Recent reporting from MIT Technology Review highlights a quiet but rapidly accelerating shift in how small businesses operate, as artificial intelligence tools move from experimental novelties to practical back-office assistants. Across industries, entrepreneurs who once struggled to juggle accounting, scheduling, marketing, and research alongside their core work are now turning to large language models to handle a growing share of routine administrative tasks. The change is not dramatic in appearance, but it is increasingly fundamental in effect: AI is becoming the invisible office layer that many small operations could never afford to hire.
The appeal is straightforward. Running a small business has always required a wide and often overwhelming range of skills. Owners are expected to act as managers, marketers, financial administrators, researchers, and customer service representatives, often simultaneously and without formal support staff. While larger companies can distribute these responsibilities across specialized teams, smaller operators typically rely on personal bandwidth and improvisation. In this gap between ambition and capacity, AI tools are beginning to offer what feels like a partial solution: a flexible, always-available assistant capable of handling repetitive, structured, or time-consuming work.
One case study featured in the MIT Technology Review report illustrates this shift through the experience of Sam Finnegan-Dehn, a London-based fundraising professional who also works as a private tutor in mathematics and philosophy. Although tutoring is his primary passion, it is only a fraction of the labor involved in sustaining his side business. Beyond teaching students, he is responsible for designing lesson plans, sourcing academic materials, tracking student progress, generating assignments, managing communications, and issuing invoices. Like many independent educators and freelancers, he found that administrative demands were limiting his ability to expand his client base, despite strong demand for his services.
To address this constraint, Finnegan-Dehn began experimenting with AI tools as a form of digital assistant. Over time, he transitioned from general-purpose models like ChatGPT and Claude to Notion AI, which integrates directly into his existing digital workspace. His tutoring practice already relied heavily on Notion for organizing notes, reading lists, and student records, making the platform’s AI features a natural extension rather than an external tool. In practice, he describes the system as a kind of second memory that helps connect fragmented information scattered across different notebooks, allowing him to retrieve and synthesize insights more efficiently than manual searching.
Rather than using AI to generate teaching materials from scratch, Finnegan-Dehn relies on it primarily for administrative augmentation. One of its most significant functions is recording and summarizing tutoring sessions, with the consent of his students. These automated summaries allow him to reflect on patterns in student performance and adjust his teaching strategies accordingly. If the AI-generated notes suggest that a particular explanation or technique is not resonating, he can revise his approach in future sessions. In this sense, AI is not replacing pedagogical judgment but reinforcing it with structured feedback.
Beyond session analysis, Notion AI also supports his broader business operations, including goal-setting, invoicing, lesson documentation, and even drafting social media posts. One of the most useful applications, he explains, is its ability to translate abstract ambitions into actionable steps. When he sets a long-term objective, such as increasing his client base by a specific number before the end of the year, he can input that goal into the system and ask the AI to generate a structured roadmap. The model then proposes incremental tasks based on his existing profile and activity data, helping him bridge the gap between intention and execution.
The platform enabling these workflows, Notion, has evolved significantly in recent years. Originally known as a flexible note-taking and productivity application, it introduced its AI add-on in late 2023. Since then, it has expanded to include integrations with email systems, calendar tools, and more advanced automation features, including agent-like functionality. These capabilities allow it to interact with multiple digital environments simultaneously, effectively positioning it as a centralized assistant for knowledge work. However, this increased access has also raised concerns about data privacy and the extent to which sensitive business information is processed by external systems.
The broader market for AI-driven small business tools is also expanding beyond general productivity platforms. In the retail and craft sector, for instance, specialized systems are being developed to automate highly specific workflows. One example referenced in the MIT Technology Review coverage is Grandma’s Quilt Shop in Yuma, Arizona, which uses a platform called Rain designed for craft-based businesses. The software assists with generating product descriptions, inventory listings, and pricing recommendations for fabric designs, significantly reducing the time required to prepare items for sale. According to the shop’s owners, listing efficiency improved by as much as 60 to 80 percent after adopting the tool, demonstrating how targeted AI applications can directly impact operational throughput.
Despite these gains, the adoption of AI in small business contexts is not without friction. Finnegan-Dehn himself has described some of Notion AI’s outputs as occasionally clunky, noting that the system does not always align perfectly with his expectations or workflow logic. Cost is another consideration, with AI add-ons often priced as recurring monthly subscriptions that may be difficult to justify for businesses operating on narrow margins. These limitations underscore a central tension in the current AI landscape: while the tools can reduce labor in certain areas, they also introduce new dependencies and ongoing expenses that must be carefully weighed.
Security and privacy concerns further complicate adoption. As AI systems become more deeply integrated into operational workflows, they gain access to increasingly sensitive business data, from client communications to financial records. MIT Technology Review notes that this raises questions about how such information is stored, processed, and potentially used for model training. Some experts suggest that businesses with even modestly sensitive information may benefit from using local or open-source AI models that operate directly on personal devices rather than transmitting data to cloud-based services.
