tools make up the stack. All three have free plans sufficient to test and run small batches.
| Tool | What It Does | Free Plan | Free Plan Limits | Source |
|---|---|---|---|---|
| Hermes Agent | Open-source AI agent that orchestrates the research, analysis, scoring, and writing | Free tier available; self-host on a $5/mo VPS or run locally | Depends on your LLM provider costs; the agent itself is MIT-licensed and free | hermes-agent.nousresearch.com |
| Apify | Marketplace of web scraping "actors" — pre-built scrapers for Google Maps, Facebook Ads, websites, screenshots | $5 in monthly compute credits; no credit card required | Credits expire monthly (no rollover); 5 datacenter proxy IPs; 8 GB max actor RAM | apify.com |
| Airtable | No-code database to store leads with rich fields, attachments, and views | Free: up to 5 editors, 1,000 records per base, unlimited bases | No automations on free plan; 1,000 records per base | airtable.com |
Total cost for your first batch: $0. The Apify free plan's $5 monthly credits cover roughly 2,500 Google Maps place results at ~$2 per 1,000 places (Confirmed, per Apify's free plan documentation). Airtable's 1,000-record limit per base is enough for several hundred leads with room to spare. Hermes Agent is open-source under the MIT license (Confirmed, Nous Research).
Which Apify Actors Do You Need?
Apify's marketplace hosts thousands of actors. Four are essential for this pipeline. Reference them by their exact actor IDs so the agent never guesses.
| Purpose | Actor ID | What It Returns | Pricing |
|---|---|---|---|
| Find businesses on Google Maps | compass/crawler-google-places |
Business name, address, phone, website, Google rating, review count, category, coordinates | Pay-per-event (~$2/1,000 places on free plan) |
| Screenshot each website | apify/screenshot-url |
A PNG screenshot of the business homepage | Free actor; you pay only platform compute |
| Check if they run Facebook ads | automation-lab/facebook-ads-library |
Active/inactive ad status, ad creatives, start dates | $0.005/run start + $0.0005/ad scraped |
| Crawl and read website content | apify/website-content-crawler |
Full-text content as Markdown, page metadata, social links | Pay-per-usage (~$0.20–$5/1,000 pages) |
These actor IDs are verified against the Apify Store as of July 2026. The Google Maps scraper (compass/crawler-google-places) is the most popular Maps actor on the platform, with over 6,400 total users (Confirmed, Apify actor page). The Website Content Crawler is Apify's own flagship content actor, maintained by Apify themselves (Confirmed, apify.com/apify/website-content-crawler).
Important: Apify credits do not roll over. Unused credits expire at the end of each billing cycle (Confirmed, Apify free plan docs). Run your pipeline regularly to use what you get.
How Do You Set Up Hermes Agent for the First Time?
Hermes Agent is an open-source, self-improving AI agent developed by Nous Research and first released in February 2026. It runs from a terminal, a desktop app, or a web dashboard, and supports persistent memory, reusable skills, cron jobs, browser automation, and MCP (Model Context Protocol) tool integrations (Confirmed, hermes-agent.nousresearch.com).
Step 1: Install Hermes Agent
You can install it locally on macOS, Windows, or Linux, or host it on a cloud VPS. For the automated pipeline to run on a schedule, you need it running on a server or your local machine staying awake.
# Install via terminal (macOS/Linux)
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
# Or download the desktop app from the website
# Or deploy on a VPS (one-click install available on Hostinger)
During setup, select your LLM provider (OpenAI, Anthropic, or others), connect your messaging platforms if desired, and enable the tools you need (web search, browser, terminal).
Step 2: Connect MCP Servers for Apify and Airtable
Hermes Agent uses MCP (Model Context Protocol) to connect to external tools. You need two MCP servers:
Airtable MCP — gives the agent read/write access to your Airtable base:
- In Airtable, go to Profile → Builder Hub → Personal Access Tokens
- Create a token named "Hermes" with read/write permissions for records and bases
- In Hermes, tell it: "Set up Airtable MCP with this token: [paste token]"
Apify MCP — gives the agent access to Apify actors:
- In Apify, go to Settings → API & Integrations → copy your personal API token
- In Hermes Dashboard, go to Keys → Settings → Custom Keys → add "Apify token"
- Tell Hermes: "Set up Apify MCP server with this token: [paste token]"
After setup, verify both connections by asking Hermes to list Airtable fields and available Apify actors. You should see your Airtable columns and a list of Apify actors.
How Do You Build the Lead Generation Pipeline Step by Step?
Step 1: Create Your Airtable Base
Set up a table in Airtable with these columns to capture the full lead record:
| Column Name | Type | Purpose |
|---|---|---|
| Business Name | Single line text | Company name |
| Niche | Single line text | Business category |
| City | Single line text | Location |
| Address | Long text | Full address |
| Phone | Phone | Contact number |
| If found on site | ||
| Website | URL | Homepage URL |
| Google Rating | Number | Star rating (0–5) |
| Review Count | Number | Total reviews |
| Website Verdict | Long text | Agent's assessment of site quality |
| Running Ads | Checkbox | Yes/no — are they on Facebook Ads Library? |
| Ad Notes | Long text | Details of ad activity |
| URL | Facebook page link | |
| Social Activity | Single select | Active / Inactive / No Presence |
| Social Notes | Long text | Instagram, Facebook activity summary |
| Gaps Found | Long text | Specific problems found |
| Recommended Services | Long text | Services they likely need |
| Priority | Single select | High / Medium / Low |
| Pitch | Long text | Personalized outreach opener |
| Score | Number | 0–100 lead score |
Step 2: Run the Pipeline Manually First
Before automating, walk the agent through the pipeline once so you can verify the output quality. Send these messages to Hermes in sequence:
Message 1: "List all the fields in my Airtable Leads table." (Confirms the connection works.)
Message 2: "Find 5 dentists in Austin using the Apify actor compass/crawler-google-places. For each, get their name, website, phone, Google rating, and review count."
The agent calls the Google Maps scraper, retrieves the results, and lists them.
Message 3: "For each of these dentists: (1) Take a screenshot of their website using apify/screenshot-url. (2) Check if they are running ads using automation-lab/facebook-ads-library. (3) Read their website and check social activity using apify/website-content-crawler. Then populate all fields in my Airtable."
The agent runs all four actors, analyzes the results, and writes a complete row for each business into Airtable — including a website verdict ("outdated, no mobile optimization"), ad status, social presence, gaps found, recommended services, priority, and a personalized pitch.
Step 3: Define Your Lead Scoring Framework
This is where you add your own judgment. The scoring criteria below is an original framework — adjust the weights for your service.
| Dimension | Points | What to Evaluate | Why It Matters |
|---|---|---|---|
| Website quality | 0–25 | Is the site mobile-friendly, modern, fast? Screenshot + content crawl | A bad website = clear service opportunity |
| Ad activity | 0–20 | Are they running Facebook or Instagram ads? | Not running ads = untapped channel |
| Social presence | 0–20 | Are they posting on Facebook/Instagram? How recently? | Dormant social = management gap |
| Google listing | 0–15 | Is the profile complete? How many reviews? What rating? | Low rating or few reviews = reputation work needed |
| Review sentiment | 0–10 | Are there negative reviews mentioning specific problems? | Complaints reveal exactly what to sell |
| Contactability | 0–10 | Is there a phone, email, or contact form? | Determines outreach channel options |
| Total | 0–100 |
Score interpretation:
- 70–100: High priority — multiple clear gaps, pitch is straightforward
- 40–69: Medium priority — some opportunities, needs a more nuanced pitch
- 0–39: Low priority — they are already doing well; table for later
This scoring framework is the information gain this guide contributes. Most lead-gen tutorials stop at "find businesses and save them." Scoring against a transparent rubric means your pipeline does not just find leads — it ranks them by what they actually need.
Step 4: Turn the Pipeline Into a Reusable Skill
Once you are satisfied with the manual output, tell Hermes to save the entire workflow as a skill:
"Turn everything we just did into a reusable skill called 'lead-finder.' Use these actors:
compass/crawler-google-placesto find businesses,apify/screenshot-urlto screenshot the site,automation-lab/facebook-ads-libraryto check ads, andapify/website-content-crawlerto rate the site and socials. Score each on a scale of 0 to 100 using these criteria: [paste your scoring framework]. Write a personalized opener referencing the specific gaps found. Save everything to my Airtable Leads table."
Hermes Agent's skill system is one of its defining features — it writes reusable workflow files to its memory so you can invoke the same multi-step process with a single command later (Confirmed, hermes-agent.nousresearch.com; AIToolTier review).
Now, any time you want leads, you say:
"Run lead-finder for plumbers in Chicago for 5 leads."
One line. Different niche, different city, same pipeline. The skill handles everything.
Step 5: Automate It With a Cron Schedule
The final step is to stop asking the agent manually. Hermes Agent has a built-in cron scheduler — describe the schedule in natural language and it creates the job (Confirmed, hermes-tutorials.dev):
"Every day at 2am and 11pm, run lead-finder and rotate through this list: dentists in Austin, plumbers in Chicago, electricians in Dallas. Add new leads to my Airtable."
This creates a scheduled job that executes the skill automatically at those times. For cron to work, Hermes Agent needs to be running on a server (or your local machine must stay awake). If you are running it on a $5/month VPS, it stays on permanently.
You can also check up on what it does — see our guide to how AI agent task boards work for managing autonomous workflows like this one.
How Much Does This Actually Cost Per Lead?
Here is a realistic cost breakdown for a batch of 50 leads using the Apify free plan:
| Step | Actor | Volume | Estimated Cost |
|---|---|---|---|
| Find businesses | compass/crawler-google-places |
50 places | ~$0.10 (at ~$2/1,000 places) |
| Screenshot websites | apify/screenshot-url |
50 URLs | ~$0.15 (platform compute only) |
| Check Facebook ads | automation-lab/facebook-ads-library |
50 searches | ~$0.28 ($0.005/run × 50 + ad scrape) |
| Crawl website content | apify/website-content-crawler |
50 sites, ~2 pages each | ~$0.10 (lightweight pages) |
| Total for 50 leads | ~$0.63 |
That is roughly $0.01 per lead — and it stays within the free plan's $5 monthly credit allowance. For comparison, paid lead-list services and manual prospecting research typically cost $0.15–$5.00 per lead (Reported, per use-apify.com's lead-gen playbook).
Airtable cost: $0 on the free plan as long as you stay under 1,000 records per base. If you outgrow it, the Team plan starts at $20/user/month (Confirmed, Airtable pricing).
What Can You Do With the Scored Lead Data?
Once your pipeline is running and accumulating scored leads in Airtable, several next steps follow naturally:
Build an audit report
For high-score leads (70+), turn the agent's gap analysis into a one-page audit report. The data is already in Airtable — the website screenshot, the social assessment, the ad status, the recommended services. Package it as a PDF and attach it to your outreach email as proof you actually looked at their business.
Extend the pipeline to outreach
Hermes Agent can also handle the outreach — sending emails or messages to qualified leads, tracking responses, and following up. This requires connecting your email or messaging platform as another MCP server. For a comparison of how autonomous AI agents handle multi-step business workflows, see our analysis of AI agent orchestration patterns.
Use the pitch as a warm intro, not a cold email
The pitch the agent writes is specific: "I noticed your website is not mobile-optimized and you are not running any Facebook ads — here is what that is costing you." That converts a cold outreach into a warm one because you are referencing concrete findings, not a generic template. The businesses most likely to respond are the ones with the most gaps — exactly the ones the pipeline surfaces as high-score leads. Our deeper analysis of when not to use AI applies here too: use the agent for the research grind, but keep a human in the loop for the actual outreach until you trust the quality.
What Are the Pitfalls and Limitations of This Approach?
Hallucination risk in analysis
The agent evaluates website quality and social activity based on scraped content. If a scraper fails silently (returns empty data), the agent may still write a verdict. Always spot-check the first batch manually, and if you're not sure when AI analysis helps vs hurts, our deterministic-first rule explains how to grade agent claims.
Apify credits expire monthly
Unlike most SaaS platforms, Apify free credits do not roll over. If you do not use the $5 in a given month, it is gone. Schedule a small run early in the month to confirm the pipeline works, then larger runs later (Confirmed, Apify free plan documentation).
Airtable's 1,000-record limit
If you are generating leads across many niches and cities, you will hit the 1,000 records per base limit on the free plan. Options: archive old leads to a separate base, delete low-score leads periodically, or upgrade to Team ($20/user/month) for 50,000 records per base (Confirmed, Airtable pricing).
Cron requires a running server
If you are running Hermes Agent on your local machine, it must stay awake for cron jobs to fire. For reliable automation, host it on a VPS. If you're exploring self-hosting options, our guide to building an AI agent operating system covers the infrastructure decisions.
Email addresses are rarely in Google Maps data
Google Maps listings include phone numbers and websites but usually not email addresses. To get emails, you need a separate email enrichment actor or tool — add it as a fifth stage to your pipeline once the core four stages are working.
Scraping legal considerations
Business listing data from Google Maps is generally considered public. The Facebook Ads Library is explicitly public — Meta built it for transparency (Confirmed, facebook.com/ads/library). However, how you use the data matters: respect each platform's terms of service, and avoid scraping personal data (individual reviewer info, for example) without a legitimate purpose.
What This Means for You
If you sell a service to local businesses — web design, SEO, social media management, advertising, bookkeeping, anything — this pipeline replaces your manual prospecting with a system that does the research, analysis, and pitch-writing for you. The free tiers of all three tools are enough to validate the concept and generate your first 200+ leads. You own the pipeline because it is built on open-source and marketplace components, not a locked SaaS that raises prices or changes features without notice.
Start with one niche in one city. Run the pipeline manually first, verify the output quality, then turn it into a skill and schedule it. Within a week you will have a self-running lead engine that produces scored, pitched prospects on a schedule you control.
FAQ
Q: What is automated lead generation with an AI agent? A: It is a pipeline where an autonomous AI agent discovers businesses (via web scrapers like Apify's Google Maps actor), analyzes each prospect's website, social media, and ad activity, scores them on need, writes a personalized pitch, and saves the full record to a database — all on a schedule, without manual intervention per lead.
Q: How much does the Hermes Agent + Apify + Airtable stack cost? A: $0 on the free plans. Apify gives $5 in monthly compute credits (enough for ~2,500 Google Maps results). Airtable's free plan allows 1,000 records per base with 5 editors. Hermes Agent is open-source and free under the MIT license. Paid upgrades start at $29/month for Apify (Starter) and $20/user/month for Airtable (Team).
Q: Can this work for any type of business, not just dentists? A: Yes. The pipeline is generic: you pass it a business type and a city, and the Google Maps scraper finds matching businesses. "Plumbers in Chicago," "electricians in Dallas," "landscapers in Phoenix" — the same skill handles all of them with no configuration changes.
Q: Do I need to know how to code to set this up? A: No. Hermes Agent accepts natural-language instructions — you describe the pipeline in plain English and it connects the MCP servers, calls the Apify actors, and writes to Airtable. The scoring criteria and pitch templates are also defined in natural language when you create the skill.
Q: How often should I run the pipeline? A: Start with twice daily (e.g., 2 AM and 11 PM) for one niche-city combination at a time. Each run of 5 leads costs about $0.06 in Apify credits, so twice daily for a month is ~$3.60 — well within the $5 free allowance. Scale up frequency or batch size once you confirm quality.
Q: What is the difference between this and buying a lead list? A: A purchased lead list is static — names, phones, and maybe emails, with no context about what the business needs. This pipeline produces fresh, scored leads with an evidence-based pitch: it knows the website is outdated, they are not running ads, and their last Facebook post was 4 months ago. You reach out with a specific reason, not a generic template.
Discussion
0 comments