Running a sufferance and bonded warehouse near the Port of Montreal, I deal with hundreds of incoming shipments, customs documents, and client communications every week. Until recently, most of this was manual. Over the past few months, I've automated significant parts of our operation using a combination of AI tools, custom scripts, and integrations. Here's what we built and what worked.
The Problem
Our business has three main bottlenecks:
- Customer outreach — We send hundreds of personalized emails per week to potential clients in Europe, but writing each one manually is impossible at scale.
- Customs documentation — Every shipment requires accurate tariff classification and CARM-compliant filings.
- Lead qualification — Inquiries come in from a website contact form, email, and now a chatbot, but sorting them by priority took hours per day.
What We Built
1. Automated Email Outreach Pipeline
We built a system that:
- Scrapes public business directories for potential clients (logistics companies, freight forwarders in EU countries)
- Extracts contact information and enriches it with company data
- Generates personalized cold outreach emails using a template engine
- Sends emails through Resend API at a controlled rate (50,000/day capacity)
- Automatically schedules follow-up sequences (Day 3, Day 7, Day 14)
- Stops the sequence if a contact replies or unsubscribes
The result: instead of sending 50 manual emails per week, we now send 3,000+ targeted emails monthly with zero manual effort.
2. AI Customer Support Chatbot
We integrated an AI chatbot (Crisp + custom knowledge base) on our website that:
- Answers common questions about customs, CETA tariffs, and warehousing
- Trained on a curated knowledge base of our policies and services
- Routes complex inquiries to human staff
- Captures leads automatically into our CRM
Conversion rate from chatbot to qualified lead: about 12%.
3. Real-Time Keyword Research with Google Ads API
For SEO, we integrated the Google Ads API to:
- Pull real monthly search volumes for industry keywords
- Identify high-opportunity terms (low competition, high intent)
- Auto-generate SEO content focused on these terms
- Skip low-volume terms entirely (avoiding wasted content production)
Result: 350 keywords analyzed, only 57 had real search volume. Our content team now focuses exclusively on those 57.
4. Automated SEO Content Pipeline
A cron job runs daily at 8 AM Montreal time and:
- Pulls trending logistics news from RSS feeds
- Uses AI to rewrite the news from a Canadian warehousing perspective
- Binds each article to a real keyword from our database
- Auto-generates internal links to money pages
- Publishes the article and notifies Google via the Indexing API
- Posts a summary to our Google Business Profile
We're publishing 2-3 SEO-optimized articles daily without manual intervention.
Lessons Learned
1. Start with the bottleneck. We started with email outreach because it was eating the most time. Don't automate things that aren't actually painful.
2. AI is great for content, but humans should review prompts. We use templates and AI for speed, but every prompt was carefully designed. Pure AI output without human guidance produces generic results that don't convert.
3. Measure everything. Without analytics, you can't tell if your automation is helping or hurting. We track open rates, reply rates, conversion rates, and adjust constantly.
4. Don't over-engineer. Our entire automation stack runs on simple cron jobs and a single Postgres database. No Kubernetes, no microservices, no Kafka. Just bash scripts and Node.js. Total infrastructure cost: under $50/month.
Tech Stack
- Backend: Next.js + Prisma + PostgreSQL
- Email: Resend (replaced Gmail SMTP after hitting limits)
- Hosting: Single VPS with PM2 + cron
- AI: Anthropic Claude API for content generation
- Search Data: Google Ads API for keyword research
- Chatbot: Crisp with custom knowledge base
- Web Scraping: Custom Python scripts with rate limiting
What's Next
We're currently working on:
- Predictive shipment routing based on historical data
- Auto-generated quote system using cargo type and volume
- Multi-language support for Spanish, German, and French clients
Building a real business requires more than just code — but smart automation lets a small team punch way above its weight. If you're running a small logistics or B2B operation and feel overwhelmed, start by writing down the three things you do most often. Two of them can probably be automated.
Tony Gu is the Managing Director of FENGYE LOGISTICS, a CBSA-licensed sufferance and bonded warehouse near the Port of Montreal. Learn more at fywarehouse.com.
Top comments (0)