AI in brand research: an accelerator, not an expert
AI shortens the rough survey phase, but positioning decisions must still be anchored in real human evidence.
AI excels at scanning the market, gathering competitors, and drafting many positioning hypotheses for the team to discuss. However, AI cannot replace real customer interviews and lacks the ability to read the cultural context of Vietnam to make final decisions. Sinh Vũ uses AI to quickly outline the framework, while the final positioning is still determined by humans based on verified data.
- Analyze competitors and quickly compile industry reports.
- Sketching multiple positioning hypotheses as discussion material.
- narrow research budget needs to move quickly on the framework
- Make the final positioning decision without real customers.
- need to read deep insights from real interviews
- forecasting future behavior based solely on aggregated data
AI is not a strategic expert, but it is a sufficiently good research assistant to significantly shorten the rough work. The issue is not whether to use it or not, but to use it correctly and understand its limitations. Sinh Vũ shares this perspective so you do not waste the tool or let it lead you astray.
What can AI do in the research phase?
In the preparation stage, AI can process large volumes of public documents in a short time. Specifically, it excels at the following tasks:
- Review competitors, gathering public information about their positioning, messaging, and distribution channels.
- Summarize industry reports and expert articles so the team doesn't have to read everything from the beginning.
- Draft 5 to 7 different positioning hypotheses as material for discussion, rather than starting from a blank page.
- Establishing a comparison framework (table, matrix) to facilitate a more structured discussion from the start.
All of these tasks belong to the raw phase, the preparation of materials. AI moves faster than humans here, and that is its true value.
AI cannot replace any part
The boundaries are clearer than many think. There are three parts that AI cannot replace in brand strategy research:
- Real customer interviews. AI can create simulated survey respondents (synthetic respondents, meaning virtual profiles based on aggregated data), but this tool only reflects observed behavior from the past. When you need to understand true motivations, concerns, or purchasing context, nothing replaces a direct interview.
- Reading the cultural context of Vietnam. Many brand decisions depend on local understandings, from how Vietnamese people read signals of credibility to how they distinguish between "local" brands and "foreign" brands. AI has not been trained deeply enough in this context to make accurate judgments.
- Final decision-making. Beautiful hypotheses from AI still require someone with sufficient judgment and responsibility to select, adjust, and finalize. This step cannot be delegated to tools.
Specific risks to note
Sinh Vũ observes that the following mistakes are repeated frequently during the research phase:
- Summary information as verified facts. AI synthesizes from multiple sources, but does not always cite accurately. A figure from a reputable report and a figure from an intermediary blog may be presented identically. Always trace back to the source before use.
- Assign data collected from blogs as research from a major company. This is the most common form of hallucination (data hallucination, meaning the model fabricates or misattributes sources) in current AI tools. No single technique can completely prevent this.
- Amplifying existing biases. If the training data of the model is skewed towards a certain customer group, the AI-generated customer profile will reflect that bias, not the actual market reality.
- Received many hypotheses but no one has the appetite to filter. AI can generate ten positioning hypotheses, but if the team lacks a neutral and experienced leader to distinguish feasible hypotheses from appealing ones, the result remains null.
Transparency in methodology is essential when using aggregated data: clearly state when and where simulated data is used so that readers understand the limitations of the conclusions.
STRAT7, B2B International on market research with synthetic data
When to use AI, when not to?
Use AI to quickly outline: Analyze competitors, gather industry reports, draft initial positioning hypotheses, and create comparison tables. Particularly useful when research budgets are limited and materials are needed before internal workshops.
No relying on AI for final decisions: Final positioning decisions, deep insights from real buyers, forecasting future market behavior, or judging local context. In these areas, at least a few interviews with real customers are needed to anchor conclusions.
The viewpoint of Sinh Vũ
In the consulting and workshop service (S5), Sinh Vũ uses AI to prepare a comparative framework and competitor summary before the meeting. However, the true value of that session lies elsewhere: in the neutral facilitator who keeps the discussion data-driven, in the interviews with the founder and key team members, and in the positioning decision made by your team in the room, based on verified evidence. These are things that AI cannot replace and should not replace.
The correct mindset about AI at this stage is: it is a research assistant that helps the team arrive at discussions with more materials and less rough preparation time. Nothing more, nothing less.
Decision checklist
Topic: AI in brand strategy and research. Sinh Vũ guide, sinhvu.com
Select each item you find appropriate, then print or save as PDF to take with you.
If you have marked most of the signs above, this is the time to discuss in more detail. Sinh Vũ can help you review and propose a direction.
References
STRAT7, Synthetic Data in Market Research. B2B International, AI in Market Research: Limitations of Synthetic Data. Academic survey on reducing hallucination in LLM (RAG, reasoning, agentic), arXiv. Practical perspective from Sinh Vũ Studio.
Frequently asked questions
Can AI fully replace customer interviews?
No. AI can simulate survey respondents (synthetic respondents), but reliability is only temporarily stable with current behavior, while future predictions quickly lose accuracy. Interviewing real customers remains an irreplaceable layer of evidence, especially when understanding motivations and the context of purchasing decisions.
Will using AI for market research reduce costs?
The raw part (gathering documents, creating competitor comparison tables, drafting hypotheses) can be expedited significantly, thereby reducing labor hours in the preparation stage. However, if you skip interviewing real customers to save time, the risk of making incorrect positioning decisions will be much more costly than the initial savings.
How can you tell if a metric provided by AI is reliable?
The simplest principle: trace back to the original source, do not stop at an AI summary. If AI cites a number but cannot provide the original report, or refers to an intermediary blog, that number cannot be used in strategic documents. Sinh Vũ always requires source verification before including any data in the positioning document.