I found a market opportunity by accident while scrolling Reddit.

Someone posted asking for alternatives to InkFrog, an eBay listing tool that's shutting down with 200,000+ users scrambling for replacements. Instead of guessing what to build, I used AI to do deep competitive research.

Here's the step-by-step process I used to turn a random Reddit post into a prioritized product roadmap.

I was browsing Reddit to promote my app when I stumbled on users panicking about InkFrog shutting down. The tool had features similar to mine, so I saw an opportunity. But I didn't want to guess what to build.

I opened Claude and asked it to help me craft a research prompt. I gave it the Reddit thread, InkFrog's website, and shutdown details. Claude built me a detailed research brief targeting the top 20% of features I should focus on.

Then I ran that same prompt in three different AI models: Claude, ChatGPT with deep research, and Gemini Pro. Each gave me different insights because they're trained differently and have different strengths.

The real magic happened when I brought all three reports back to Claude for synthesis. It found the commonalities and built me a unified action plan.

The key insight: this wasn't a feature race, it was a migration race. Users won't pick the tool with the most features. They'll pick the one that gets their listings working before June 1st.

Claude ranked everything by priority. P0 was InkFrog CSV import with image rehosting, a landing page, and master profiles for bulk operations. It even drafted Reddit posts and go-to-market copy.

One warning: the AI initially suggested I pretend to be an InkFrog user in my Reddit responses. I had to correct that. Never fabricate experience you don't have.

The takeaway: when you spot a market opportunity, don't guess what people want. Use multiple AI models to research it systematically. Each AI sees different patterns, but together they give you a complete picture.

BONUS Prompt

Use the prompt below to do this type of research. Mach sure to run it on more than one LLM.

💡 Tips for running this across multiple LLMs:

The only thing you need to change run-to-run is the Context section. Everything else stays the same.

Jensen Huang announced that every NVIDIA software engineer now works alongside AI agents. He says this boosts productivity rather than replacing jobs. Adobe is building an integrated agentic workflow that combines AI models, teams, and brand governance into one system.

Joe's Read: This affects any company with software teams or creative workflows that wants to stay competitive on productivity.

Adobe executives outlined four industry changes: content bottlenecks are shifting from output to human judgment, traditional search is being replaced by AI answers, AI agents can handle entire content supply chains, and strategic AI partnerships are now as critical as product development. Companies are integrating external AI capabilities instead of building everything in-house.

Joe's Read: This affects marketing teams and content creators who need to rethink their search strategy and content distribution approach.

A new framework helps decide which AI tasks to run locally versus in the cloud based on privacy, frequency, and cost. The Mac mini M4 Pro with 64GB memory is recommended as the entry-level personal AI computer. The approach includes scoring tasks and generating phased build plans to avoid expensive, underused hardware purchases.

Joe's Read: This affects small business owners considering local AI setups who want to avoid wasting money on hardware they won't actually use. One of the biggest perks of going this route is privacy (provided you install a free, open-source, local LLM like Ollama)

Claude Cowork lets you create a personalized AI writing skill that mimics your brand voice. You feed it writing samples and answer interview questions until it reaches 95% confidence in matching your style. It can then generate social media posts that sound authentic and require less editing.

Caveat: You'll still need to review and refine the outputs, and it requires ongoing feedback to maintain quality over time.

Read more: https://substack.com/redirect/957e0820-b336-458d-b530-e56b6f396625?j=eyJ1IjoiNWR3bDZrIn0.wq1OENISdHa6Qx7b41P_1bMxVfT9393NhO0K0hlECqc

The biggest mistake people make with AI research is using just one model. Each AI has blind spots. ChatGPT might miss emotional nuances that Claude picks up. Gemini might catch technical details the others miss.

I learned this the hard way. My first research attempts with single models felt incomplete. When I started running the same prompts across three different AIs, the quality jumped dramatically.

It's like asking three different experts for advice, then having a fourth synthesize their recommendations. The synthesis step is critical. Don't just collect three reports and try to merge them yourself. Feed them all back to your preferred AI and let it find the patterns.

This approach works for any market research, not just competitor analysis. Product validation, customer interviews, industry trends. Multiple perspectives always beat a single viewpoint, even when that viewpoint comes from AI.

There's a specific reason why I say that you need to have a “Council of LLMs” that will be there to help you synthesize data. However, I am going to give you one big warning. You're not going to get the expected results if you are using the free accounts for any of these LLMs.

Sure, sometimes they offer a little bit of access to their higher premium models, but for work like this, it may not be enough. I am a huge advocate for getting the paid plans, even if it's the cheapest one, because it's going to give you the best access. I'll be honest, I've never gotten a plan higher than the base $20, so it doesn't have to be crazy.

Think about it this way, those $20 are more valuable than a $20 Netflix subscription (which I don't even think they have a $20 plan anymore because they went up in price.) Now, if you're really in a pinch and you just want to stick with one, I am going to recommend Google and in particular, Google Workspace.

And there's a specific reason for that. If you do Google Workspace (and you tie it to your personal domain or if you have a business domain), it's roughly $26.40 monthly, but you get access to the whole gamut of their tools. So not only are you getting Gmail, Calendar, Docs, etc. with the advanced business features, you also get access to the pro versions of Gemini, as well as AI Studio, VEO3, Nano Banana Pro, and Notebook LM.

And Notebook LM has to be the biggest differentiator here. If you're talking about learning, whether it's about AI or just whatever it is that you're passionate about, Notebook LM is the tool that can teach you something in a variety of ways that works best for you.

Prefer an audio podcast?

Check.

Do you prefer an infographic?

It can also do that.

Do you want an interactive slide?

It can also handle that.

Plus a plethora of other things. But needless to say, it's important to have variety, especially when you're doing research and validation for something you want to build.

n8n — My go-to for standard and AI assisted workflow automations.

VoiceInk — To date, I have saved 8 hours and 28 minutes (around 139,810 keystrokes) using VoiceInk. If you are looking to make the most of your time, start using your voice.

Blotato — Blotato handles getting your content everywhere. I use it to repurpose issues into social posts without doing it manually platform by platform. It is the backbone of my social media system.

Beehiiv — What you're reading right now is published on Beehiiv. If you're thinking about starting a newsletter or moving off a clunky platform, this is the one I'd recommend. 20% off your first 3 months with my link.

Google Workspace — Runs quietly in the background of everything I do. Email, docs, shared drives for client work. 14-day trial and 10% off your first year.

Descript — If you're using VoiceInk to capture ideas and drafts, Descript is where the video and audio side of content creation gets cleaned up. 50% off your first two months on the Creator Plan.