What small firms want before they trust an AI agent

My MSc final project was a business plan for ProAgentix, a company I proposed but never set up. It would run AI agents on a firm’s own devices and use the cloud only when needed, so that the data stayed under the firm’s control. The plan was aimed at software and IT firms in London with one to nine staff.

Before writing the plan, I wanted to know whether small firms would want it. Through Prolific, a UK online research panel, I ran a survey aimed at decision-makers in London micro-businesses. It drew 102 responses, and I followed up with interviews.

The first surprise was cost. Most respondents (73.5%) had spent between £0 and £50 on AI tools in the previous month, and only about one in ten had spent more than £200.

The second surprise was the order of the worries. Asked for the biggest obstacle to adopting AI agents, a third chose accuracy (34%), a quarter privacy (26%) and a fifth the cost of usage (22%). Privacy still mattered. Three in four (75%) rated local control of their data as necessary or critical. In the interviews, the same point came up in practical terms, including from one person who described checking every answer an AI tool gave before using it.

Nearly half (48%) chose the cheapest plan, at £120 a month, and a third (33%) would not buy any plan. In the interviews, the same £120 struck one person as a large cost and another as too cheap to be credible. What decided it was what the tool would free them to do, and whether they could try it first and leave if it did not work.

The sample has limits. Respondents came from an online panel and described themselves, and about one in six did not match the target profile or gave no role. Their answers remain in the figures, so the results are indicative. It was enough, though, to change the emphasis of the plan. The version I submitted led with results people could see and an easy start. Keeping data local became the reason to choose the product, though no longer the whole case for it.