AI Consulting in Thailand: A 90-Day Pilot
A 90-Day AI Pilot for Thailand
Hero artwork: AI-generated brand illustration using Parikshit Khanna’s likeness.
AI consulting in Thailand should bridge the distance between an exciting demonstration and a reliable business workflow. That requires more than a model subscription. It requires a process owner, authorised data, Thai-language evaluation, employee training and a clear decision at the end of the pilot.
A 90-day format is long enough to establish a baseline, test the workflow and correct weaknesses, but short enough to keep senior attention. It also gives procurement, technology, privacy and frontline teams one shared timetable.
Select a pilot with measurable value
Start with the process, not the tool. Map where time is lost, where information becomes inconsistent and where employees repeat low-value work. Suitable early candidates may include internal knowledge retrieval, first-draft summaries, sales-support material or a copilot that suggests—not sends—customer responses.
Define success before configuration. If the goal is faster service, measure existing response time, correction rate and customer escalation. If the goal is knowledge access, record search time and the rate at which staff find an approved source. Include the full cost of licences, integration, review and training when estimating return.
Select a use case whose errors can be detected and corrected. High-impact automated decisions deserve more extensive legal, ethical and technical assessment and are rarely a sensible first experiment for an organisation beginning its governance journey.
Convert Thailand’s PDPA into daily controls
Thailand’s Personal Data Protection Act B.E. 2562 (2019) regulates the collection, use and disclosure of personal data and provides rights to data subjects. For a pilot, the relevant question is not whether “the AI” stores data in the abstract. It is how information moves through prompts, files, retrieval indexes, logs, integrations and support systems.
The pilot team should document purpose and legal basis, minimise personal data, restrict access, define retention, assess security and vendors, and establish routes for requests and incidents. Sensitive information and cross-border transfers require careful attention. Employees should know which tools are approved and which information is prohibited.
Thai legal counsel and the organisation’s data protection officer should confirm the applicable requirements. An AI consultant can implement approved controls and evidence, but should not present a pilot as a legal compliance certificate.
Test Thai language as a product requirement
Thai evaluation cannot be reduced to translating an English prompt. The language’s script, lack of spaces between words, polite particles, formality, specialised terminology and transliterated foreign words can affect retrieval and generation. A system may sound fluent while selecting the wrong policy or softening a critical warning.
Create a Thai test set with representative questions, source documents and expected evidence. Include colloquial and formal inputs where relevant, spelling variants, dates, addresses and code-switching. Measure source accuracy, completeness, tone and the system’s ability to abstain or escalate.
Native or professionally qualified reviewers should examine customer-facing and high-impact outputs. Compare Thai and English results separately rather than averaging them into one reassuring score.
A commercial 90-day pilot plan
Stage | Days | Activity | Required output |
Diagnose | 1–15 | Interviews, current-state measures, use-case and data-risk scoring | Pilot charter, baseline, owner and approved scope |
Design | 16–30 | PDPA data map, vendor review, Thai evaluation plan, human-review design | Control register, test set and acceptance thresholds |
Build and prepare | 31–50 | Limited configuration, integrations, user guide and cohort training | Working pilot and trained users in a controlled environment |
Operate and test | 51–75 | Live-like scenarios, error review, security tests and feedback | Results across quality, risk, adoption, time and cost |
Decide | 76–90 | Remediation, benefits case and executive review | Stop, revise or scale decision with next-quarter plan |
Scope and price should be confirmed after discovery. The proposal must identify whether software licences, translation, development, penetration testing, legal review and travel are included or excluded.
Align with Thailand’s governance direction
Thailand’s Electronic Transactions Development Agency operates an AI Governance Clinic and publishes research and recommendations, including executive and generative AI governance guidance. These materials offer a useful local reference for oversight, risk and responsible application.
They should inform—not replace—the organisation’s own assessment. Governance becomes real when named people make decisions, controls appear in the workflow and evidence is reviewed. A policy that employees cannot apply under time pressure is not an effective control.
Practical tips for Thailand teams
Give the pilot one accountable business owner and one clear outcome.
Keep the first use case narrow enough to test deeply in 90 days.
Test Thai-language retrieval and output independently from English.
Use approved source documents and require citations where practical.
Train reviewers to identify confident but unsupported responses.
Log corrections, incidents and escalation time, not just usage volume.
Recheck the workflow when a model, vendor, data source or audience changes.
Fund expansion only after controls and benefits meet agreed thresholds.
Build internal ownership before scale
Executives need a decision dashboard; users need role-specific practice; reviewers need quality criteria; privacy and technology teams need data, security and incident procedures. Train these groups for their responsibilities instead of giving everyone the same generic session.
Parikshit Khanna is the founder of Digital Training Jet, an AI trainer and strategic consultant, and a TEDx speaker. Documented programmes are associated with Masters’ Union, CHRIST University, IIT Delhi and IIT Roorkee. This short biography is factual and does not imply institutional endorsement.
For a Thailand readiness audit, leadership workshop or 90-day pilot proposal, get in touch with Parikshit Khanna.
Frequently asked questions
Does Thailand’s PDPA apply to a short pilot?
Duration does not remove data-protection duties. If the pilot processes personal data, the organisation should assess the real data flow, purpose, legal basis, controls, vendor roles and transfers.
Why test Thai and English separately?
Performance can differ by language, source material and user style. Separate results reveal whether a good overall average is hiding poor Thai-language accuracy or escalation.
What should the pilot measure?
Use outcome and risk measures: cycle time, quality, correction rate, source accuracy, unsupported responses, adoption, incidents and total operating cost.
What happens after day 90?
Leadership should choose to stop, revise or scale. Scaling should have a budget, accountable owner, remediation list, monitoring plan and training schedule—not merely more licences.
Sources
Regulatory information checked on 11 October 2026. This article is general information, not legal advice.
