A client came to me with a research problem.
They run a supplement brand — creatine specifically — and they wanted to understand the market. Not in a vague, “what’s the industry doing” way. They wanted to know what consumers actually think about every major competitor. What they love. What frustrates them. Where there’s white space nobody has filled.
The thorough way to answer that question would be to read through thousands of Amazon reviews across a dozen products. Note the patterns. Cross-reference the complaints. Pull out the quotes that keep showing up.
Nobody does that. It takes weeks. You’d need a researcher or an analyst, and even then, you’re going to skim.
So I built an agent to do it instead.
What the agent did
The agent scraped all the Amazon reviews for every major competitor — roughly 10,000 reviews total across multiple SKUs.
It ran sentiment analysis on the reviews. Grouped them by theme. Flagged the recurring complaints. Identified the features consumers praised most. And surfaced the gaps — things multiple reviewers asked for that none of the products actually deliver.
Time to run: about 20 minutes.
Manual equivalent: somewhere between two weeks and a month, depending on how thorough you want to be. And even at the end of a month, a human analyst is going to miss things.
The output was a structured market report that told them exactly where the opportunities were.
The pattern keeps showing up
I saw the same dynamic when I was organizing an event and needed press coverage.
I needed to find journalists who cover Austin business and tech, understand what they write about, and reach out with pitches that actually fit their beat. Do it wrong and you’re just sending cold emails that get ignored. Do it right and you get coverage.
The old way to do this research: search each journalist by name, read their recent articles, check their LinkedIn, figure out the angle. For 30 journalists, that’s probably 15-20 hours of work — if you’re fast and organized.
I did it with AI deep research instead. Found the relevant journalists, pulled their recent work, identified the story angles that matched their coverage areas, and built the outreach list.
A few hours instead of 20+.
Austin Business Journal ended up covering the event.
Why research is the real AI opportunity
Most businesses are using AI to write things. Email drafts. Social captions. Blog posts. Marketing copy.
That’s fine. But writing is already fast. A decent writer can produce a solid draft in an hour. The improvement AI delivers there is real but incremental.
Research is different. Research is slow. Research is where you lose weeks.
And “slow” isn’t the only problem. The other problem is that expensive research often gets skipped or shortcuts get taken. You need to understand your market, so you read 50 reviews instead of 5,000. You need to know your competitors, so you look at their homepage instead of analyzing all their customer feedback. The quality of the research scales with how much time you put in — and nobody has that time.
AI changes the equation. You’re not just doing the same research faster. You’re doing research that wasn’t economically viable before.
10,000 reviews? No analyst is reading all of those without shortcuts. No team has the bandwidth. But an agent doesn’t get tired, doesn’t skim, doesn’t stop at page 3.
Research tasks AI can handle now
Here’s a short list of research categories that work well with AI agents today:
Competitor intelligence
Scrape reviews, collect feedback, identify patterns. Works across any consumer product category where public review data exists.
Customer sentiment analysis
Feed in support tickets, reviews, survey responses, or interview transcripts. AI finds the patterns faster than any analyst would.
Lead enrichment before outreach
Pull background on a company, its team, recent news, and funding history before a pitch. Works for sales calls, investor meetings, partnership conversations.
Journalist and investor research
Identify who covers your space, what they’ve written about recently, and what angles might land — before you reach out.
Market gap analysis
Compare what consumers ask for vs. what existing products deliver. Works anywhere public feedback data exists.
One thing to try this week
Pick one research task you’re doing manually right now — or skipping because it takes too long.
It doesn’t need to be a 10,000-review project. Start smaller. A competitor’s last 50 reviews. Background on five leads before calls this week. Recent coverage from three journalists in your industry.
Build the agent, run it, see what you get. You’ll know pretty quickly whether AI can absorb that research loop — and what that frees you up to do instead.
Recommended for you
Want the full system? 25X is the flagship productivity system we teach.
Explore 25X →