AI Training for Sri Lankan Hotels and Tourism Teams

By Ammar Ahamed, Founder, Cocoon · 21 October 2026 · Part 6 of 8 in the Corporate AI in Sri Lanka series

A message at 2 am

A guest messages the hotel at two in the morning. Their flight has been delayed, they want to change their airport pickup, and their message is in a language nobody on the night shift reads well. The night manager does their best with a translation app and a lot of goodwill.

That is hospitality in Sri Lanka at its best and most stretched: warm people doing a great deal with very little support. AI does not replace the warmth. It removes some of the stretch, so the people who make guests feel welcome have more time to do it.

What the numbers say

Sri Lanka welcomed 2,362,521 tourists in 2025, the highest number on record, according to the Sri Lanka Tourism Development Authority, up from 2,053,465 in 2024. Tourism earned USD 3,219.2 million in 2025, by the Central Bank's figures.

2026 has been harder. Arrivals for January to September were 1,693,679, behind the 2025 pace. When growth slows, competition shifts to the things guests notice: how quickly you reply, how you handle a complaint, what your reviews say, and how good the experience feels for the price.

Tourism is also a major employer. SLTDA's 2023 statistical report put total tourism employment at around 430,000 people, about 205,000 of them in direct jobs. Many properties also know how hard it is to keep experienced staff, which makes every hour of their time more valuable.

Where AI helps a hotel team

What must stay human

Hospitality is the one industry where we would argue hardest against over-automating. A guest who wrote to you in a moment of stress should get a reply that reads like a person cared, because a person did.

So the rule for hotel teams is simple. AI drafts. People host. A tool can produce a first reply in seconds, but someone who knows the property, the guest and the situation reads it, changes it and sends it. Over time, the best teams build a small library of prompts that sound like their property, so the drafts get closer to their own voice.

The same goes for anything involving safety, health or allergies. Those answers come from people with the facts, every time.

Reviews: a quick, visible win

Review responses are a good place to start training a hotel team, because the results are public and measurable within weeks.

Two uses work well. The first is drafting a thoughtful reply to each review, especially the difficult ones, which staff often leave unanswered because they are hard to write. The second is reading a month of reviews at once and pulling out the patterns: which complaints keep coming back, which staff get named for good service, and which parts of the stay guests mention most.

Sales, groups and events

Sales teams in Sri Lankan hotels spend a lot of time on proposals for corporate events, weddings, conferences and tour operators. Much of each proposal is rewritten from the last one.

AI helps here by drafting a tailored proposal from a short brief, researching a company before a sales call using its public information, and turning meeting notes into a clear follow-up email. One rule matters: rates, availability and terms always come from your own systems and your own rate sheets, never from an AI's guess.

Guest data and the PDPA

Hotels hold some of the most sensitive personal data in any business: passport details, contact information, payment details, special requests that can reveal health or religion, and sometimes who a guest is travelling with.

None of that should go into an unapproved AI tool. Staff can get almost all the benefit by removing names and identifiers before using AI: "draft a reply to a guest whose pickup needs to move from 6 am to 9 am because of a delayed flight" works perfectly well without a name or a booking number.

With the main duties of Sri Lanka's Personal Data Protection Act due on 1 January 2027, a one-page guideline and approved tools should come before any training. We explain both in the PDPA and AI at work.

A training plan for a property or group

Brief the general manager and heads of department first, and agree two or three outcomes. In hotels these are often reply times, review scores and the time it takes to send a sales proposal.

Train front office, reservations and guest relations together, on real but anonymised messages and reviews.

Then sales and marketing, on real briefs and proposals with rates removed.

Then F&B, HR and finance, each on their own repeated work.

Put a champion at every property. For hotel groups with resorts in the south, the hill country or the east, a local champion who shares the team's language matters more than anything else. Our post on AI for teams who work in Sinhala and Tamil shows how to design for that.

Measure what guests feel. Track reply times and review responses before training and for a quarter after. Those numbers make the case for the next property. The full sequence is in our 90-day rollout plan, and our wider guide to AI for hospitality teams covers more use cases.

One thing to try this week: look at your property's unanswered reviews from the last month. Draft replies to three of them with AI, improve them yourself, and post them. Notice how long it took.

Frequently asked questions

Sources

1. SLTDA, Monthly Tourist Arrivals Reports 2026 (2025 and 2026 figures)
2. SLTDA, Monthly Tourist Arrivals Reports 2024
3. Central Bank of Sri Lanka, External Sector Performance, December 2025 (tourism earnings)
4. SLTDA, Annual Statistical Report 2023 (tourism employment)