How to Choose an AI Consulting Partner in Thailand
Eight criteria, the questions to ask and the red flags to watch for when you compare AI consulting firms in Thailand.
Choosing an AI consulting partner in Thailand comes down to evidence. Look for a team that can show how it works, test on your own Thai data, explain who does the work, handle PDPA and hosting questions plainly, and stay accountable after launch. Slide decks and demo videos are the weakest signals.
This guide is for executives, procurement and IT leads comparing two to five firms. It gives you evaluation criteria, questions to ask, a way to compare answers and the red flags that usually predict a poor result. It focuses on capability and accountability, which you should settle before any commercial discussion.
What an AI consulting partner should do
Good AI consulting covers more than building a model. At a minimum, a partner should help you do five things:
- Choose use cases worth pursuing, and say no to the rest.
- Check that your data, systems and people can support them.
- Build and test a pilot on your real data.
- Set governance, including PDPA and security.
- Move the result into daily operations and keep it working.
Some firms do only strategy, and some do only engineering. Either can be right, but you should know which you are hiring and who connects the two. A strategy report without engineering often stalls. Engineering without strategy often builds the wrong thing.
Eight criteria for choosing an AI consulting partner
Use these as a scorecard. Rate each partner from one to five on every row, and write down the evidence behind the score.
| Criterion | What good looks like | How to check |
|---|---|---|
| Thai-language capability | Tests on real Thai input, including informal and mixed Thai-English text | Ask for a live test on your documents or chats |
| Relevant experience | Can describe similar problems and what went wrong | Ask for references you can call |
| Delivery team | Names the people who will do the work and their roles | Meet them before signing |
| Method | A defined path from discovery to pilot to production | Ask for a sample plan with milestones |
| Governance | Clear approach to PDPA, access control, logging and human review | Ask for their standard governance checklist |
| Hosting options | Cloud, in-country or on-premise, with trade-offs explained | Ask where data would be processed and stored |
| Handover | Documentation, training and a named support path | Ask what you own at the end |
| Honesty | Willing to say what AI cannot do for you | Notice whether they challenge your idea |
The last row is the hardest to fake. A partner who agrees with every idea you have is selling, not advising.
Questions to ask in the first meeting
Ask open questions and listen for specifics. Vague answers are information.
On your problem
- What would you want to know about our process before proposing anything?
- Which of our ideas would you not do first, and why?
On evidence
- Can you test your approach on a sample of our Thai documents or conversations before we sign?
- How do you measure accuracy, and who decides what counts as correct?
- Tell us about a project that did not go as planned. What changed?
On people
- Who will be on our project, and what share of their time is it?
- Is any work done by a subcontractor or technology partner, and where is it done?
On data and compliance
- Which personal data will your team see, and how is access controlled?
- How do you handle the PDPA, and who is responsible for what?
- Where will our data be processed and stored during the pilot and afterwards?
On the future
- Who owns the code, models, prompts and documentation?
- If we end the relationship, can we run the system without you?
- Who monitors quality after launch?
What to look for in the answers
Specifics over adjectives. For example, "We tested on a sample of invoices and found the model failed on stamped totals, so we added a review step" is a real answer, and the sample size should be stated. "Our solution is robust and scalable" is not.
A willingness to scope small. A good partner proposes a narrow pilot with a clear test, not a company-wide transformation.
Clarity about subcontracting. Many Thai consultancies work with technology partners abroad. That is not a problem by itself. It becomes one when you do not know who touches your data or who is accountable for the result. At UPSPIRE, for example, we say plainly that we are a Bangkok company working with our technology partner Groovy Technoweb Pvt Ltd in India. Any firm you evaluate should be equally direct.
Realistic handling of Thai. Modern models handle Thai well on many tasks, but quality varies by task and document. Be cautious with anyone who promises perfect Thai handwriting recognition or flawless results without testing.
Respect for your regulators. Ask which rules the partner has worked under. Banking, healthcare and government projects bring extra requirements beyond the PDPA, such as sector regulator guidance and internal security policies. Check current guidance from the PDPC and your sector regulator, and confirm your position with counsel.
Red flags
Any one of these deserves a follow-up question. Several together are a reason to walk away.
- A fixed answer before discovery. The partner proposes a solution before understanding your process.
- No test on your data. Everything is shown on their sample data only.
- Accuracy claims with no method. A single percentage with no explanation of how it was measured.
- A black box. They will not explain how it works, what data it uses or how it fails.
- No human fallback. The design assumes the AI is always right.
- Vague team. You meet senior staff in the pitch and never see them again.
- Pressure to decide. Deadlines that exist only to close the deal.
- Lock-in by design. Proprietary formats, no documentation, no export of your data or prompts.
- Silence on PDPA. Compliance is described as "handled" without detail.
- Everything is AI. Some problems are better solved with simple rules or a better form. A partner who never says so is suspect.
Run a fair comparison
A simple process keeps the decision grounded.
- Write a one-page brief. Describe the problem, the data you have, the systems involved and what success means.
- Send the same brief to every firm. This makes answers comparable.
- Request a small test on a sample of your data, under an agreement that protects it. Mask personal data where you can.
- Score with a mixed panel. Include IT, the business owner and legal.
- Call references. Ask what went wrong and how the partner responded.
- Compare plans, not slides. The best proposal has the clearest milestones and the clearest exit points.
If you want a reference point for what a pilot can look like, our working demos use sample data to show how a legal assistant answers and cites the sections it relies on. They are demos, not client results, but they show the level of detail you can ask any partner to produce.
What can go wrong even with a good partner
Choosing well does not remove your own responsibilities.
- You do not provide a decision-maker. Projects stall when nobody on your side can approve data access or accept results.
- The scope drifts. New requests arrive weekly. Agree how changes are handled.
- Reviewers are unavailable. A pilot needs staff who check outputs. Without them, you cannot tell whether it works.
- Expectations differ. Write down in plain words what "done" means for the pilot.
If your own team is not ready, a readiness review may be the better first step. Our AI readiness checklist for Thai enterprises helps you find out.
How to start
Start with the one-page brief and a shortlist of no more than three firms. Ask each to run a small test on your data before you commit to a larger engagement.
If you would like us on that shortlist, our AI strategy and consulting practice is built around discovery, a ranked use-case portfolio and a pilot plan. You can also contact our team to talk through your brief, whether or not you choose us.
Frequently asked questions
What should I ask an AI consultant in the first meeting?
Ask who will do the work, whether they can test on your Thai data before you sign, how they measure accuracy, how they handle the PDPA, where data is hosted, and who owns the outputs after handover.
How do I know if an AI consulting firm really understands Thai?
Give them a sample of your own Thai documents or chat messages, including informal and mixed Thai-English text, and ask them to show results and failures. Polished demo data tells you little.
Should the partner be a local company?
Not necessarily. What matters is clarity about who does the work, where data is processed, and who is accountable. A local presence helps with language, regulation and support, but ask about any subcontractors or overseas teams.
What are the biggest red flags when hiring an AI consultant?
Proposing a solution before understanding your process, refusing to test on your data, quoting accuracy with no method, avoiding PDPA questions, and offering no human fallback or handover plan.
Do we need a consultant or can our developers do it?
If you already know the use case and have the skills, targeted support may be enough. A consultant is most useful for choosing use cases, setting governance and measuring results.
Ready to put this to work?
Turn this playbook into a working system
From a first agent to a national platform, the same team plans it, builds it and runs it with you.
