What I Built With Claude's Newest Model in One Afternoon (And What It Replaced)
I've been experimenting with Claude's newest model, Fable. I’m kind of blown away by what I’ve been able to accomplish with just a few hours of experimentation.
Here's the situation. Going fractional means I no longer have a huge tech stack or an embedded operations team. I'm a solopreneur. It’s me. That’s it. No research team, no data analyst, no teammate to hand a project to and say, "See what you can find and get back to me in a couple weeks." If I want market intelligence, I either buy it, outsource it, or go without it. Most of the time, as a fractional marketer working across multiple clients, I've gone without it. Not because it isn't valuable. I live by data and market insights. But because the cost and the timeline just wouldn’t make sense for one person.
So I decided to see what would happen if I treated Fable like a research partner plus off-the-shelf software that I could license. (Note: until now, I’ve created skills, used cowork, and collaborated with Claude mostly for one-off research or writing, but I’ve never attempted anything this robust.)
The prompt that started it
I gave it a role and a task, and a sense of what kind of outcome I was looking for. I started with “You're a research analyst working with me, a fractional CMO serving K-12 and higher ed edtech companies. Build a brand tracker scanning Reddit and trusted edtech industry voices. I want quantifiable share of voice, brand themes, sentiment, something trackable over time, and exportable to blog content. Ask clarifying questions before you start.”
That last line was the key to what became very effective collaboration. Fable came back with multiple-choice questions with pre-selected fixed options, including a recommendation. I liked that I could choose an option “as is” or clarify further. These options included things like “Open discovery or a fixed watchlist of brands? Which segments matter? Which sources actually count as trusted?” Those questions settled the whole foundation before a single search happened.
What came out of it
Four iterations later, in one working session, I had a two-file, browser-based dashboard tracking 17 brands and 6 industry themes across 4 market segments. It scores Share of Voice (SoV) three ways (earned, owned, and total, to make sure that a brand's own PR or content publication never gets counted towards SoV or sentiment). It clusters sentiment with actual citations behind every score. In one of my later iterations, I had the idea to explore how a company’s campaign activity will influence brand SoV and earned sentiment over time, so Claude created an "amplification ratio."
I find this extremely exciting, especially with back-to-school season on the immediate horizon, which is one of the biggest marketing spend times of year across K-12 and higher education. The amplification measurement shows whether a brand's campaign spend is actually landing with the market or just shouting into the void.
Lastly, I want to be able to measure a client’s performance in contrast, even if a client of mine isn’t ranking organically within the top brands for SoV. This has been extremely helpful because it helps keep conversations grounded in data as compared to a client’s perception, and also provides immediate insight for how a client can improve.
Again, one of the biggest takeaways for me was the collaborative back and forth with Claude using Fable to get to an output I’m genuinely excited about. Claude truly became a coworker I was iterating with. I'd ask "what would you change to make this as accurate as possible?" and it would name real weaknesses (SEO listicle farms inflating the counts, a brand's own blog counting toward its own share of voice) and fix them. When we tightened the source list, total tracked mentions dropped from 88 to 54. That felt like a step back, until I realized smaller and accurate tracking beats bigger and noisy every time.
What this actually replaced (or enabled)
Here's the part I keep coming back to. If I'd wanted this same capability without Claude, I had two options, and neither is reasonable for a solo operator.
Option one: hire it out. A single deep competitive intelligence report from a research partner typically runs somewhere between $5,000 and $15,000 for multi-competitor analysis (data from a 2026 market research cost breakdown from MX8 Labs), and freelance market researchers themselves bill $50 to $300 an hour. For what it’s worth, I’ve spent much more than this in the past. The output is often a massively detailed deck featuring a one-time snapshot, unlike what I built with Claude, which updates weekly and is immediately relevant and actionable.
Option two: buy the software. Enterprise social listening platforms like Brandwatch, Meltwater, and Talkwalker are the tools built for exactly this kind of tracking, and none of them are built for a company of one. Vendr's 2026 pricing data puts Brandwatch's median annual contract around $50,000, with entry pricing on Meltwater starting around $15,000 to $25,000 a year and Talkwalker's base plans around $9,600 a year, all requiring 12-month commitments. These are great tools, and I’ve licensed them when running large teams with a seven-figure budget. This just isn’t reasonable for me right now, and it’s a little early for most of my clients to invest in these tools.
Result: I can now associate campaign activity to Share of Voice and brand sentiment dynamically and across an evolving competitive set every week. I built something that does the job expensive tools are meant to do, and then some in one afternoon for the cost of a Claude subscription.
That's the real story here. It's not that AI wrote something for me. It's that AI made a category of work financially and logistically possible for a solopreneur that simply wasn't possible before.
What I'd say if asked…
Claude didn’t replace my judgment. It didn't decide which sources counted as trustworthy; I did. It didn't decide that a client with zero earned mentions (yet) needed a comparative spot on the dashboard; I did. What it replaced was the weeks of manual research, the research agency I couldn't afford to hire, and the enterprise software contract I'd never sign as a business of one.
If you're running a business solo, I'd encourage you to try the same thing I did: Think big about the things you used to do or tools you used to have access to and experiment. Then iterate, and expand even further. It’s like manifesting - how big can you imagine your Claude?
I'll be running this tracker weekly from here, and I'm planning to share some of what it surfaces (starting with a genuinely interesting finding: the biggest conversation in edtech right now isn't about any one brand, it's about vendor trust and data security) in upcoming posts.
Let's Talk
If you're a fellow solopreneur or fractional leader curious about how you're using AI to close the resource gap, I'd love to hear about it.
Sources
Vendr, "Brandwatch Software Pricing & Plans 2026" and "Meltwater Software Pricing & Plans 2026," 2026. vendr.com/marketplace
MX8 Labs, "What Market Research Actually Costs in 2026: A Transparent Breakdown," May 2026. mx8labs.com