Bain 指出，上一波投资「回报不足」，可用的节省池比假设更小，而当前一波 AI 投资的商业案例也多是建立在预测而非实际数据上。虽然有些公司用已实现的节省来资助生成式 AI 与 agentic AI，新一轮支出的最大资金来源仍是「有针对性的节省」——有 44% 的公司这么表示。
Bain 认为，AI 专案表现不佳的首要原因不是预算或优先顺序，而是企业无法可靠取得自己的资料；即使全球已投入数千亿美元做资料现代化，资料结构与可存取性仍是障碍。其建议是：不要等到把所有资料都整理好才开始，而应先用现成资料喂给模型，再让 AI 协助整理其余资料。
Bain & Co.’s global survey shows that automation-driven cost savings are broadly missing expectations. Completed in April, the survey covered 951 companies with more than $100 million in revenue across nine sectors; among firms that measured AI savings, the largest share (40%) achieved reductions of 10% or less, even though most had expected much stronger improvement.
Bain said the previous wave of investment underdelivered, leaving a smaller savings pool than assumed, and warned that the current AI investment case was built on projections rather than actual results. Although some companies are funding generative and agentic AI with realized savings, the largest share of next-wave funding came from targeted savings, cited by 44% of respondents.
Bain argued that the main reason AI programs underperform is not budgets or competing priorities but unreliable access to a company’s own data. Despite hundreds of billions of dollars spent globally on data modernization, structure and accessibility remain barriers; its prescription is to start with available data, feed it into models first, and then use AI to help organize the rest.