Every importer who's wired money to China knows the moment: the supplier replies "No problem, my friend. We will try our best." And you stare at it wondering — what did they actually just agree to?
Usually: less than you thought. "No problem" is not a commitment. "We will try our best" is a soft refusal with a smile. "The price will rise next week" is often theater. And "please pay to our manager's personal account, company account has FX issues" is the opening move of the classic deposit grab.
These patterns are consistent enough that they're checkable. So I built the Supplier Reply Decoder — paste any supplier email, and it scans for the patterns that matter, in English and Chinese, entirely in your browser.
What it actually does
1. Bilingual pattern scan, client-side. Seven pattern families: vague commitments (no problem / 没问题), dodge patterns ("we will check with our engineer"), pressure tactics ("raw material price is going up"), payment-account red flags (personal account / account change), factory-vs-trading-company identity signals, specificity metrics (does the reply contain actual numbers, dates, quantities?), and question-response rate. Chinese phrases are matched natively — because the tell often hides in the language the supplier writes when they think you can't read it.
2. A risk band, not a verdict. The output groups findings into none / low / high / dealbreaker. Payment-account changes are dealbreakers regardless of how warm the rest of the email is. That asymmetry is deliberate: charm is a tactic, not a signal.
3. A generated AI prompt that carries the scan's findings. This is the part I care about most. Generic "analyze this supplier email" prompts give you generic answers. The decoder injects what it found into the prompt it builds for you — so when you paste it into ChatGPT/Claude/Gemini, the AI starts from "these five patterns were detected, confirm or debunk each in context" instead of from zero. It's a two-stage pipeline: cheap regex does the detection, the LLM does the judgment.
4. Zero dependencies, zero tracking. One HTML file, vanilla JS. The email you paste never leaves your browser — there is no server. A sourcing tool that phones home with your supplier correspondence would be ironic.
The design bet
Most "AI + email" tools want you to paste your text into their API. I'd rather give people a prompt than a login. The scan runs free forever, the AI step uses whatever chatbot the reader already has, and the value compounds: the prompt template teaches the pattern vocabulary — after a few uses, you start seeing "we will try our best" the way the scanner sees it.
It's part of a free sourcing toolkit (landed-cost calculator, supplier red-flag checker, CNY deadline tool) built on the same principles: browser-only, no signup, bilingual where it matters.
Paste your trickiest supplier reply into the decoder and see what the scan catches — the account-switch pattern alone has saved people their deposits.
(The complete system — the 12-phrase decode table, 25 negotiation templates, the 10 AI sourcing prompts — is in The China Sourcing Playbook.)
Top comments (0)