Market research meets AI: The 3-step framework
From information overload to actionable insights in days, not weeks
Summary
Traditional market research is breaking under the weight of information overload. Drawing from a year of hands-on experimentation, we've developed a practical 3-step AI research framework that cuts research time by 50% while improving quality, using just prompts and AI tools. Our approach helps you move from drowning in data to surfacing genuine insights - no coding or technical expertise required! 🙌
The modern research challenge: We're all information hoarders
It's late in the evening, and your desktop is a battlefield of open browser tabs, fragmented notes, and half-processed information. That game-changing market insight you found yesterday? It's now buried somewhere between tabs 47 and 48. Whether you're a salesperson tracking competitors, a product manager sizing new opportunities, or an entrepreneur validating ideas - we're all drowning in information while racing against deadlines.
While the market research industry is valued at over $80B globally, producing accurate research has become increasingly challenging due to information overload, lack of credible sources and shortage of time.
Introducing our handcrafted AI research framework
Over the last year, while researching our multiple ‘next-big’ ideas, we have spent many sleepless nights. But thanks to the advancements in Generative AI, we have formulated a systematic and fast method to do research.
In this blog, we will talk about:
Our research flow that cuts down our research time by half
Which AI tools to use for the best results
Some intelligent prompting techniques to 10X your research
💡 Oh and let’s clear this upfront: No code, no AI agents, no fancy jargon at all. Just smart prompting and a combination of some AI tools.
Let’s dive in! 🏃🏻♀️➡️
Step 1: Using AI for secondary research
The first phase builds on traditional secondary research but supercharges it with AI. Rather than starting with generic Google searches or jumping straight to ChatGPT, begin with specialized tools designed for deep research, such as Perplexity.ai, Gigabrain or Stanford's Storm.
Here is a sample prompt that you can try right away in Perplexity for your next research project. Let’s look at an e-commerce example.
E-COMMERCE MARKET SIZING: SAMPLE PROMPT FOR PERPLEXITY
Analyze the Indian e-commerce market with a specific focus on:
1. Market size, and regional breakdowns (past 2 years)
2. CAGR projections for next 5 years
3. Market segmentation by:
- Product verticals
- User demographics
- Focus area (horizontal vs. vertical)
4. Key players, their GMV and market shares
5. Player-wise positioning, strengths and weaknesses
6. Emerging market opportunities and threats
Include only: Verified market reports, Investment analyses, Government/regulatory data, Industry association reports
Required: All data points must be from 2023-2024💡Pro tip: To get super credible sources, add advanced search strings at the end of your prompt. Example: filetype:pdf "your topic" "2023" "market size" site:.edu OR site:.gov
We recommend you save all collected data in a folder or a Google drive, including industry reports, Reddit threads, Youtube podcast links, etc.
Once your data is organized, you’d need to choose the right tool for synthesis:
For a quick analysis with 5-6 documents - Use Claude Projects or Custom GPTs
Upload reports in your Claude project, connect your output Google doc with an outline in place (just section headings are ok), and let Claude summarize the reports one by one for you!
For deep analysis of multiple reports, Youtube video links, etc. - NotebookLM
Use NotebookLM to analyze multiple reports and create summaries, FAQs, and cross-document insights
💡 Pro tip: Save any useful web pages as PDFs, upload to the Claude Project and let Claude summarize them for you.
Step 2: Using AI for primary research
Alright, so the fun begins now. Remember that part where you request users for in-depth interviews or to fill your surveys? Now, you can simulate that right in your browser! Of course, that doesn’t undermine the importance of real interviews, but it sparks imagination and gives you a great starting point.
Creating user personas with AI
Let’s go back to our Claude Project. Start a new chat within the project and generate detailed customer personas with AI. Here’s a sample prompt:
GENERATING USER PERSONAS WITH CLAUDE: SAMPLE PROMPT
You are a senior market research manager in Indian e-commerce. First, create the most prominent 2-3 user personas including:
- Demographics & psychographics: Age, income, values, tech habits
- Daily routine: Schedule, pain points, decision-making triggers, trusted sources
- Product usage: Current solutions, brand loyalty, price sensitivity,
preferences
- Buying process: Research, influences, criteria, post-purchase behavior
Next, for each persona, include key objections, essential features, deal-breakers, and brand perception.
Confirm with me at every step before moving forward.Expanding your research
With your personas ready, you can now:
Run mock customer interviews by having AI roleplay each persona
Generate interview questionnaires with AI for real user research
Analyze interview transcripts and survey responses with AI
💡 Pro tip: Always end your AI interactions with "What else should I know that I haven't asked?" - you'll be surprised by the insights!
At every step, remember to ask Claude to summarize the insights and manually copy and paste them to your output Google document in the Claude project.
Step 3: Summarizing the insights
By now, you’d have the following documents in your Claude project:
(1) Industry reports from Step 1
(2) Your user personas and mock chats with users from Step 2, saved in a document by you in Project Knowledge
(3) Transcripts or results from primary interviews and surveys from Step 2, saved in Project Knowledge
(4) Your output from each of the above in your output Google document, again linked to Project Knowledge
Next, you’d want to create a neat presentation that is shareable with others. No surprises - but AI can be super useful here too.😃
Your Claude Project already has all the intelligence now. So just ask it to create a visual presentation. Look at a sample prompt below for the same:
CREATING AN HTML/CSS PRESENTATION WITH CLAUDE: SAMPLE PROMPT
You are an expert product manager in Indian e-commerce. Based on all our research available in the Project Knowledge section, create a concise presentation to present to the company leadership, focused on:
1. Market Size & Growth over the next 5 years
2. Competition Landscape: Key players, market shares, positioning, strengths, weaknesses, etc.
3. Target Customer Analysis: Buying behaviour, preferences, etc.4. Product Opportunity
5. Go-to-Market Strategy
6. Key Risks & Mitigation
Format: Maximum 10 slides, data-driven with clear visualizations for stakeholder review. Provide the output in HTML/CSS format. Follow a white and dark blue color scheme.
First, share the contents of the slides in bullet points for me to approve. Only then, go ahead and create the HTML/CSS output. And voila, see the below sample output from another of our internal projects. Sure, it could do with some formatting fixes (prompt Claude to do this) but it is almost ready! 🥳
💡 Pro tip: Once your presentation is ready, share it with Claude for a mock presentation run. Ask Claude to role play as your target stakeholder (e.g., "You are the Chief Product Officer at a leading Indian e-commerce company") and critique your presentation or ask challenging questions. This helps pressure-test your narrative and prepare for the actual presentation.
What’s next?
If you've made it this far, well done! Remember these key takeaways:
Document your process and tools for reproducibility
Cross-check AI’s output with multiple sources, especially for the important data points
Stay open to diverse perspectives to avoid tunnel vision
Ready to transform the way you research? Pick your next project, experiment with these tools, and take the leap from data overload to insight-driven success.
In our next post, we'll explore strategies to supercharge your workflows. See you next time! 👋