内容来源：https://aibusiness.com/generative-ai/what-ai-s-growing-share-tech-spending-means-enterprises-2027

内容总结：

国际知名研究机构Forrester近日发布报告预测，到2027年，人工智能将占企业技术支出的约20%。该机构分析师布莱恩·霍普金斯在接受采访时指出，当前AI支出的增长几乎全部来自基础设施和算力领域，包括服务器、芯片、云服务、AI模型访问及软件订阅。与此同时，生成式AI正改变企业软件付费模式，基于token的消费计费、使用费和积分等新方式使成本更难预测，智能体成本还随任务复杂度和算力需求波动。

霍普金斯表示，企业软件支出增速低于通胀水平，采购能力受到侵蚀。为腾出AI投资资金，企业正压降IT服务合同价格，要求供应商让渡AI带来的效率红利，并将软件开发、技术支持和客户服务等环节收回自营以削减外部成本。他同时提醒，安全领域不应成为削减对象——攻击者正利用AI智能体以前所未有的规模发现和利用漏洞，安全风险日趋紧迫；此外，企业长期拖延的ERP基础设施及Salesforce等平台现代化改造也亟待投入，这些问题正与AI成本上升叠加，对IT预算形成双重压力。

在投资回报评估方面，霍普金斯将AI的财务成果分为收入增长、效率与成本下降、风险管理改善三类，并对应短期生产力、中期参与度和长期转型三个时间维度。他指出，生产力最容易衡量，也是企业最常报告的AI收益，但其对利润的贡献排名靠后；参与度关注改善客户与员工同企业的互动方式，可提升交易频率和满意度；转型则涉及业务流程重塑、产品交付提速和新产品开发，风险更高但可形成持久竞争优势。仅聚焦生产力的企业可能错失参与度和转型层面的更大机会，而获取这些收益需要跨部门协调、共享优先级并约定成果指标。

霍普金斯还提出“信任税”概念，用以描述智能体AI带来的隐性成本。企业除自动化流程外，往往还需对智能体进行测试和红队演练、加装防护栏与监控、维护集中登记与治理审查，并在其与客户或合作伙伴交互时验证身份，由此产生额外开销，行业目前仍在探索合理支出水平。

在投资优先级上，霍普金斯认为应聚焦安全、数据和AI平台三大方向，但企业不能等到三者完全就绪才启动AI应用，尤其是在快速变革的行业。他建议企业先制定AI战略，明确主要业务挑战、竞争优势和预期成果，再转化为可衡量目标并建立投资组合，由业务高管对结果负责、组合经理负责资金安排。他以石油公司为例指出，若以每桶成本、产量和安全指标衡量绩效，AI投资就应与之挂钩，进而明确实现目标所需的基础工作。

展望未来，霍普金斯表示，在预计充满挑战的2027年，企业仍需在安全、负责任AI、治理、FinOps以及智能体访问和消费管控等方面持续投入。部分企业将认识到AI在参与度和转型方面的价值，能够构建商业论证并获得资金；仍将AI仅视为生产力工具的企业则可能陷入困境。他预计，不少企业或将收缩投入，而少数企业将凭借正确战略继续推进，这一分化对财务的全面影响可能要到2028至2029年才会充分显现。

中文翻译：

由谷歌云赞助
如何选择你的首批生成式人工智能应用场景
要开始使用生成式人工智能，首先应聚焦于能够改善人类信息体验的领域。
人工智能支出正在重塑企业IT预算。首席信息官如何管理不断上升的成本、确定投资优先级，并衡量超越生产力提升的回报。
人工智能投资的不断增长正促使企业重新审视IT预算。Forrester预测，到2027年，人工智能将占企业技术支出的约20%。这一估算收录于10月21日发布的报告中，其中包含技术供应商的基础设施支出；这并不意味着普通企业会将自身IT预算中同等比例的资金投入人工智能。
在本问答中，Forrester分析师布莱恩·霍普金斯探讨了这给企业预算带来的压力、首席信息官如何确定支出优先级，以及如何判断人工智能投资是否能带来持久价值。
是什么推动了人工智能支出的增长？
布莱恩·霍普金斯：几乎所有的增长都来自基础设施和算力——服务器、芯片、云服务、人工智能模型访问和软件订阅。
人工智能也在改变企业为软件付费的方式。传统上，组织可以估算一款应用的年度成本并据此编制预算。而有了生成式人工智能，成本更难预测。供应商正在引入基于token的消费模式、使用费和数据额度。而且智能体的成本会随任务复杂度和所需算力而变化。
首席信息官从哪里找钱来投资人工智能？
霍普金斯：企业软件支出的增长速度低于通胀率，购买力正在被侵蚀。
为弥补这一缺口，企业正在施压要求降低IT服务合同价格，理由是供应商应将人工智能驱动的效率提升让利给客户；同时将软件开发、技术支持和客户服务收回内部，以削减外部成本。
企业应避免削减哪些技术支出？
霍普金斯：安全显然是最明显的一项。各组织此前因为将安全视为成本中心而推迟了安全投资。但随着攻击者利用人工智能智能体以前所未有的规模发现并利用漏洞，风险正变得更加紧迫。
企业也一直没有对长期存在、成本高昂的ERP基础设施以及Salesforce等平台进行现代化改造。首席信息官多年来一直呼吁在这些领域投资，但企业往往推迟了这些决策。
这些长期积累的问题正在汇聚，而与此同时，人工智能成本的上升又给IT预算带来了额外压力。
企业在人工智能投资回报率方面存在哪些误区？
霍普金斯：我们看到三种财务成果——收入增长、效率提升/成本降低，以及更好的风险管理——它们被归入三个时间维度：生产力（短期）、互动（中期）和转型（长期）。
生产力最容易衡量。企业可以追踪人工智能是否减少了员工撰写客户邮件的时间，或加快了客服工单的处理速度。生产力是最常被提及的人工智能收益，但其对利润的贡献排名却低得多。
互动涉及改善客户和员工与企业的交互方式。例如，帮助客户更轻松地获得答案并完成交易，可以提高交易频率并提升满意度。
转型涉及重新思考业务流程、加速产品交付、创造新产品或在竞争对手之前进入市场。这类投资风险更大，但可以创造持久的竞争优势。
只关注生产力的企业可能会错失互动和转型中更大的机会。要获得这些收益，还需要跨部门协调，统一优先级并商定成果目标。
企业如何核算人工智能那些不太显眼的成本？
霍普金斯：我们将这些成本称为“信任税”，而智能体人工智能正将其推向前台。
企业可能以为智能体只需自动化某个流程，但往往还需要对其进行测试和红队演练、添加护栏和监控、维护带治理审查的集中注册表，并在其与客户或合作伙伴交互时验证其身份。
这些要求带来了额外的间接成本，行业仍在试图确定组织应预期支出多少。
首席信息官应优先在哪些方面投资人工智能？
霍普金斯：有三个主要投资优先级：安全、数据和人工智能平台。但首席信息官不能等到这三个领域全部到位才启动人工智能应用，尤其是在经历快速颠覆的行业中。
企业应首先制定人工智能战略，明确其主要业务挑战、竞争优势和期望成果。然后应将战略转化为可衡量的目标，并建立投资组合，由一位业务高管对结果负责，由一位组合经理负责资金安排。
例如，一家石油公司可能通过每桶成本、产量和安全性来衡量绩效。人工智能投资应与改善这些指标挂钩。一旦企业理解了某项投资将如何影响业务指标，就可以确定实现该成果所需的基础性工作。
企业会在没有明确回报的情况下继续增加人工智能预算吗？
霍普金斯：在很可能充满挑战的2027年，各组织需要在安全、负责任的人工智能、治理、FinOps以及智能体访问和消费管控方面进行投资。
一些企业将认识到人工智能通过互动和转型所能带来的价值。它们将能够构建商业论证并获得资金。而那些仍将人工智能仅视为又一种生产力工具的企业，则更可能陷入困境。
许多企业可能会收缩，而少数企业则会以正确的战略继续推进——这一分化带来的完整财务影响可能要到2028至2029年才会显现。
编者按：本访谈经过编辑以确保清晰和简洁。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
AI spending is reshaping enterprise IT budgets. How CIOs can manage rising costs, prioritize investments and measure returns beyond productivity gains.
Growing investment in AI is challenging enterprises to rethink IT budgets. Forrester forecasts that AI will account for about 20% of business technology spending in 2027. The estimate, included in an Oct. 21 report, includes infrastructure spending by technology vendors; it doesn't mean the average enterprise will devote that share of its own IT budget to AI.
In this Q&A, Brian Hopkins, an analyst at Forrester, discusses the pressures this is putting on enterprise budgets, how CIOs can prioritize spending, and how to judge whether AI investments deliver lasting value.
What's driving the increase in AI spending?
Brian Hopkins: Almost all of the growth comes from infrastructure and computing capacity — servers, chips, cloud services, AI model access and software subscriptions.
AI is also changing how companies pay for software. Traditionally, organizations could estimate an application's annual cost and budget accordingly. With generative AI, costs are harder to predict. Vendors are introducing token-based consumption, usage charges and credits. And agent costs vary with task complexity and the computing power required.
Where can CIOs find money for AI investments?
Hopkins: Enterprise software spending is rising more slowly than inflation, eroding purchasing power.
To compensate, companies are pushing for lower prices on IT services contracts, arguing providers should pass on AI-driven efficiency gains, and are bringing software development, support, and customer service in-house to cut external costs.
Where should enterprises avoid cutting technology spending?
Hopkins: Security is the obvious one. Organizations have delayed security investments because they view security as a cost center. But risks are becoming more urgent as attackers use AI agents to find and exploit vulnerabilities at a scale we haven't seen before.
Companies also haven't modernized longstanding, expensive ERP infrastructure and platforms such as Salesforce. CIOs have been asking for investment in these areas for years, but businesses have often delayed those decisions.
These longstanding issues are converging just as rising AI costs put additional pressure on IT budgets.
What are companies getting wrong about the ROI of AI?
Hopkins: We see three financial outcomes — revenue growth, efficiency/cost reduction, and better risk management — organized into three time horizons: productivity (short term), engagement (medium term), and transformation (long term).
Productivity is easiest to measure. Companies can track whether AI reduces the time employees spend writing customer emails or speeds up customer service case resolution. Productivity is the most commonly reported AI benefit, yet its bottom-line impact ranks much lower.
Engagement involves improving how customers and employees interact with a business. For example, helping customers get answers and complete transactions more easily can increase transaction frequency and improve satisfaction.
Transformation involves rethinking business processes, accelerating product delivery, creating new products or entering markets ahead of competitors. These investments carry more risk but can create a lasting competitive advantage.
Companies that focus only on productivity risk missing the larger opportunities in engagement and transformation. Capturing those benefits also requires coordination across departments, with shared priorities and agreed-upon outcomes.
How are companies accounting for AI's less visible costs?
Hopkins: We call those costs the "trust tax," and agentic AI is bringing them to the forefront.
Companies might expect an agent to simply automate a process but often need to also test and red-team it, add guardrails and monitoring, maintain a centralized registry with governance review, and verify its identity if it interacts with customers or partners.
These requirements create additional overhead, and the industry is still trying to determine how much organizations should expect to spend.
Where should CIOs prioritize AI investments?
Hopkins: There are three main investment priorities: security, data and the AI platform. But CIOs can't delay AI adoption until all three areas are fully addressed, especially in industries undergoing rapid disruption.
Companies should start with an AI strategy that identifies their major business challenges, competitive advantages and desired outcomes. They should translate that strategy into measurable targets and create an investment portfolio, with a business executive accountable for results and a portfolio manager responsible for funding.
For example, an oil company might measure performance through cost per barrel, production and safety. AI investments should be tied to improving those metrics. Once a company understands how an investment will affect a business metric, it can identify the foundational work required to achieve that outcome.
Will companies continue to increase AI budgets without clear returns?
Hopkins: Organizations need to invest in security, responsible AI, governance, FinOps and controls over agent access and consumption in what is likely to be a challenging 2027.
Some companies will recognize the value AI can deliver through engagement and transformation. They'll be able to build business cases and secure funding. Companies that remain focused on AI as just another productivity tool are more likely to struggle.
Many companies may pull back, while a smaller group pushes forward with the right strategy — a divide whose full financial impact likely won't be visible until 2028–2029.
Editor’s note: This interview has been edited for clarity and conciseness.