All case studies
Confidential investment holding Hospitality division

Hospitality: a structured operations layer on top of WhatsApp

AI turns hotel chat traffic into a searchable operations database, request lifecycle, duplicate detection and manager reporting without changing how staff communicate.

AI implementation AI Acceleration Professional services

1,219

maintenance requests extracted from chat

792

duplicate messages identified (about 16% of the queue)

1.3 min

median time to request acknowledgement

From constraint to outcome

What was blocked, what changed, and what the client gained.

01

Problem

Hundreds of daily maintenance requests arrived in chat with no reliable way to link duplicates, track a request lifecycle or see recurring room issues.

02

System

A dedicated WhatsApp number reads existing work groups, classifies messages and media, links replies to requests, and exposes operational data through chat and scheduled reports.

03

Outcome

The hotel has an operating database and measurable response times without requiring a new app, staff retraining or a new process.

The challenge

The hospitality division handled 600–700 maintenance requests a day, with no way to detect duplicates. Accounting and HR work were also largely manual.

What we built

  • Structure without a new app. Staff continue using the WhatsApp groups they already know. A dedicated number in each work group reads messages and turns them into database records without changing how people write.
  • Request lifecycle and duplicate control. Messages are classified by operational type, room, department and severity; photos are described and voice messages transcribed. “Received” and “Done” replies create an opened → received → done lifecycle. Repeat messages about the same issue attach to the open request instead of creating a new dispatch.
  • Operational answers and reporting. Managers can ask questions in Russian in chat. The system queries the database for every figure and the model only phrases the answer. Daily and weekly reports cover requests, acknowledgement and completion times, oldest open work and unconfirmed items; urgent events are surfaced immediately.

What is already measured

  • 13,879 messages collected across 19 work groups, including every floor, engineering and housekeeping group.
  • 1,219 maintenance requests extracted and classified: 796 breakdowns, 298 preventive-maintenance jobs, 271 supply requests, 22 incidents and 90 urgent messages. 473 rooms now have their own history.
  • 3,368 photos read and described automatically.
  • 792 duplicate messages identified: about 16% of the queue. 111 rooms were flagged as problem rooms, with the highest at 13 requests.
  • Median time to acknowledgement is 1.3 minutes; median completion is 15.6 minutes. Staff kept their existing groups, so adoption required no new app or training.

The outcome

The division can see a clean request queue, recurrent room problems and real response times from the workflow it already uses. AI observes, structures, reports and answers; dispatch decisions remain with people.