AI Concierge for a Regional Resort Group — 91% Self-Service, 321 Leads in 5 Weeks

The Client

A one-person industrial valve, pump, and fitting distributor — B2B, no sales team, no back-office staff. The owner personally answered every one of roughly 50 daily LINE chats: pricing questions, quotation requests, payment slips, shipment tracking.

The Challenge

The Solution

A LINE OA chatbot was built to search the live catalog and answer routine questions instantly — pricing, availability, spec lookups across brands and standards — draft quotations and invoices as ready-to-send PDFs, read incoming product photos, verify payment slips, and track shipment numbers. The goal wasn't to replace the owner's judgment, just to take the repetitive share off his plate so he could focus on the conversations that actually need him.

From a Simpler Scope to What the Job Actually Needed

The deal closed against a simpler scope: a bot that answers basic questions and generates PDF quotations. The first working call surfaced three layers of complexity that weren't in the original agreement — linking brand, model, and spec-sheet files together, picking the correct sizing standard per brand (since some brands treat it as an option on one model and a completely different model elsewhere), and the owner's real expectation that customers get an actual spec file to open, not just a text description. Rather than stopping to rescope and rebuild from zero, the owner chose to keep testing the live bot himself, round after round, sending feedback continuously while the build adjusted incrementally — instead of waiting for a complete written spec before starting.

Key Technical Decisions

Built in 7 Days Against a 14-Day Plan

What made the bot sound like the actual owner rather than a generic script was training it on real chat history — 1,827 real conversations analyzed in detail, down to noticing the owner used one particular sentence-ending particle over another by a ratio of roughly 19 to 1, then pulling real past replies into 15 of the most common question categories (pricing, quotations, stock availability, shipping cost, payment) as the bot's standard answers.

The Results

Where the 90% Figure Actually Comes From

The 90% wasn't measured by an automated tracking system after go-live — the project closed as a self-hosted handover before there was accumulated usage data. It comes from two sources instead: first, an analysis of 1,827 real historical conversations (49,311 customer messages) found that 12 of the 15 most common question categories — about 90% of message volume — already had answers the bot covered before handover, with the remaining 3 categories (shipping status, spec requests, invoice name changes) closed with ready-made answers before go-live. Second, a prepared 31-question technical test suite passed 31/31, and the owner's own hands-on test on his phone passed 21/21 — covering direct model-name lookups, remembering repeat customers without re-asking questions, and keeping the discount ceiling confidential from customers.

Being Honest About What's Still Open

This project closed as a self-hosted handover — the owner chose to install and run the system himself rather than continue a monthly support contract, so there's no long-term usage data (like real monthly chat volume or measured cost savings after go-live). The 90% figure comes from the delivery-testing report, not 30-day-plus post-launch tracking.

Lessons Learned