One of the first things I automated that actually moved the needle: lead qualification. Before this workflow, I was manually reading every form submission, copy-pasting into a spreadsheet, and deciding if a lead was worth a 30-minute call. It was eating 2–3 hours a week — not a lot until you realize that compounds over months. Here's the exact flow I built in n8n: Step 1: Trigger on new form submissionTypeform / Tally / custom form → n8n webhook. Every lead entry fires the workflow immediately. Step 2: AI scoringSend the lead data to an AI node (Claude). Prompt it to score the lead 1–10 based on: business type and size, problem described, any budget signals in their free-text answers, urgency indicators. It also writes a one-sentence summary explaining why the lead matters or doesn't. Step 3: Routing logic Score 7–10 → tagged "Hot" → Slack ping to me + automated calendar link sent via email immediately Score 4–6 → tagged "Warm" → added to a 3-email nurture drip over 7 days Score 1–3 → tagged "Cold" → CRM entry + 30-day follow-up task, no immediate action Step 4: CRM entryEvery lead lands in HubSpot / Airtable with the score, AI summary, and tag already filled in. Zero manual data entry. Step 5: Morning Slack digest8am daily: a scheduled workflow pulls the last 24h of leads and sends a summary — X hot, X warm, X cold, top 3 worth reviewing today. Result: went from ~2 hours/day on lead review to about 10 minutes. Hot leads come to me automatically. Warm ones get nurtured without me touching them. The biggest mistake I see people make when building this: they use a single yes/no "is this a good lead?" prompt. The magic is the numeric score + the one-line reasoning — it gives you a paper trail and trains your eye for what actually converts over time. What does your lead qualification workflow look like right now — manual, semi-automated, or fully hands-off?

Discovered Claude Code subagents last month and it completely changed how I work. Having specialized agents that handle different parts of a project — research, coding, testing — is like having a team. Anyone else experimenting with multi-agent setups?

6 months ago I was manually researching leads, writing emails one by one, and getting maybe a 2% reply rate. Now I run a fully automated cold email system built in n8n that: ✅ Scrapes leads from multiple sources (LinkedIn, Apollo, Google Maps) ✅ Enriches each lead with AI — finds pain points, recent news, tech stack ✅ Writes a hyper-personalized first line for every single email ✅ Sequences follow-ups automatically based on behavior ✅ Books calls directly into my calendar Here's the exact workflow breakdown: Step 1 — Lead Sourcing I use a multi-source scraper that pulls leads from Apollo, LinkedIn Sales Nav export, and Google Maps depending on the niche. The data gets cleaned and deduped automatically before anything else runs. Step 2 — AI Enrichment Each lead gets run through a Claude prompt that researches the company, identifies the most likely pain point based on their industry + size, and pulls any recent trigger events (new funding, hiring spree, etc.). This takes about 8 seconds per lead. Step 3 — Personalized Email Generation Another Claude node writes the first line and email body using the enrichment data. Not a template — actually unique for each person. The difference in reply rates is insane. Step 4 — Sending + Sequencing Emails go out via Instantly (or Smartlead) with a 3-step sequence. If someone opens but doesn't reply after 3 days, they get a follow-up. If they click a link, they get a different follow-up. All logic runs in n8n. Step 5 — CRM + Calendar Integration Replies that contain certain keywords (interested, let's chat, tell me more) get flagged and pushed to HubSpot. Interested leads get a Calendly link automatically. Booked meetings sync back to my Google Calendar. The whole system runs on a $20/month n8n cloud instance. No code needed beyond the workflow setup. Reply rate went from ~2% to 8-12% depending on the niche. That's a 4-6x improvement just from better personalization at scale. Drop a comment if you want me to break down any specific node or the AI prompt I use for enrichment — happy to share the details 👇

6 months ago I was manually writing every cold email. It was taking me 3-4 hours a day and the results were inconsistent. Now I have a fully automated system running in n8n that: 1. Scrapes leads from multiple sources (LinkedIn, Apollo, company websites) 2. Enriches each lead with AI — company context, pain points, recent news 3. Writes a unique personalized first line for every single email 4. Sends via instantly.ai with proper warm-up and rotation 5. Logs everything back to a Google Sheet automatically The whole thing runs every morning at 7am without me touching it. Here's what the workflow actually looks like: Trigger → Pull new leads from source → Deduplicate against existing list → Enrich with Clay/AI (job title, company size, recent funding, tech stack) → Generate personalized opener with GPT-4o → Format email → Send via SMTP → Update CRM The personalization is what makes the difference. Generic "I noticed you work at [Company]" lines get ignored. But when you reference something specific — a recent LinkedIn post, a product launch, a hiring pattern — reply rates jump significantly. What I learned building this: - Enrichment quality matters more than volume. 50 well-researched emails beat 500 generic ones every time. - Always warm up sending domains for at least 2-3 weeks before using them - Build in a daily send cap per domain (we use 30-40/day per inbox) - Use spintax for subject lines to avoid spam filters - The follow-up sequence is where most deals actually happen — automate that too This system books us 8-12 calls per month on complete autopilot. The initial build took about a weekend to put together. Drop a comment if you want me to break down any specific part of this — happy to go deeper on the enrichment step or the GPT prompt structure that generates the personalized openers.

A few months ago I was spending 3-4 hours a day manually researching leads, writing personalised emails, and following up. Classic agency grind. Then I built a fully automated cold email workflow in n8n that handles the entire process — from lead research to personalised copy to sending — without me touching it. Here's the exact flow: Step 1 — Lead inputI feed a list of business names + URLs into a Google Sheet. n8n watches the sheet and picks up new rows automatically. Step 2 — AI enrichmentFor each lead, n8n scrapes the company website and pulls key info: what they do, recent blog posts, any pain points mentioned. This context goes to GPT-4o. Step 3 — Personalised email generationGPT-4o writes a custom opening line referencing something specific about that company — not generic "I loved your website" stuff. Real specifics. Then it slots that into a proven email framework. Step 4 — Send + trackThe email goes out via SMTP (or whatever sending tool you use). Lead status in the sheet updates automatically to "Sent." Step 5 — Follow-up sequenceIf no reply after 3 days, a follow-up gets triggered automatically. Different angle, shorter copy. Works like a drip. The result? I went from spending ~4 hours/day on outreach to about 15 minutes reviewing what went out. Reply rates actually went UP because the personalisation improved. The whole thing runs on n8n (self-hosted or cloud), costs almost nothing to operate, and took me about a weekend to build the first version. If you're running an AI agency or selling any kind of B2B service, this is probably the highest-ROI automation you can build right now. Drop a comment below if you want me to break down any specific step in more detail — happy to go deeper on the AI prompt structure or the follow-up logic 👇