Sam Finnegan-Dehn has two jobs. By day, he works in fundraising for a charity from his home in London. By night and weekend, he's a math and philosophy tutor for university students—a side hustle fueled by academic degrees and genuine passion for the subjects. But here's the thing nobody warns you about when you go freelance: the teaching is maybe 30% of the actual work. The rest is lesson planning, research, invoicing, client management, and chasing down leads on new reading material. That overhead adds up fast, especially when you're squeezing tutoring hours around a full-time gig.

Teaching AI to Handle Your Admin

So Finnegan-Dehn did what any resourceful operator would do—he found a digital assistant that doesn't need coffee breaks or health insurance. Specifically, he uses Notion AI to handle the secretarial tasks scattered across his digital notebooks: tracking client progress notes, flagging relevant new research, and maintaining an organized record of where each student stands in their learning journey. He describes it as "a second memory" that connects ideas he's written down across different notebook tabs—essentially turning a pile of scribbled observations into something searchable and actionable. After experimenting with Claude and ChatGPT, he settled on Notion AI because it plugs directly into his existing workflow rather than requiring him to copy-paste content into yet another interface. The setup isn't glamorous, but it's practical. Finnegan-Dehn uses the tool's meeting recorder (with client consent) to generate automated summaries of tutoring sessions. If the summary indicates a particular teaching technique wasn't landing with a student, he adjusts his approach next time. Beyond that, Notion AI helps draft lesson notes, handle invoicing, and even sync social media posts—all tasks that demand attention but not creativity. For goal-setting specifically, Finnegan-Dehn feeds the system a "North Star" objective—say, hitting a certain client count by year-end—and asks it to generate the concrete steps needed to get there based on his profile data.

Real Numbers From Real Businesses

This isn't just one tutor's experiment. Grandma's Quilt Shop in Yuma, Arizona, uses Rain, an AI platform with software specifically built for craft businesses, to generate inventory descriptions and pricing for their fabric designs. The owners claim the tool cuts item-listing time by 60 to 80%—a meaningful efficiency gain when you're managing hundreds of SKUs. The pattern emerging across these case studies is consistent: AI excels at rote, structured tasks that follow predictable patterns. Transcription, summarization, inventory description generation, invoice drafting—these are all areas where a language model can produce acceptable output with minimal supervision.

The Fine Print Nobody Talks About

But let's be real about the tradeoffs. Finnegan-Dehn himself calls parts of Notion AI "clunky" at times—idiosyncrasies that break immersion when you're trying to move fast. And there's the matter of cost: $20 per month for the Notion AI add-on. For a solopreneur already counting pennies, that's not trivial. The article's user tips section gets into territory that more businesses should heed before diving in: LLMs learn from your inputs, so think carefully about what data you're feeding third-party platforms and whether their ecosystems actually fit how you work. AI hallucinates—meaning accuracy-critical tasks still need human oversight. Sometimes existing off-the-shelf tools like Shopify or Square beat whatever you'd "vibe-code" with a chatbot. And for anything sensitive—whether that's client PII or proprietary business data—running an open-source model locally might be smarter than sending everything to OpenAI's servers.

Key Takeaways

  • AI tools like Notion AI excel at rote, structured tasks: summarization, invoicing, inventory descriptions, and cross-document syncing
  • Integration matters—the best tool is the one that fits your existing workflow without requiring extra friction
  • Real efficiency gains are documented: Grandma's Quilt Shop claims 60-80% time reduction on item listings using industry-specific AI
  • Not all tasks suit AI: hallucinations mean accuracy-critical work still needs human review
  • Data privacy concerns are legitimate—consider local open-source models for sensitive information rather than proprietary APIs

The Bottom Line

Look, the promise of AI for small business isn't about replacing judgment or creativity—it's about reclaiming hours you'd otherwise spend on busywork that drains you. Finnegan-Dehn's model works because he identified exactly where his bandwidth was getting eaten up and deployed a tool purpose-built to handle that specific gap. The businesses that'll get burned are the ones that adopt AI reactively, chasing hype instead of auditing their actual workflow pain points first.