Our Take
AI works best on the work no one wants to do, but small business owners often can't afford to find out which tasks those are without paying month-to-month.
Why it matters
Solo operators and micro-businesses have neither the budget for specialists nor the time to test new tools. The MIT case studies show where AI actually sticks—but also where it creates new overhead.
Do this week
Audit: map your weekly routine in 15-minute blocks, tag the ones that don't require judgment, and test one AI tool on exactly those blocks for two weeks before committing to the subscription.
A tutor and a craft shop found different AI fits
Sam Finnegan-Dehn tutors philosophy and math from London while working full-time in fundraising. He uses Notion AI to do what he calls "secretarial" work: recording client sessions with consent, summarizing them to spot teaching gaps, drafting lesson notes, syncing invoices, and generating social media posts. The tool costs $20 per month (company-reported).
Grandma's Quilt Shop in Yuma, Arizona uses Rain, an AI suite built for craft vendors, to generate product descriptions and pricing. The owners report it cuts inventory listing time by 60 to 80% (company-reported).
Both cases share a pattern: the AI handles rote work. Finnegan-Dehn still creates the teaching strategy; Rain still leaves pricing decisions to humans. Neither business is using AI to do the core skill.
Where the advice falls short
MIT Technology Review offers a checklist: audit what skills you lack, don't feed sensitive data to cloud models, use local LLMs when privacy matters. All sound. But the guide glosses over the time cost of evaluation itself.
A solo operator trying four different AI tools at $15-25 per month burns cash and attention while learning which one fits their workflow. Finnegan-Dehn tested Claude and ChatGPT before landing on Notion AI because it integrated with his existing notes app. That integration mattered more than the model. Many small business owners won't have the luxury of that trial period, and many won't discover the fit until they've already paid three months of subscriptions.
The other unspoken cost: AI hallucinates. Finnegan-Dehn described Notion AI as "clunky" at times. A tutor can catch a bad lesson summary. An inventory system generating product descriptions might publish errors before anyone notices. The advice to keep humans in charge of accuracy tasks is correct. It is also a reminder that AI doesn't eliminate the need for oversight; it changes where oversight is required.
Start with the job no one wants
Both case studies work because neither business asked AI to replace a core competency. Finnegan-Dehn doesn't ask Notion AI to write philosophy lessons. Grandma's Quilt Shop doesn't ask Rain to pick which fabrics to stock. Instead, they use AI for the admin tail.
Before signing up, list the work you actively dislike or that has no creative payoff. Does it take less than 30 minutes per week? If yes, the subscription almost certainly costs more than just doing it yourself. If it takes 3+ hours, the math moves in favor of testing. Pick one tool, commit to two weeks, and measure whether you actually freed up time or just added another tab to check.
And consider the ecosystem lock-in. Finnegan-Dehn chose Notion AI partly because his notes already lived in Notion. If you commit to a tool, you're committing to the platform. That switching cost is real even if the subscription is cheap.