Small-business operators are quietly turning general-purpose AI tools into back-office staff, with Notion AI's $20-per-month add-on and vertical tools like Rain doing the work a solo proprietor would otherwise hire out. Sam Finnegan-Dehn, a London tutor who moonlights teaching math and philosophy on top of a charity fundraising job, has routed invoicing, lesson notes, goal-setting, and social media drafting through Notion AI. Grandma's Quilt Shop in Yuma, Arizona, has cut the time it takes to list inventory items by 60 to 80% using Rain, a software suite tailored to craft companies. The pattern, documented in MIT Technology Review's Making AI Work newsletter on June 2, 2026, is the unflashy half of the AI economy: not frontier reasoning, but rote work being absorbed by general models.
Finnegan-Dehn tested both Claude and ChatGPT before landing on Notion AI, choosing it because it sits inside the notebook app where his tutoring notes already live. The integration matters more than raw model quality for his use case. He uses the tool to record client meetings with their consent, generate automated summaries, and refine his teaching approach based on what the summaries surface. If a technique appears in the summary as ineffective, he adjusts for the next session.
Goal-setting is the other workflow he leans on. He writes down a "North Star" — a target number of clients by the end of the year — then asks the AI to generate the concrete intermediate steps based on the profile it has built from his notes. He then picks which tasks to tackle first. It is the kind of structured planning a human assistant or business coach would charge meaningfully more than $20 a month to provide.
Key facts
- 01Notion AI, released in late 2023, now costs $20 per month and integrates email, calendar, and a newly released agent.
- 02Grandma's Quilt Shop in Yuma, Arizona uses Rain to cut item-listing time by 60 to 80%.
- 03London-based tutor Sam Finnegan-Dehn tested Claude and ChatGPT before settling on Notion AI for its tighter integration with his existing notes.
- 04Notion AI records client meetings with consent and produces summaries used to adjust teaching strategy.
- 05Reporting appeared in MIT Technology Review's Making AI Work newsletter on June 2, 2026.
Notion, long established in note-taking software, released its AI add-on in late 2023 and has since extended it with an email client, calendar integrations, and a newly released agent. That access has raised privacy concerns — the model sees whatever lives in the workspace — but it is also what makes the tool useful as a virtual assistant rather than a chat window. The tasks it targets are largely rote: syncing information across documents, searching old scribbles, drafting routine correspondence.
Rain illustrates the other path: vertical tools built for one industry. Grandma's Quilt Shop uses it to generate inventory descriptions and pricing for fabric designs, the kind of repetitive copywriting that eats hours in a small retail operation. The owners' claim of a 60 to 80% time reduction on listing work is the sort of productivity number that, applied across a craft retailer's full catalog, translates directly into either more inventory turned over or fewer hours worked.
Finnegan-Dehn is candid that Notion AI is not seamless — he describes some of its behavior as "clunky" — and the $20 monthly cost is not free money for a part-time tutor. The calculation small operators have to make is whether the time saved exceeds the subscription plus the hours spent learning the tool and cleaning up its mistakes. For tasks where accuracy matters, hallucinations are not theoretical, and the tutor keeps human judgment in the loop on anything client-facing.
There is also a deliberate carve-out for tasks where AI is the wrong tool. Payment processing is the example flagged in the MIT Technology Review piece: using an established platform like Shopify or Square is safer than vibe-coding a checkout flow with an LLM. The framework being modeled is not "replace everything with AI" but "identify the skills you lack in-house and see whether AI can handle them or train you up," with humans retained where errors are expensive.
Data sensitivity is the other constraint. The reporting notes that proprietary chatbots collect prompt data, and that for any genuinely sensitive material a local open-source model running on a laptop or small desktop is now a viable alternative to ChatGPT or Claude. That option did not exist as a practical matter two years ago. The proliferation of small models that run on consumer hardware — covered repeatedly on this site, most recently with Osaurus passing 112,000 downloads — has quietly made local inference a default option for privacy-conscious operators.
The skeptical read is that none of this is a productivity revolution measured in the macro data yet. A tutor saving a few hours a week and a quilt shop listing items faster do not show up in GDP. Notion AI also competes against free or near-free workflows — a spreadsheet, a calendar, a human assistant for a few hours a month — and the $20 monthly cost is recurring whether or not the operator uses it heavily that month. The privacy concerns flagged around Notion's deeper integrations are real and unresolved.
The more interesting signal is distributional. The biggest AI stories of the past month have been about hyperscaler capex — Alphabet raising $80B for AI buildout, Nvidia's RTX Spark push into the $200B CPU market — but the marginal user gain from those investments is concentrated at large enterprises with the engineering staff to deploy custom systems. A London tutor and a Yuma quilt shop are getting comparable workflow lift from $20-per-month off-the-shelf tools. If that pattern holds, the small-business AI market is less about who builds the best frontier model and more about which products land cleanly inside the notebooks, point-of-sale systems, and inventory tools that solo operators already use.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
