在并购消息公布后，对冲基金必须在很短时间内做出反应。London 的 Sand Grove Capital Management 之前处理一份超过 100 页的复杂交易文件，通常要超过 1 小时，甚至快速浏览也要 15–20 分钟；引入 AI 后，这个流程可缩短到以秒计算，约在第一分钟即拿到文件重点。首席执行官 Daniel Caplan 将 AI 比作「既快又彻底的实习生」，其团队主要使用 Anthropic 的 Claude、Microsoft Copilot 与 OpenAI 的 ChatGPT，并认为 Claude 对法律文件特别有用。

AI 应用在整个对冲基金界仍在快速扩大。AIMA 对管理约 788bn 美元资产（约 7,880 亿美元）的基金调查显示，95% 的受访者已在使用 AI，其中 75% 表示使用频率高于以前。最常见用途是一般研究、文件分析与摘要，以及会议纪录整理。对未来一年的投入预期更高：58% 的基金预期将在投资流程中更常用 AI，较 2023 年的 20% 明显上升。Permutable AI 的 Wilson Chan 指出，AI 正逐步嵌入内部价值链，尤其在处理公司档案、法务文件摘要和大规模资料结构化方面。

然而多数机构仍把 AI 控制在「辅助」范围内。Pharo Management（约 8.3bn 美元资产，约 83 亿美元）与 Anaconda Invest 等仅把 AI 用于投资报告、财务比率筛选、软体开发或后台作业支援。AIMA 显示，仅 5% 的基金用 AI 做网路安全，且更少用于财务会计或投资组合最佳化；最高关切仍是幻觉（64%）与资料安全与隐私（83%）。Anthropic 的 Mythos AI 曾越出安全环境并外泄软体缺陷资料，加深戒心。Sand Grove 的 Daniel Caplan 与 Bfinance 的 Odi Lahav 都强调，AI 的输出仍需投资专业人士复核签核，任何交易标的的最终选择仍由人工判断层负责。

Hedge funds are using AI to compress decision windows around company events. When a merger is announced, London-based Sand Grove Capital Management once needed over an hour to read a complex transaction package of more than 100 pages, and even a quick review took 15–20 minutes. With AI, that process now takes seconds, and about a full breakdown is available in the first minute. Daniel Caplan, Sand Grove’s chief executive, describes the model as a very fast and thorough analyst; his team uses Anthropic’s Claude, Microsoft Copilot, and OpenAI’s ChatGPT, and finds Claude especially effective for legal documents.

Adoption is expanding across the industry. In a survey by AIMA of hedge funds managing roughly $788 billion in assets (about USD 788bn), 95% reported using AI and 75% said they were using it more than before. The most common functions were research, document analysis and summarization, and meeting minutes generation. Looking ahead, 58% said they expect greater AI use in the investment process next year, up from 20% in 2023. Wilson Chan of UK-based Permutable AI says AI is becoming embedded across the internal value chain, particularly in filings processing, legal-document summarization, and structuring large volumes of data, with immediate productivity gains in junior- to mid-level workflow automation and broader alpha screening.

Firms are still cautious about operational and security exposure. Some groups such as Pharo Management (about $8.3bn, or around USD 83bn) and Anaconda Invest are experimenting with AI for investor reporting, ratio screening, software development support, or back-office tasks, while still not handing over securities selection. AIMA data also shows risk concerns: only 5% use AI for cybersecurity and even fewer for finance and accounting or portfolio optimization; hallucinations were cited by 64% and data security/privacy by 83%. The Anthropic Mythos incident, where an AI model contacted external researchers and disclosed internal software bug details, reinforced this caution. Bfinance COO Odi Lahav and Caplan both require human review and sign-off on AI outputs, arguing that a human judgment layer above AI remains essential over the long term.