Samuel Oladele
Certified Automation Expert | Voice Agents | API Expert
New to Contra
Samuel is building their profile!
I built an AI email agent for a towing company. Here's what actually made it hard. 🧵
The client was drowning in emails. Vehicle releases, invoice disputes, insurance adjusters, frustrated customers. Repetitive, time-sensitive, and legally sensitive. A perfect automation target.
I built the whole thing in n8n with GPT Here's how the pipeline works:
1 → Gmail trigger polls every minute for new emails
2 → Filter out self-sent emails to prevent reply loops
3 → AI classifies urgency, category, tone, and sender intent
4 → Business hours check (America/Toronto timezone)
5 → High urgency → instant Telegram alert to the owner
6 → 30-min wait, then fetch the full thread
7 → Check if the last reply was already theirs — skip if so
8 → AI drafts and sends a contextual acknowledgment
The hardest part wasn't the pipeline. It was the system prompt.
Towing companies operate in a legally grey space — releasing a vehicle, confirming a payment, or acknowledging a balance can carry real liability. So the AI had to be useful without ever being authoritative. Every response needed to feel human and helpful, while deferring every actual decision to billing or dispatch.
I spent more time on prompt guardrails than on the n8n nodes. Things like: don't use names from the incoming greeting (they're usually addressing internal staff, not the sender), never validate a claim even if the sender says it's already been approved, and route Interac deposit notifications silently — no reply, just a Telegram ping to the owner.
Thread memory via a custom session key keeps the AI context-aware across multi-message conversations. Structured output parsers keep responses predictable. The whole thing runs unattended.
Clients now get a reply in under a minute. The team gets a Telegram summary. No one touches a routine email anymore.
If you're an automation dev, the unglamorous truth is: the plumbing is easy. Understanding the client's legal and operational risk is where the real work lives.
#n8n (https://www.linkedin.com/search/results/all/?keywords=%23n8n&origin=HASH_TAG_FROM_FEED) #AIAgents (https://www.linkedin.com/search/results/all/?keywords=%23aiagents&origin=HASH_TAG_FROM_FEED) #WorkflowAutomation (https://www.linkedin.com/search/results/all/?keywords=%23workflowautomation&origin=HASH_TAG_FROM_FEED) #PromptEngineering (https://www.linkedin.com/search/results/all/?keywords=%23promptengineering&origin=HASH_TAG_FROM_FEED) #LLM (https://www.linkedin.com/search/results/all/?keywords=%23llm&origin=HASH_TAG_FROM_FEED) #Automation (https://www.linkedin.com/search/results/all/?keywords=%23automation&origin=HASH_TAG_FROM_FEED) #BuildInPublic (https://www.linkedin.com/search/results/all/?keywords=%23buildinpublic&origin=HASH_TAG_FROM_FEED) #OpenAI (https://www.linkedin.com/search/results/all/?keywords=%23openai&origin=HASH_TAG_FROM_FEED) #NoCode (https://www.linkedin.com/search/results/all/?keywords=%23nocode&origin=HASH_TAG_FROM_FEED) #FreelanceDev (https://www.linkedin.com/search/results/all/?keywords=%23freelancedev&origin=HASH_TAG_FROM_FEED)
0
7
I built a fully automated B2B lead generation pipeline that runs while I sleep.
Here's what it does in one workflow:
→ Scrapes Google Maps for businesses by industry, location & radius
→ Filters out duplicates and already-processed leads
→ Uses GPT to visit each company website and qualify the lead
→ Extracts the decision-maker's name, email, and phone number
→ Pushes qualified leads into Google Sheets and Zoho Bigin as contacts + companies
→ Flags leads that don't have a verified company email
Zero manual prospecting. Zero copy-pasting between tabs.
The whole thing runs on n8n and costs a fraction of what any lead database subscription would.
Automation isn't just for big tech companies. If you know your ICP, you can build this.
#n8n #Automation #SalesAutomation #LeadGeneration #B2BSales #NoCode
0
12
My client was spending hours manually searching Google Maps, copy-pasting emails, and cold outreaching prospects one by one.
I automated the entire thing in n8n.
Here's the exact workflow I built 🧵
The client sells GPS tracking solutions to logistics and fleet companies. The challenge? Finding the right businesses, getting the right contact, and sending a personalized pitch — at scale.
So I built a 5-step engine:
① Form Trigger — client fills in keyword, city, industry, radius
② Apify scrapes Google Maps → returns businesses with websites
③ GPT visits each website, scores the lead, extracts the decision maker's name, position, and contact details
④ Qualified leads auto-save to Google Sheets
⑤ GPT personalizes the email → Gmail sends it automatically
The AI doesn't just scrape. It thinks.
It identifies fleet signals ("mentions delivery vans"), finds pain points, and even writes the recommended pitch angle for each company.
Result: From 4 hours of manual work → fully automated in under 15 minutes per run.
This is what selling automation services looks like in 2026.
♻️ Repost if this was useful
💬 Comment "workflow" if you want to see the n8n JSON
#n8n #Automation #AIAgents #LeadGeneration #FreelanceAutomation #B2BSales8
0
6
Most clinics are still losing appointments to missed calls and manual follow-ups.
I built a voice AI receptionist that handles the entire booking flow automatically.
Here's what it does:
📞 Answers inbound calls in a natural, human-like voice (powered by ElevenLabs)
🧠 Recognises returning callers by checking a connected Google Sheet in real time
📅 Books appointments directly — no human intervention required
🔔 Instantly alerts the clinic team the moment a booking is confirmed
⚙️ Orchestrated end-to-end through n8n automation workflows
The result?
Front desk staff reclaim hours every week. No calls go unanswered. No leads fall through the cracks.
If your clinic is still relying on manual booking or a basic contact form, there's a better way.
Built with: ElevenLabs · n8n · Google Sheets · Google Calendar
0
13