Overseas Lead-Gen Automation for an Exporter
Background
In early 2026 — over Chinese New Year — we took on a small project. The client: a factory owner who makes work gloves. He already had overseas business, but overseas lead generation was his bottleneck.
It wasn't the product he didn't understand — it was the overseas buyer. He didn't know how to find them, how to talk to them, or how to tell if they actually needed his gloves.
When he called us, his brief was very concrete:
"Find overseas customers who might need work gloves. Send me the list on my phone so I can scan it. Once I confirm, send the emails."
We split this into three sequential steps.
Three Steps to Automate "Finding Customers"
Step 1: Find companies
We used the Google Maps API, querying by industry + geography keywords to retrieve potential buyer companies. Each query returned a batch of raw records: company name, address, industry, phone, website.
Step 2: Judge "are they a likely customer"
For each company, we automatically scanned: are there signs of relevant procurement, what does their website sell, how big are they, does the geography make sense — to judge whether this company might actually need work gloves.
This step is heavy HTTP scraping + content parsing + business reasoning. Skip it and any mail goes straight to spam.
Step 3: Write the outreach
Customers passing step 2 entered the email-writing stage. Every email is written around that specific company — what they do, how big they are, why they might need work gloves, what our price point looks like, what problem we solve.
One small detail: the client didn't initially know that overseas business runs on email. We did the research, then told him: "Yes, email is the main battlefield." So we also helped him set up the email system from scratch.
All three pieces matter
The system has three parts: find → qualify → email. Skip any one and the automation falls apart — no step 1, no pipeline; no step 2, you spam; no step 3, you send "Hi, we make gloves" into the void.
How He Reviews It All on His Phone
Writing the email is half the work. The owner isn't at his desk all day — he needs a way to scan, edit, and approve on his phone.
Around New Year, a new tool had just gone viral: OpenClaw. We integrated it immediately, hooking the whole pipeline into his Feishu (the Lark-based team chat).
Through Feishu cards pushed to his phone, he could see:
- How many potential customers were surfaced today
- Each company's snapshot + the email draft
- One-tap Send / Edit / Skip
The human approval gate sits right in the middle of the workflow — AI does the finding, qualifying, and drafting, but before anything goes out, he eyeballs it. Effectively: AI does the hunting, he decides the direction.
Results
After launch, the owner told us something we were pretty pleased about:
"Before, I could write ten emails a day and not know who I was sending them to. Now I send a hundred a day and get about ten replies."
- Manual era: ~10 emails/day, < 5 replies
- AI automation: 100 emails/day, ~10 replies
Cost-per-reply dropped to roughly 1/10. Outreach throughput lifted ~10×.
This project also validated one of our B-side automation principles: finding the customer is always step one — and always the most expensive step. The value of automation isn't "saving the few seconds of clicking send" — it's automating the two most expensive steps: who to contact and what to say.
Final Thoughts
The project isn't big, but a few things came out of it that we're proud of:
- Small projects can still drive real AI value. One owner + one automation stack = 10× outreach. AI doesn't need a giant project to deliver.
- The human approval gate is the key. Not every email should auto-send. He eyeballed it; AI hunted, he decided direction. The workflow holds.
- Use new tools early. OpenClaw had been live about a week when we plugged it in. The client gets to feel "I'm using the newest stuff," and we don't accumulate tool or tech debt.