AI for Cambodian businesses should begin with one measurable work problem, not a large software purchase. This guide shows hotel managers, retailers, schools, NGOs and professional organisations how to select a suitable use case, protect information and measure whether AI improves work for local customers, foreign residents, tourists, international buyers, investors or donors. Written by: Renee
Start with a business task, not an AI trend
The most useful first question is not “Which AI tool should we buy?” It is “Which repeated task is slowing the team down, creating avoidable errors or delaying a customer response?”
Potential tasks include drafting a first version of a bilingual enquiry response, summarising meeting notes, organising donor documents, checking a product catalogue for missing information, preparing interview questions or turning a long policy document into an internal checklist.
These tasks have different levels of risk. A draft social media caption is usually easier to review than a response about medical services, a safeguarding issue, immigration, finance or a grant commitment. Start with work where a trained member of staff can check the output before it reaches another person.
What international evidence suggests, and what it does not prove in Cambodia
A useful starting point is the State of AI in Sales report, published by HubSpot and based on responses from more than 600 sales professionals across business-to-business and business-to-consumer teams. It reports that 43% of sales professionals surveyed use AI at work, while 74% of marketing professionals surveyed use it. The report also records several practical uses: 36% use AI for forecasting, lead scoring or pipeline analysis, 22% for lead qualification, and 20% for prospect outreach.
These figures show which types of work organisations elsewhere have investigated. They do not establish the rate of AI adoption in Cambodia, or prove that a particular tool will improve results for a Cambodian organisation. The sensible response is to treat them as ideas for controlled local testing, then use your own enquiry, sales, service or programme data to make a decision.
Which Cambodian organisations should investigate AI first?
AI is most worth investigating where staff handle a high volume of similar information, repeat the same process regularly and can review the result before it is used.
| Organisation | Suitable first test | Measure |
|---|---|---|
| Hotels, guesthouses and tour operators | Draft replies to common booking or itinerary questions in Khmer and English | Response time, booking enquiries, correction rate and guest satisfaction |
| Retailers and distributors | Clean product descriptions and prepare customer service reply drafts | Time per listing, catalogue errors, response time and completed orders |
| Schools and training providers | Create first drafts of lesson outlines, parent updates or course FAQs | Preparation time, teacher edits, enrolment enquiries and factual errors |
| NGOs and development organisations | Summarise approved documents and create internal evidence checklists | Search time, missing references, review time and reporting corrections |
| Professional service firms | Classify enquiries and prepare meeting summaries for staff review | Follow-up time, qualified enquiries, missed actions and conversion rate |
This is a prioritisation framework, not a claim that any sector will automatically benefit. A small organisation with low enquiry volume may gain more from improving its enquiry form or filing system than from introducing AI. A larger team with repetitive administration may have a clearer case for a trial.
Adapt the test for each audience in Cambodia
Local Cambodian customers
Do not assume that an English-first workflow can simply be translated. Check Khmer spelling, names, prices, dates, locations, units and the level of politeness expected by your customers. Keep a human review step for messages involving complaints, refunds, health, education or sensitive personal information.
Measure whether the revised process produces fewer clarification questions and faster completed enquiries. Ask customers a short question after the interaction, such as whether the answer was clear and whether they knew what to do next.
Foreign residents and tourists
Foreign residents may need detailed practical information, while tourists often need concise answers about availability, location, transport, opening times, cancellation terms or what is included. Separate these audiences in your records rather than asking one generic system to answer everyone.
Use English-language drafts only when staff can verify the details. A polished answer that contains an incorrect price, location or booking condition can create more work and damage confidence.
International buyers, investors and donors
For international buyers, investors and donors, AI can help locate information in approved documents or produce a first draft of a summary. It should not invent financial figures, impact results, procurement claims, legal commitments or monitoring evidence.
Require every external document to be checked against the underlying source. Keep a simple record of who approved the final version, which documents were used and when the information was last checked.
Khmer-language and English-language audiences
Decide whether the workflow needs two separate language versions or a single bilingual document. Review both versions for meaning, not only grammar. A translation can be linguistically correct while still being unsuitable for a local customer, parent, guest or community partner.
Choose a low-risk pilot with a clear baseline
Before testing, record the current process for at least one working week, or long enough to capture a normal cycle. Use a simple baseline:
- Number of enquiries, documents or cases handled.
- Average time spent on each task.
- Number of corrections, escalations or missed follow-ups.
- Response time from enquiry to first useful reply.
- Commercial or programme outcome, such as a completed booking, sale, enrolment, meeting or approved report.
Then define the pilot in one sentence: “For four weeks, two staff members will use a reviewed AI draft for English and Khmer booking enquiries, with the target of reducing drafting time without increasing factual corrections.” This is more useful than a general target such as “use AI more often”.
Keep a control process where possible. One team member can continue with the existing method for selected enquiries while another uses the new workflow. The comparison does not need to be perfect, but it should use similar enquiry types and the same measurement period.
Set rules before staff put information into a tool
Every organisation should decide what staff may and may not enter. The rules should cover customer details, employee information, donor records, beneficiary information, contracts, financial data, unpublished research and confidential partner material.
A basic internal policy should answer five questions:
- Which tools are approved?
- What information must never be entered?
- Who checks AI-generated text, translations or summaries?
- How are errors reported and corrected?
- How are prompts, outputs and final documents stored?
Do not use an AI output as evidence merely because it sounds confident. Ask the system to work only from documents the organisation has supplied, where the tool supports that method, and make staff verify names, numbers, dates, quotations and commitments.
Measure the result that matters to the organisation
Time saved is useful, but it is not the only outcome. A faster process can still be harmful if staff spend more time correcting errors or if customers receive less helpful replies.
Use a small scorecard with one operational measure, one quality measure and one outcome measure:
| Measure type | Example question | Useful evidence |
|---|---|---|
| Operational | Did the task take less time? | Minutes per enquiry, document or case |
| Quality | Was the final output accurate and appropriate? | Corrections, escalations and reviewer scores |
| Audience | Did the recipient understand the next step? | Clarification requests, complaints and short feedback |
| Commercial or programme | Did the process support the intended result? | Bookings, sales, enrolments, meetings, approvals or donor submissions |
| Risk | Did the trial create a privacy or compliance issue? | Incidents, access records and policy breaches |
The HubSpot report says 98% of surveyed sales professionals make at least some edits to AI-generated text. That finding supports a practical rule for Cambodian teams: treat AI as a drafting or analysis assistant, not as an unsupervised author. The person responsible for the customer, project or document remains responsible for checking the result.
When to stop, improve or expand the pilot
After the agreed test period, compare the baseline with the pilot. Continue only if the workflow produces a worthwhile improvement without creating unacceptable quality, privacy or reputational risks.
- Stop if the tool introduces serious factual, privacy or safeguarding problems.
- Improve if staff save time but need better source documents, prompts, training or review rules.
- Expand if the result is repeatable, staff understand the controls and the benefit is larger than the ongoing cost.
For a hotel, expansion might mean applying a tested reply workflow to several properties. For a retailer, it might mean improving a larger catalogue after checking a sample. For an NGO, it might mean extending a document-search process only after confirming that citations and source versions remain reliable.
A practical 30-day action plan for Cambodian organisations
- Days 1 to 5: list repetitive tasks and select one with low risk and a clear owner.
- Days 6 to 10: record the baseline, define the audience, prepare approved source material and set review rules.
- Days 11 to 20: run the pilot with a small number of trained staff. Record time, corrections and outcomes.
- Days 21 to 25: compare the pilot with the existing process and ask staff or recipients for focused feedback.
- Days 26 to 30: decide whether to stop, revise or expand. Document the decision and update the internal policy.
The strongest case for AI is not that it is new. It is that a specific Cambodian organisation can show, using its own records, that a controlled workflow saves time or improves service without lowering accuracy or accountability.
