Automate with AI and scripts.
Lead gen is the highest leverage area to apply AI right now. Once leads are flowing in, manual enrichment, qualification, and follow up will eat hours every week. This module shows you exactly where Claude and your favourite LLM save the most time, and gives you a downloadable toolkit of prompts and scripts to plug in directly.
A working lead gen system produces a steady flow of new leads. That's a good problem. The next problem: each lead requires research, qualification, scoring, and a personalised first response. Done manually, that's 10 to 20 minutes per lead. At 100 leads a month, that's 16 to 33 hours of repetitive work eating into your week.
Most of this work fits exactly what LLMs are good at: processing structured information, applying rules, and writing personalised text. Apps Script and Zapier handle the orchestration. The result: a lead can be enriched, scored, tagged, and handed off to your sales workflow in under 60 seconds, with no human touch required for the unqualified ones.
Where AI plugs into your existing system.
Lead
captured
Existing
Enrich
AI + script
Score
AI
Personalise
follow up
AI
CRM
handoff
Existing
Steps 01 and 05 you've already built in Module 02. Steps 02, 03, and 04 are where AI does the work. Each one is covered below.
Step 02. Enrich the lead
Your form asked for the bare minimum. Now you need everything else you'd ask if you had unlimited form space: company size, industry, role, recent news, social profiles, fit signals. Manually, this takes 10 to 15 minutes per lead. Automated, it takes seconds.
Company enrichment from email or domain
Take the email domain, look up company name, size, industry, location, and recent news. Feed everything to Claude or your favourite LLM with a structured prompt. Returns a clean profile in JSON.
↓ 10 min per lead
Role and seniority detection
From email + LinkedIn URL (if collected), determine job title, seniority, decision making authority. Useful when you didn't collect job title in the form to keep conversion high.
↓ 5 min per lead
Intent signals from notes or context
If your form has a "Tell us about your situation" field, AI can extract specific pain points, urgency level, and use case from free text answers. Tags the lead automatically.
↓ 5 min per lead
Junk lead detection
Catch the obvious bad leads: fake names, throwaway emails, competitor companies, students, irrelevant industries. AI flags them before they reach your sales team.
↓ 100% of wasted time on junk
- A trigger. Apps Script, Zapier, or your CRM watches for new lead submissions.
- A data lookup. Optional. Pulls company info from public sources or paid APIs (Clearbit, Apollo, Hunter).
- An AI call. Sends what you have to Claude (or your favourite LLM) with a structured prompt. Returns enriched data as JSON, ready to write back into the CRM.
Step 03. Score the lead
Once you've enriched, you can score. The point of scoring is to prioritise where your sales time goes first. Hot leads to call right now. Warm leads to follow up. Cold leads to the nurture sequence. Junk to the bin.
A simple lead scoring rubric you can hand to Claude:
Four buckets, one decision per lead.
Strong fit + high intent. Right industry, right company size, decision maker, immediate timeline. Route to sales for same day call.
Good fit, lower urgency. Right profile but timeline is 1 to 3 months. Personalised follow up, then nurture sequence.
Weak fit or early stage. Wrong company size, just exploring, or unclear use case. Drop into long form nurture sequence.
Not a real lead. Competitor, student, fake details, or completely off profile. Filter out, don't waste sales time.
You are a lead scoring assistant for [your company]. Score the following lead as HOT, WARM, COLD, or JUNK based on: - Company fit: [your ideal customer profile] - Role fit: [decision maker vs influencer vs end user] - Timeline signals from form responses - Budget signals from form responses Return JSON with: score, reasoning (1 sentence), suggested next action. Lead data: {paste enriched lead JSON here}
Old school point based scoring (10 points for senior title, 5 points for matching industry) misses nuance. AI can read a free text answer like "we've been looking for something like this for 6 months" and recognise high intent immediately. The same prompt scales from 10 leads a month to 10,000, with consistent reasoning across every one.
Step 04. Personalise the follow up
Generic templates feel generic. Real personalisation lifts reply rates significantly. Done manually, writing a personalised follow up takes 5 to 10 minutes per lead. Done with AI plus your enriched data, it takes 5 seconds and the result is often better than what you'd write at the end of a long day.
Three follow up flows that benefit most from AI:
First touch email after form submission
Reference their company, the specific magnet they downloaded, and one observation about their situation from the enrichment. Send within 5 minutes for max impact. AI drafts it, you approve and send (or auto send if you trust the prompt).
LinkedIn connection note
If you reach out on LinkedIn after they've downloaded a guide, AI can craft a 200 character connection note that references something specific about them. Pre filtered by your scoring so you only do it for HOT and WARM leads.
Re engagement after silence
Lead went cold? AI can generate a contextual nudge based on what's happened since their last interaction. New product update, relevant case study, or just a check in framed around their original problem.
Lead Generation toolkit: prompts, scripts, and templates.
Everything you need to put this module into practice. Copy paste ready prompts for Claude. Apps Script snippets for orchestration. Zapier templates. Plug them into your existing CRM and stack.
Lead enrichment prompt for Claude
Lead scoring prompt with rubric
Personalised follow up email prompt
Apps Script: form to enrichment trigger
Apps Script: write enriched data to CRM
Zapier templates and webhook setup
The stack and the trade offs
You don't need a complex stack to run this. The three core pieces:
An LLM with API access
Claude API is what we recommend, but ChatGPT, Gemini, or any other commercial LLM API works. Cost is typically $0.01 to $0.10 per lead processed, depending on prompt length and model.
An automation layer
Apps Script if you live in Google Sheets and want maximum control (free). Zapier or Make if you want a visual builder ($20 to $50 a month). Both work, pick what you'll actually maintain.
Your CRM as the destination
Whatever CRM you picked in Module 02. Enriched data writes back as custom fields. Scored leads route to the right pipeline stage. No new tools needed.
Don't fully auto send messages until you've reviewed at least 50 outputs. AI can hallucinate, get tone wrong, or miss context. Start with "AI drafts, human reviews" mode for the first few weeks.
Privacy and data handling. If you're sending lead data to an LLM API, check your data processing agreement and your privacy policy. For B2B SaaS in Europe especially, GDPR matters.
Don't over engineer. Adding 6 steps to the pipeline that each shave 30 seconds is rarely worth the maintenance cost. Start with one or two automations. Add more only when the manual work has clearly become repetitive.
What you should have at the end of this module
An LLM API key (Claude or your favourite) and an automation tool (Apps Script or Zapier)
An enrichment workflow that adds company, role, and intent data to every new lead
A scoring prompt that classifies leads as HOT, WARM, COLD, or JUNK
A personalised first touch email generated and sent within 5 minutes of submission
The downloaded toolkit with prompts and scripts ready to plug in
You've got the system. Now feed it consistently.
Pick your channel, build your magnet, run your campaigns, automate the operational grind. The brands that win with paid social lead gen aren't the ones with the smartest tactics, they're the ones who run the system every month and let the compounds compound. After 6 months of consistent running, you'll have a database, a scoring engine, and a follow up system most agencies would charge thousands to build.
Back to the Course Library →