AI adoption at SMEs still 20.4%, but 83.2% who try it cite time saved

The article's starting point is a strain many small firms recognize: hiring is hard, and a handful of people are absorbing inquiries and order entry on top of their real jobs. That is the gap it says AI agents — which plan steps and operate tools rather than just answering questions — are meant to close. The body draws the contrast twice: generative AI leaves the copying and pasting to a person, while RPA stops when a form or a request breaks the pattern, so the exception-heavy work RPA never reached becomes the candidate to hand over. It also notes that Microsoft 365 and Google Workspace already bundle agent-building features that can be instructed in Japanese, meaning a firm does not need to wait for an IT hire to begin.

The middle of the article stacks evidence. SME Support Japan's survey puts adoption at 20.4% and, among adopters, work-efficiency or time savings at 83.2%, with labor-shortage response second at 33.9%. Shinkin Central Bank's survey found the most common use among active generative-AI adopters was drafting and summarizing emails, reports and minutes, at 74.0%. Three named cases follow: Theoria House Clinic, REFLAME and Okada Kenma. Notably, the first two are Salesforce Agentforce case studies, so the figures come from the vendor — useful as internal-proposal material, but worth reading as vendor-published rather than independent.

What the piece ultimately hinges on is discipline rather than technology. Its closing advice is to pick one task, write down the steps, pilot with two or three people, set permission and human-review rules, and measure before broadening. Whether that restraint holds — and whether the Digitalization/AI Introduction Subsidy 2026, the SME Labor-Saving Investment Subsidy and the Human Resources Development Promotion Subsidy actually fit a given firm's setup — is likely to decide which SMEs capture the time savings the survey describes and which simply buy unused licenses.