AI can speed up the slow, text-heavy parts of market research: summarising documents, sorting open-ended survey answers, drafting interview guides and structuring findings. It cannot be trusted as a source of facts, and it cannot replace talking to real customers. The safe pattern is to use AI to process material you supply and to check every claim that could change a decision.
Where AI genuinely helps
| Task | How AI helps | What you still do |
|---|---|---|
| Desk research | Summarises long reports you provide and pulls out themes | Read the originals for any figure you will quote |
| Survey comments | Groups free-text answers into themes | Spot-check themes against raw responses |
| Interview prep | Drafts question lists and screening criteria | Edit for bias and relevance |
| Interview notes | Cleans transcripts and tags topics | Review quotes in context |
| Competitor scans | Organises public information you have collected | Verify against the competitor’s own pages |
| Reporting | Drafts structure and plain-language summaries | Confirm every number and conclusion |
In each case the AI works on material you have gathered, which keeps the source clear and makes checking possible.
The main risk: confident but wrong output
Generative AI can produce fluent text that is simply incorrect. NIST’s publication on generative AI risk, AI 600-1, is a companion profile to its AI Risk Management Framework, aimed at organisations addressing trustworthiness in the design, development, use and evaluation of AI. NIST describes the framework as voluntary, and notes that the Generative AI Profile can help identify the unique risks posed by generative AI.
For research, the practical consequence is simple: treat any statistic, quote, market size or named source an AI gives you as unverified until you have found it in the original document. Invented citations are a known failure mode of language models, which is why this rule matters.
A safer workflow
- Define the decision first. Write down what you need to decide and what evidence would change your mind. This prevents collecting interesting but useless material.
- Collect sources yourself. Gather official statistics, industry reports, your own sales data and customer feedback. Prefer primary sources such as the Australian Bureau of Statistics, regulators and platform documentation.
- Give the AI only what it needs. Paste or upload the specific material and ask it to summarise or categorise that material, not to recall facts from memory.
- Ask for traceability. Request that each point cites the passage it came from, then check a sample.
- Verify claims that matter. Any number you will present to a client, board or customer should be traced to the original page.
- Keep a human decision point. A named person reviews the findings and decides what to do.7. Record what you did. Note the tool, date and prompt so the work can be repeated and audited.
Privacy and customer data
Be careful with what you paste into a tool. Customer lists, interview recordings and survey responses can contain personal information, and Australian businesses may have obligations under the Privacy Act and the Australian Privacy Principles. Check the tool’s data handling terms, remove or anonymise identifying details where you can, and seek professional advice if you are unsure. This is general information, not legal advice. For wider automation questions, see how AI automation can improve business operations.
What AI cannot do
- Replace real customers. Synthetic personas and simulated interviews reflect patterns in text, not your actual market. Use them to brainstorm hypotheses at most, then test with real people.
- Know your local market. General models may be thin on small local markets, new products and recent changes.
- Supply original evidence. Only surveys, interviews, sales data and experiments create new evidence. See how to measure creative campaign performance beyond views for how to test rather than assume.
Connecting research to strategy
Research is only useful if it changes a plan. Use findings to sharpen audience, positioning and channel choices, as described in building a digital marketing strategy for a growing business. Keep the research small and decision-led rather than chasing a complete picture.
Where Radiance Impact fits
Radiance Impact combines creative production, AI, digital technology and marketing, and approaches AI as a tool that needs checking rather than a source of answers. To learn more, see /about or get in touch through /contact.
Sources
- NIST AI 600-1: Artificial Intelligence Risk Management Framework, Generative AI Profile, National Institute of Standards and Technology
- AI Risk Management Framework, National Institute of Standards and Technology
