Building With AI: What It Can and Cannot Do for Your Venture

By Kasper-Tanakrit Sermsuksan · Global Venture I: Playbook · Block E · Facts checked 12 Nov 2026

You are here: E-07 · Block E · Understanding the Opportunities · article 7 of 7 · Form E · Next: F-01 Startup Funding Types

In short: AI is the fastest research assistant, drafting partner and prototyping tool a founder has ever had — but it cannot talk to your customers, give you evidence or make your judgment calls. Use it to learn a field in an afternoon, draft faster and build a testable MVP in days. Check everything it tells you, because it makes things up and can leak what you paste into it. I recommend three simple rules: use AI for research and drafting, write your own reflections and interview notes, and keep a note of where you used it.

Try it in 10 minutes: Here is a test I like. Ask an AI tool for three facts about your market, with sources. Open each source yourself. Mark each fact "held up", "wrong" or "could not find". Most people are surprised by the last column — and that column is why this article exists.

Here is a number that should change how you work this year. About 243 million people in Southeast Asia now use generative AI — around 58% of the working population¹.

Your peers use it. Your competitors use it.

Using it is no longer an advantage. It is the floor.

Remember the bathroom break from Market Research for Startups? A candidate once told me how she excused herself from an investor meeting, locked the restroom door and spent three minutes Googling. She walked back in ready.²

Today those three minutes would take thirty seconds in an AI chat.

The tool got faster. The skill did not change: know what you do not know, check it fast, come back ready to contribute. I wrote this to help you keep the skill while you use the tool.

How can startups use AI?

In three main ways:

• Research: learn an unfamiliar field, its terms and its rules before you walk into an interview.
• Drafting: get to a first version of an email, a UVP, a landing page or a memo section faster.
• Prototyping: build mockups, clickable screens and simple apps for an MVP test in days, not months.

Here is how I would use each one.

1. Research: learn a field fast

• Explain an unfamiliar industry, its terms and its business models before an interview.
• Summarize long reports and policies — then open the original for anything you will quote.
• Draft a list of possible competitors, regulations and stakeholders for you to check.
• Prepare better interview questions: "What would a nurse in a Thai community hospital find frustrating about…?" gives you hypotheses to test, not answers.

2. Drafting: get to a first version faster

• First drafts of outreach emails, a UVP, a landing page, a pitch outline or sections of a 6-page memo.
• Rewriting for clarity, translating between Thai and English, and checking tone for a partner or a funder.
• Playing the critic: "Here is my business model. Give me the five strongest objections an investor would raise."

3. Prototyping: make it tangible

• Mockups, product images, short videos and clickable screens for an MVP test.
• Simple working apps, chatbots, forms and landing pages, even if you are not a strong coder.
• Many versions of the same idea in one evening, so you can test which one customers react to.

People believe what they can see. AI lets you show them something real in days instead of months.

That is a genuine gift. Use it.

Why does every founder now need to know how to build with AI?

Because the money and the attention have already moved. The same e-Conomy SEA report says AI-focused companies took about 80% of private funding in the region in the first half of 2026, up from 32% a year earlier¹.

I take two lessons from this. First, using AI is the baseline — like using a search engine. Second, "we use AI" is not a value proposition. It is a feature everyone can claim.

I make that point in Industry Trend Analysis, and it matters even more now.

To me, the question is not whether you use AI. It is whether you use it to go faster toward real evidence — or to produce convincing-looking work with no evidence underneath.

What can AI not do for your venture?

It cannot talk to your customers

An AI can write a perfect "customer persona" in ten seconds. It is not evidence. It is a guess, built from what other people wrote on the internet — written confidently enough to feel true.

Your customers' real words, their past behavior and their commitments are evidence. Nothing else is.

At the end of one course, a student told me that the assignments that pushed them out of their comfort zone to talk to real people felt uncomfortable at first, and turned into one of the most valuable parts of the semester. No tool does that part for you.

It cannot make your judgment calls

AI can list ten options. It cannot decide which risk your team is willing to take, which customer you choose to serve first, or when to stop.

That is judgment. It comes from evidence, values and experience — yours.

It cannot build trust for you

Trust is still built person to person. Relationships, reputation and warm introductions — your Social Capital — are still built person to person. An AI-written thank-you letter that sounds like every other AI-written thank-you letter does the opposite of what you want.

It cannot make your work different

If everyone asks the same tool for a value proposition, everyone gets a similar value proposition.

Your edge is what only your team knows: what you heard in your interviews, what you saw on site, what your field and your cross-discipline friends taught you.

What are the risks of using AI as a founder?

Five risks I would keep in mind:

• Hallucination. AI tools can invent facts, numbers, quotes, studies and links that look completely real. Never put a number, a source or a quote into your deck, memo or form unless you have opened the original source yourself.
• Privacy and confidentiality. What you paste into a public AI tool may be stored or used to train it, depending on the tool and its settings. Do not paste interview notes with names, a founder's confidential numbers, or a partner's documents without permission. Remove names first.
• Out-of-date answers. Many tools do not know about the latest programs, rules or prices. Grant calls, trade agreements and regulations change every year — check the official page.
• Bias. These tools know some places and languages much better than others. In my experience, answers about Thai provinces, ASEAN consumers or local rules are often weaker than they sound.
• Skill loss. If AI writes your reflection, you skip the thinking the reflection was for. You lose the learning and keep only the output.

What rules should a founder set for using AI?

I recommend three lines. They are the same ones I ask of my own students.

• Use it freely for research and drafting.
• Write the thinking yourself: your reflections, your interview notes and your letters — thank-you letters, letters to your future self — in your own words.
• Say where you used it. Keep one line on what AI did, every time.

Why these lines? Because reflections and interview notes are the evidence of your learning. Everything else is a tool for getting there faster.

✦ If you are doing this as part of Global Venture I: Playbook

These three lines are the course AI policy. The full rule sits in About Global Venture I: Playbook. Your Venture update, interview notes and letters stay in your own words, and every form has an "AI used for:" field. Fill it in every time.

How to do it: build with AI, the right way

1. Use AI to prepare, not to replace. Before each interview, ask AI to explain the field and draft questions. Then go and talk to the real person.
2. Build the MVP fast. Use AI tools to make your prototype or landing page. Spend the time you save on putting it in front of more people.
3. Check every fact. For every number or source AI gives you, open the original. No source opened, no fact used.
4. Protect people. Remove names and confidential details before pasting anything from your interviews.
5. Write the thinking yourself. Draft reflections and interview notes by hand or in your own words first. Use AI only to fix grammar if you need it — and say so.
6. Log it. Keep one line on what AI did (for example, "AI used for: explaining agricultural terms before interview 2; first draft of landing-page copy").

Reality check

It is very tempting to let AI do the uncomfortable parts — write the reflection, invent the customer insight, polish the thank-you letter until it sounds like everyone else's. It is also tempting to trust an answer because it sounds confident.

In my experience, most students who get into trouble with AI did not cheat on purpose. They just stopped checking.

Other people may notice. More importantly, you will notice — later, when you realize you cannot explain your own venture without your notes.

If you are struggling at this stage, you are like most people. It means you are doing the real work.

Frequently asked questions

Can AI replace customer interviews?
No, and I would not let it. AI can help you prepare questions and understand a field. Only real people give you evidence about real problems.

How should founders use AI to build an MVP?
Use it to make the test faster — a landing page, a mockup, product images, a simple chatbot — then put it in front of real people with a pass line decided in advance. What to test first is covered in MVP & Prototype.

Can I use ChatGPT, Gemini or Claude to work on my venture?
Yes, for research and drafting. I recommend writing your own reflections and interview notes, and keeping a note of where you used AI.

How do I check if AI made something up?
My rule: open the original source for every fact, number, quote and link. If you cannot find it, do not use it.

Is it safe to paste my interview notes into an AI tool?
Only after removing names and confidential details, and only if the person agreed to how their information would be used.

✦ A Note for Global Venture I Students · Block E: Understanding the Opportunities · Form E

Block E is where you build your MVP, so it is where I expect AI to help most — and where it is easiest to fool yourself. Use it to go faster toward real people, not to avoid them.

Can I use ChatGPT, Gemini or Claude in Global Venture I? Yes, for research and drafting. Your Venture update, interview notes and letters stay in your own words, and you fill in the "AI used for:" field on every form. If you skip the checking, your instructor may notice.

Do this before class:

1. Use an AI tool to build or improve one part of your MVP this week. In Form E, write what you asked, what it gave you, and what you changed by hand.
2. Take three facts AI gave you about your market. Open the original source for each. Note which ones held up, which did not, and which you could not find.
3. Venture update (your own, in Form E; your team keeps one Venture Log): tool applied · what changed in the idea · evidence link · readiness self-score · AI used for. Book Log: reread your Chapter 18 entry — what is your version of the three-minute bathroom break now?

Pre-class check (in Form E):

1. Name two things AI can do for your venture and two things it cannot.
2. Why is an AI-generated customer persona not evidence?
3. What are the three parts of the course AI policy?

Your progress: Customer readiness to level 3, with 4 as the stretch (faster, more tangible MVP tests with real customers) · Technology readiness: record only. How we measure progress

Based on the KTH Innovation Readiness Level™ by KTH Innovation, licensed CC BY-NC-SA 4.0 (kthinnovationreadinesslevel.com).

Skills you practice: Technological literacy · AI and big data · ethical and sustainable thinking · taking the initiative.

Keep reading

• Startup Funding Types — next, in Block F: how ventures get funded
• MVP & Prototype — what to build, and what to test first
• The Interview Masterclass — the part AI cannot do
• The Interview Log & Evidence Rules — what counts as evidence (and what does not)

All articles in Block E · Understanding the Opportunities

• E-01 Industry Trend Analysis
• E-02 Social Capital in ASEAN
• E-03 MVP & Prototype
• E-04 Unique Value Proposition
• E-05 Regional Frameworks
• E-06 The Cross-Discipline Friend Interview
• E-07 Building With AI (you are here)

Further reading

• Mollick, E. (2024). Co-intelligence: Living and working with AI. Portfolio.
• Agrawal, A., Gans, J., & Goldfarb, A. (2018). Prediction machines: The simple economics of artificial intelligence. Harvard Business Review Press.

References

Web pages accessed Nov 2026 (checked 12 Nov 2026).

1. Google, Temasek, & Bain & Company. (2026). e-Conomy SEA 2026: Building the intelligent economy. Google. https://economysea.withgoogle.com/
2. Sermsuksan, K.-T. (2023). The Startup Mindset [in Thai], Chapter 18. Wish Publishing (สำนักพิมพ์วิช). English: the author's own translation, not yet published.

About the author: Kasper-Tanakrit Sermsuksan is the Founder of SEA Bridge and a Visiting Lecturer in Computer Engineering and Digital Technology at Chulalongkorn University's Faculty of Engineering. Learn more →