Field notes from China's export economy

AI systems that survive contact with a real small business.

I build AI workflows for small companies that sell, source, or operate across borders — and I publish what actually happens: the workflows, the failures, and the numbers. No vendor demos. No thought leadership. Field notes.

  • 16 years running cross-border commerce before AI was a pitch deck
  • 4 years deploying AI inside real export & manufacturing SMEs
  • One rule every number published here comes from a live system
From a live deployment

An export trading company stopped losing overnight inquiries.

Overseas buyers write at 3 a.m. China time. Before: first human reply the next afternoon, and multi-product inquiries from the same buyer were answered piecemeal by different people. After: every inquiry is parsed, deduplicated per buyer across mailboxes, grounded against a verified product knowledge base, and a draft reply waits for human sign-off by morning — with hard rules on what the AI may never promise. Full teardown with numbers coming as a field note.

Company size
~20 people
Workflow
Inquiry → qualified lead
Human handoff
Price, lead time, certification
Status
In production
Work with me

Start with an AI Workflow Audit.

One week. I interview your team, map 3–5 of your repetitive workflows, and give you a prioritized 90-day plan with honest cost and risk estimates — including the workflows where AI is not the answer. If we build together afterwards, the audit fee is credited in full.

Who writes this

I'm Li Long — founder of Houkongfan, a one-person AI implementation practice based in Jiangyin, China, in the middle of one of the world's densest manufacturing and export regions. I spent 16 years building cross-border e-commerce businesses before spending the last four putting AI to work inside real SMEs.

This site exists because almost everything written about "AI for business" is written by people selling software. I implement, then I write down what happened — including the parts that failed.