Japan AI results lag: 9% see outsized gains, PwC finds

The article frames AI agents not as a single product choice but as a measurement problem. It lists eight expected effects — four quantitative ones like cutting work hours and labor costs, and four qualitative ones like reducing errors and freeing staff for higher-value work — and pairs them with Japanese and global case studies where the numbers were disclosed. Panasonic Connect cut 788,000 hours in fiscal 2025 through its ConnectAI assistant, a figure reached through 3.61 million uses at 33 minutes saved each; Klarna's assistant handled 2.3 million conversations in its first month and cut resolution time from 11 minutes to under two. Yokohama Bank, Jaroc and TAPP show the same pattern at different scales, from a major bank down to a roughly 50-person manufacturer.

The counterweight comes from the survey data. Only 23% of companies in McKinsey's state-of-AI survey have scaled AI agents anywhere in the organization, and of the 39% who said AI affected EBIT, most put that contribution below 5%. Klarna is the body's own cautionary case: after its AI-first push, its CEO told Bloomberg that quality had fallen and the company resumed hiring human operators. Gartner's forecast that over 40% of agentic AI projects will be canceled by the end of 2027, citing rising costs, unclear business value and weak risk controls, points at the same weak spot.

Taken together, the case studies and the survey results suggest the difference lies less in the tool than in whether a company defines its metrics and picks the right work before deploying. What remains open is whether Japanese firms, whose adoption rate is already high while their reported results are the lowest of six countries surveyed, will close that gap by reworking processes rather than simply distributing more tools.