人工智能繁荣的可行性取决于大型企业是否发现人工智能模型足够有用且有利可图，以维持OpenAI和Anthropic等模型制造商的生计。硅谷最初预测，白领办公任务将迅速被自动化的代币（tokens）而非人类员工所取代，但这一转变的进展比预期要慢得多。现代企业数据杂乱无章，难以整合到人工智能模型中，这引发了关于自动化策略、开源替代方案、代币成本套期保值以及投资回报率的复杂运营问题。

为了弥补这一应用差距并解决关键的现金流问题，人工智能模型制造商正越来越多地进入专业咨询领域。OpenAI最近收购了一家英国小型科技咨询公司，而OpenAI和Anthropic都在5月份与大型私募股权公司建立了合资企业，以帮助重振其基金中表现不佳的软件投资组合公司。此外，埃森哲（Accenture）等咨询巨头正在对大量员工进行这些技术的培训，包括在12月份宣布对其80万名员工中的3万名进行Anthropic的Claude模型培训。

然而，金融市场对这种服务密集的模式仍持怀疑态度，埃森哲的股价今年已缩水过半，而数据分析提供商Palantir也遭遇了类似的下跌。崇尚个人天才的传统人工智能实验室在管理实际企业整合所需的大量劳动力方面面临着文化障碍。最终，这些举措的成功取决于科技公司能否在老牌咨询公司成功掌握人工智能之前，先掌握普通企业错综复杂的运营细节。

The viability of the artificial intelligence boom hinges on whether major corporations find AI models useful and profitable enough to sustain model-makers like OpenAI and Anthropic. Silicon Valley initially predicted that white-collar office tasks would quickly be replaced by automated tokens rather than human workers, but this transition has progressed much slower than expected. Modern enterprise data is disorganized and difficult to integrate into AI models, raising complex operational questions about automation strategies, open-source alternatives, token cost hedging, and return on investment.

To bridge this implementation gap and address critical cash flow issues, AI model-makers are increasingly entering the professional consulting space. OpenAI recently acquired a small British tech consultancy, while both OpenAI and Anthropic established joint ventures in May with large private-equity firms to help revitalize underperforming software portfolio companies. Furthermore, consulting giants like Accenture are training large numbers of workers on these technologies, including a December announcement to train 30,000 of its 800,000 employees on Anthropic's Claude model.

However, financial markets remain skeptical of this service-heavy approach, as Accenture's share price has lost more than half of its value this year, and data analytics provider Palantir has experienced a similar decline. Traditional AI labs, which prize individual genius, face cultural hurdles in managing the large workforces required for hands-on corporate integration. Ultimately, the success of these initiatives depends on whether tech firms can master the intricate operational nuances of ordinary businesses before incumbent consulting firms successfully master artificial intelligence.

Source: Silicon Valley has much to learn from the spreadsheet jockeys it despises

Subtitle: Even if financial markets disagree

Dateline: 6月 25, 2026 09:14 上午