CasesReal-estate operating platform · mid- to long-term housing

Scattered housing operations, built into one platform — with arrival-date checking on top

Bookings, documents, conversations and settlement were scattered across channels. We built mid- to long-term housing operations into a single platform, and on top of it the system reads conversations every day to find bookings whose real arrival date differs from the registered one. A person still applies them.

4
Interfaces built · guest web · guest app · back office · operator app
0→1
Scattered channels → one platform
Where operations went · scattered channels → one platform → checking on top
BEFORE
Scattered ops
  • Conversations scattered per channel
  • Contracts, letters, receipts by hand
  • Real arrival dates only inside chats
A manager had to read every conversation, every day
BUILD
Into one platform
  • Bookings · rooms · payments
  • Conversations · document issuing
  • Guest web & app + back office & operator app
Operations become machine-readable
RUN
Checking on top
  • Reads only new conversations daily
  • Proposes diverging arrival dates, with grounds
  • Escalates imminent cases in stages
No automatic application — a person must approve
What we built, as an object model — users, operations, and the checking on top (drag / zoom)
ontology graph loading…
Challenge · What was stuck

The registered check-in date was derived from the school term — it was not the real arrival date. The real arrival information lived only inside chats held in several languages, so a manager had to read every conversation, every day. Contracts, proof-of-residence letters and receipts were produced by hand, and bookings arriving through outside platforms — along with pre-existing customers — sat outside the system entirely.

Solution · What we built
  1. Built the whole operation into one platform — guest web (multilingual), guest app, operator back office, operator app. Bookings, rooms, payments, conversations and documents sit on one spine.
  2. Turned documents into an issuing flow. Contracts, proof-of-residence letters and receipts come out filled in and go straight into the conversation.
  3. Gathered scattered customers into one thread — bookings from outside channels are absorbed and answered on the same screen, and pre-existing customers join the same ledger. Messages are auto-translated into the other party's language.
  4. Put arrival-date checking on top. Each morning it reads only the new conversations on upcoming bookings, extracts the real arrival date, and where it differs from the registered value, raises a proposal quoting the conversation that grounds it. Imminent cases escalate in stages. There is no automatic application — a person must approve, and the original date stays untouched while the adjustment sits on top.
Impact · What changed
  • A full sweep surfaced bookings where the registered date and the real arrival diverged — some by weeks.
  • A manager reading every conversation daily became a manager looking at a few items awaiting approval.
  • It runs live, and the operator app is shipped to the stores.
  • The discipline is the same here — the machine proposes, a person applies.

* A real engagement, anonymized with client consent. We don't publish figures about a client's operating scale.

Apply / 적용

The domain was housing, but the prescription is domain-agnostic. If your operations are scattered across chats, sheets and paper — with 'nowhere for AI to enter' —

Common situation in your companyWhat this case prescribes
Orders, bookings and support scattered across chat, calls and sheetsOperationalize into one platform first — machine-readable
When a key person leaves, the knowledge leaves with themAccrue pricing, norms and history as the company's work language → the company owns it
Revenue only grows if headcount growsMachine handles what it can, people only the rest — a non-linear shift
You 'adopted AI' but there's no data for it to enterMake operations readable first, then put it on top
Get Started

Where is one flow stuck in your company?

In 30 minutes we can find where time leaks out, together.

Book a 30-min diagnostic