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

The Client

An independent product idea: a LINE OA that acts as an outsourced memory for files — a problem the product's owner recognized everyone runs into, sending photos, slips, documents, and links into chat and then never finding them again.

The Challenge

The Solution

Users forward anything into the LINE OA — a screenshot, a PDF, a payment slip, a link — and AI reads, summarizes, and auto-tags it on arrival. Typing a plain search like "Kasikorn slip" or "photo from Osaka" returns the right file instantly, in Thai. Every account the system runs on stays in the product owner's name, so they can operate it independently going forward rather than depending on the developer long-term.

Why 3 Phases, Each One Proven Before Moving On

The project was deliberately staged into 3 phases — foundation, then AI and search, then an admin dashboard — rather than built all at once and handed over at the end. Each phase had to pass real testing from a real phone before the next one started, with nothing assumed to "probably work": phase 1 proved files sent into LINE were actually received and stored; phase 2 proved AI could actually read files and search actually found them from a real phone; phase 3 proved the product owner could genuinely operate the system without help, not just look at a demo. Every phase shipped with its own written test report — specific enough for the product owner to verify each item themselves from a phone in 5-15 minutes before signing off on the next phase. One rule was locked in from phase 1, not bolted on later: every file lives in a private space nobody else can reach, with no public link left open — download links expire within an hour, per data-protection requirements, from day one.

Key Technical Decisions

Making Thai Search Actually Accurate — 8 Rounds, Grounded in Real Usage

The most time-consuming part of the project wasn't the interface or the upload flow — it was making Thai-language search genuinely reliable rather than passing a shallow test. The product owner set the rule that shaped every one of the 8 hardening rounds: fix it by looking at real usage data, not by guessing and adding random synonyms. Every round started by pulling real user search logs first — what people actually typed, what the system actually answered — before writing a single line of code. Every fix that shipped had real data behind it, not just a hunch that it was probably fixed.

The Results

Every Phase Proven Before Moving Forward — Not Just "Shipped and Done"

The product owner didn't sign off based on the calendar — each phase had to be proven with evidence it actually worked. Phase 1 (foundation) passed all 8 of its planned test items; phase 2 (AI and search) passed all 7; phase 3 (admin dashboard and handover) passed all 10 planned items plus a full 9-test regression pass confirming nothing broke along the way.

Lessons Learned