How to Automate B2B RFQ Responses Using AI (Without Hiring Sales Reps)
Auto-matching OE numbers, generating quotes, and replying to buyer inquiries with AI. Built and running on jjradiator.com with 112 customers.
A public record of building jjradiator.com with AI: site structure, SEO content, product data, Search Console results, inquiry workflows, and mistakes along the way.
The first articles are not theory. They explain what happened while using AI to build, publish, verify, and improve the site.
Auto-matching OE numbers, generating quotes, and replying to buyer inquiries with AI. Built and running on jjradiator.com with 112 customers.
Batch AI translation pipeline that turned a Chinese catalog into 15,000 Google-indexed English pages. Why Google Translate fails.
Capture leads from multiple channels, sort them with AI, and run human-reviewed follow-up sequences. Designed for one person.
An AI script that watches A-share bank stocks and Shanghai Gold, then pushes an alert. Same pattern for exchange rates and commodity prices.
From 3 data entry companies quoting $500 to $1400 down to $5. Batch image naming, visual verification, and the real handwriting error rate.
1 million tokens for ¥49. Is it worth it? Real usage numbers, where it beats DeepSeek, and where the plan actively hurts you.
A paid OCR service quoted me $180 for 40 pages. I used two AI APIs for $8 total. Here is what the service got wrong that AI got right.
DeepSeek does not have a vision endpoint. I built one that hooks into a separate vision API. The full wiring and what it costs.
Same 40-page catalog: all Claude = ¥28. Claude as manager, DeepSeek as worker = ¥2. The exact prompt and pattern.
Turning a 500-page PDF catalog into a structured product database with AI. Real costs, real limitations, real results.
How I used AI tools after work to help an export company launch an independent website, set up basic SEO, and get indexed by Google.
41 clicks, 661 impressions, 6.2% CTR. The numbers are small — that is why I am publishing them. A real baseline for a 15,000-page AI-built B2B site.
The exact setup, the working Python script AI wrote for me, and what broke during 500 pages of processing. No coding required.
5 prompt rules that stopped the JSON crashes, each born from a specific failure. Copy-paste templates included.
50-page controlled test. Claude is 20x more expensive but only marginally better on clean pages. The hybrid approach costs $3.
Standard OCR turns multi-column B2B supplier catalogs into garbage data. How AI vision reads pages like a human, and why it works for under $1.
A workflow for cropping thousands of product images from a PDF catalog, naming files by product code, and adding watermarks in batch.
The diagnostic method I was trained in — find the root cause, not the loudest symptom — is the same method that made the PDF pipeline actually work.
41 clicks, zero inquiries so far. Here is the funnel I built before the traffic arrived, and what I will change when the data proves me wrong.
My real tool split: Codex for execution, Claude with DeepSeek for low-cost document work, and DeepSeek for fast rough drafts.
Why AI feels like a tool moment for non-programmers, and how it helps thoughts become real pages, files, and verified work.
The other side of AI: it turns big ideas into real work, shows what already exists, and replaces blind confidence with a wider view.
HTML file verification, sitemap submission, old TCM redirects, and the filename mistake that slowed me down.
Use the updates page, RSS feed, or the contact form topic "Subscribe for updates." This is still intentionally lightweight.