Let's start with the number that matters: $550 saved, automatically, by an AI agent called Instinct, without the user doing anything beyond setting it up. According to Wired's detailed hands-on review, the agent monitored subscriptions, identified a better deal, switched it, and flagged the saving. For a solopreneur or small Edinburgh business owner juggling dozens of recurring costs, that kind of background financial vigilance is worth serious attention.
Instinct is what the industry now calls an "agentic" AI: software that doesn't just answer questions but takes action. It booked restaurant reservations, managed inbox tasks, and, in one notable moment, caught a phishing email that had slipped past standard filters and warned the user before any damage was done. These are not party tricks. These are the kinds of tasks that eat twenty minutes here, an hour there, and quietly drain a small team's week.
The appeal for independent operators is obvious. Research from McKinsey's 2024 State of AI report found that small businesses adopting AI automation tools reported reclaiming an average of 10 to 15 hours per week on administrative tasks alone. The University of Edinburgh's Centre for Technology Innovation has separately noted that Scottish SMEs remain significantly under-indexed on AI tool adoption compared to their counterparts in London and the South East, leaving genuine productivity gains on the table.
But Wired's reviewer is candid about the downsides, and those candid bits are worth reading carefully. The agent also wasted $64 on a task it misread, completing an action the user hadn't sanctioned. More significantly, granting any AI agent access to your email, calendar, and financial accounts is a meaningful security decision. The reviewer flags this explicitly: an agent with that level of access is a substantial attack surface. If the service is compromised, or if the agent misreads a malicious instruction buried in a document, the consequences could be significant. This is not a reason to avoid the technology. It is a reason to read the permissions screen properly and not connect accounts you cannot afford to expose.
The practical read for Scottish SMEs is this: agentic AI is arriving fast, and the early evidence suggests it delivers genuine value for the solo operator or small team. The National Cyber Security Centre (NCSC) published updated guidance in early 2025 on third-party AI tool risk, recommending that businesses treat AI agents with the same scrutiny they would apply to any third-party software with elevated account access. Start with low-stakes accounts. Test with tasks that have bounded consequences. Let the agent earn trust incrementally, the same way you would a new member of staff.