How to Actually Find Business Ideas with AI
Most people ask ChatGPT for ten business ideas and get a generic list. In this walkthrough I show you the opposite approach: every business starts with a problem, so we use Apify to scrape Reddit, where people love to complain, and Claude Code to find the recurring problems within a niche. Targeting recruitment, we end up with 51 documented pain points backed by 178 real quotes, plus an HTML dashboard to make sense of it all.
What you need
- An Apify account. Apify gives you $5 of free credits each and every month, which helps a lot
- Claude Code. I used Fable 5 for this, but Opus 4.8 handles this use case perfectly fine
1. What Apify is and creating a token (0:30)
Apify is a service that finds information on the web for you. Think of it as tiny robots pulling things from Instagram, Facebook, TikTok, Reddit, and so on. Create an account, then click Go to console. From the console, click Settings, then API & integrations, and create a new token. Copy that token, because we are handing it straight to our tool of the day, Claude Code.
2. The two-step Reddit search plan (1:33)
I am targeting the recruitment niche, and the search breaks into two steps. First, a broad search: use Claude Code to find the subreddits where people talk about recruiting. Second, a narrow drill-down: inside those subreddits, find what people are actually struggling with.
Why bother? Because this way you are not pretending a problem exists. You are finding people with a latent problem and building a service around it. That is the difference between a business that works and a nicely formatted list of guesses.
3. Prompting Claude Code (2:32)
Paste your Apify token into Claude Code and give it the whole plan in one message:
Hey, Claude. Right now, I'm going to need your help. This is an Apify key. I want you to do two different searches. First, I want you to search subreddits where people are talking about recruiting. And then once you find those subreddits, I want you to drill down into them and find the pain points of these people. What are they having problem with in terms of recruiting? Is it the quality of the talent? Is it not enough volume? You let me know. I will also need the sources for each one of these problems. You can use a dynamic workflow for this.
Claude Code understands it is an Apify key, fetches what it needs about Apify, and runs the two-step search. Note the last line: I asked for a dynamic workflow because dynamic workflows let you work way faster. The trade-off is that they consume your tokens at a much higher rate.
Asking for sources on every claim is not optional. That is what keeps this grounded in what people actually said instead of what the model thinks sounds plausible.
4. The results: 51 pain points from Reddit (4:01)
The workflow took about 10 to 15 minutes and came back with 51 different pain points supported by 178 different quotes. This is exactly what you need when you want to start a business or open a new vertical inside the one you already run.
5. The subreddits it found (4:21)
The broad search surfaced the places where our target audience actually hangs out: recruiter tea, modern hiring, human resources, ask HR recruiting, and ask recruiters. The same move works for IT, sales, marketing, really anything, and you can go very niche. It also works for B2C: subreddits where people talk about day-to-day problems like weight loss or getting sober.
The drill-down then took the top 195 posts from the past year and 1,800 comments for each of those subreddits and analyzed all of it.
6. The 7 main pain points (5:14)
Seven main pain points came out of the analysis. The headline: the number of applications has exploded while the number of qualified candidates has dropped. Among the rest: a skill verification crisis, where AI-polished resumes and interviews make competencies unreadable, so it is extremely hard to tell if someone is bullshitting before you meet them. A two-way ghosting epidemic, where neither side follows up. Broken employer-side processes, filtering tools that are backfiring, and a frozen market plus recruiter precarity.
7. Asking for an HTML dashboard (5:38)
Raw findings are hard to absorb, and we are visual beings. So on top of the research I asked for one more thing:
Can you please create an artifact, just an HTML file, that will present all of these findings to me?
That gives us a simple dashboard encompassing everything we just found, so we can actually see what to use.
8. Walking through the artifact (7:13)
The artifact is a simple HTML page, literally just a webpage, with the complete report. The most frequent pain points sit at the top: the AI bot application flood destroying the signal, and the two-way ghosting. Then you can go subreddit by subreddit to understand what people were saying in each one. The picture it paints is clear: there are never too few applicants, it is the opposite, way too many applicants at lower quality. So how do you find the hidden gems?
9. Turning problems into a business (8:01)
Each documented pain point is a service you could provide. From here, use Claude Code or whatever AI tool you prefer to think through solutions, then start iterating: build the first prototypes if it is a SaaS, or work out the pricing, branding, and deliverables if it is a service. AI is solving many problems, but every solution creates a new problem, and that is where we step in.
Pitfalls and tips
- Do not just ask an AI for business ideas with no data behind them. The whole point is finding latent problems that real people describe in their own words.
- Dynamic workflows are fast but burn tokens at a much higher rate. Use them knowingly.
- Always demand sources for every pain point so you can verify the quotes yourself.
Where to go next
The same Apify plus Claude Code combo powers my content research system in Build an X Parasite System with Claude Code. And if you work in the Claude ecosystem, Claude Cowork Plugins Explained in 4 Minutes shows the fastest way to bolt on ready-made capabilities.
