AI 智能体将投资组合构建从 3 天缩短至数分钟：Apollo 的 Jack McDonald 与 Rowspace 的 Michael Manapat

简要说明Brief Description

Rowspace 联合创始人 Michael Manapat 与 Apollo 的负责人、Redding Ridge 首席运营官 Jack McDonald，探讨了 AI 如何帮助结构化信贷业务整合分散的数据、在几分钟内构建投资组合，并自动生成交易报告。他们还解释了在 AI 辅助投资决策时，为什么可靠的源数据、融入业务流程的工程工作，以及对底层计算结果的信心至关重要。Rowspace co-founder Michael Manapat and Jack McDonald, a principal at Apollo and COO of Redding Ridge, discuss how AI helps a structured-credit business reconcile fragmented data, build portfolios in minutes, and automate trade reporting. They explain why reliable source data, embedded engineering work, and confidence in the underlying calculations matter when AI supports investment decisions.

目录Table of Contents

• 罗氏空间与雷丁岭
  Rowspace and Redding Ridge
• 让 AI 成为金融领域的关键任务技术
  Making AI mission-critical in finance
• 投资组合构建：从数天缩短至数分钟
  Portfolio construction: days to minutes
• 自动化交易对账
  Automating trade reconciliation
• 建立超越通用聊天机器人的信任
  Building trust beyond a general-purpose chatbot
• Apollo 的 AI 采用方法
  Apollo’s approach to AI adoption
• 金融业在采用 AI 方面处于什么阶段？
  How early is finance in AI adoption?
• 生产力工具，还是持久的竞争优势？
  Productivity tool or durable competitive advantage?
• 下一步需要改进的方面
  What needs to improve next
• 为将 AI 应用于金融领域的开发者提供建议
  Advice for builders applying AI to finance

罗空间与雷丁岭Rowspace and Redding Ridge

Michael Manapat：我是 Michael Manapat，Rowspace 的联合创始人。我们利用 AI 帮助 Apollo 等金融服务合作伙伴处理杂乱、零散的数据。我们的智能体负责数据工程工作：从不同系统中提取信息并核对数据，让投资组合经理和交易员能够更快、更准确地做出决策。今天和我一起的是 Jack McDonald，Redding Ridge 的首席运营官。Redding Ridge 是 Apollo 旗下的一家信贷业务公司。Jack，能请你介绍一下自己和这项业务吗？Michael Manapat: I’m Michael Manapat, a co-founder of Rowspace. We use AI to help financial-services partners such as Apollo deal with messy, fragmented data. Our agents do data-engineering work: extracting information and reconciling data across systems so portfolio managers and traders can make decisions faster and more accurately. I’m joined by Jack McDonald, chief operating officer of Redding Ridge, one of Apollo’s credit businesses. Jack, could you tell us about yourself and the business?

Jack McDonald：我是 Apollo 的负责人，也是 Redding Ridge Capital 的首席运营官。Redding Ridge Capital 是我们的结构化信贷和 CLO 管理公司。我们是一家独立运营公司，在结构化信贷领域管理着约 500 亿美元的资产，主要是 CLO。自 2020 年起，我一直在 Apollo 工作，先后供职于金融机构团队和资产支持团队，之后转任目前在 Redding Ridge 的职务。Jack McDonald: I’m a principal at Apollo and COO of Redding Ridge Capital, our structured-credit and CLO manager. We are a separate operating company managing about $50 billion in assets across the structured-credit ecosystem, primarily CLOs. I’ve been at Apollo since 2020, working across our financial-institutions group and asset-backed group before moving into my current role at Redding Ridge.

让 AI 成为金融领域的关键任务技术Making AI mission-critical in finance

Michael Manapat：我们发现，AI 从锦上添花到成为业务关键，之间存在着巨大的鸿沟。在 Apollo，AI 可以改变公司的运营方式。哪些工作促成了这一转变？这些工作复杂在哪里？你们现在又看到了什么？Michael Manapat: We see a wide gap between AI being nice to have and becoming mission-critical. At Apollo, it can transform how the firm operates. What kinds of work made that possible, what made the work complex, and what are you seeing now?

Jack McDonald：在 Apollo，尤其是 Redding Ridge，我们管理着许多基金实体和投资。每天都有数千个数据点从各项投资中传来。我们管理着包含约一千个借款方的贷款组合，并每天评估多个业务领域的信用风险。CLO 管理是一项积极的业务，我们会在广泛银团贷款市场进行交易，因此随时掌握所需信息至关重要。Jack McDonald: At Apollo, and specifically Redding Ridge, we manage many fund entities and investments. Thousands of data points arrive daily across our investments. We manage loan pools with roughly a thousand names and assess credits across multiple businesses every day. CLO management is active, with trading in the broadly syndicated loan market, so having the information at your fingertips is essential.

我们希望数据准确、一致，并采用可用的格式，以便在快节奏的环境中做出审慎的决策。与公开股票或公开固定收益领域可获取的信息不同，我们掌握的大量信息具有私密性，且分散在各种文档中。We want our data to be accurate, consistent, and in a usable format so we can make thoughtful decisions in a fast-paced environment. Much of the information is private and fragmented across documents, unlike the information available for public equities or public fixed income.

我们与 Rowspace 这类公司合作，他们会深入参与我们的团队，了解相关计算、我们需要的信息，以及哪个来源才是权威数据源。随后，他们会改进 AI 系统，使其能够理解并重复这些工作，从而提升我们的效率。We partner with companies such as Rowspace that sit with our teams and understand the calculations, the information we need, and which source is the source of truth. They then enhance the AI systems to understand and repeat that work, making us more efficient.

投资组合构建：从数天缩短至数分钟Portfolio construction: days to minutes

Michael Manapat：现场有很多工程师。我们刚才聊到，以前做一次消防演练可能要花团队一周甚至更久。你能举一个具体例子吗？说说过去哪些工作既费劲又耗时，而现在借助 Rowspace 和 AI 变得更轻松了？Michael Manapat: There are many engineers in the audience. We’ve talked about fire drills that used to take a team a week or more. Could you give a concrete example of work that was painful or time-consuming and is now easier with Rowspace and AI?

杰克·麦克唐纳：一个例子是投资组合构建。CLO 依照规则运作，并设有特定的测试和瀑布分配机制，规定资金流向。这些规则深藏在契约及其他法律协议中。Jack McDonald: One example is portfolio construction. CLOs are rules-based and have specific tests and waterfalls governing where money flows. Those rules live deep in indentures and other legal agreements.

投资组合经理会审阅文件、建立 Excel 模型，并构建一个包含约 200 至 400 个独立信用主体的模拟投资组合。在采用 Rowspace 之前，这项工作可能需要在 Excel 中手动完成，耗时两到三天。如今，我们可以利用最新信息，在几分钟内构建投资组合。A portfolio manager would review the documents, build Excel models, and construct a mock portfolio with around 200 to 400 individual credit names. Before adopting Rowspace, that could take two to three days of manual work in Excel. Today we can use current information to construct a portfolio in minutes.

这样就能使投资组合达到 80% 或 90% 的完成度。随后，投资组合经理会运用自己的判断完善剩余部分，并融入个人风格。这一过程涉及从文档库、交易与合规系统以及实时定价来源中提取信息。That gets the portfolio to 80% or 90% completion. The portfolio manager then applies judgment to refine the last part and make it their own. It involves pulling information from document libraries, trade and compliance systems, and real-time pricing sources.

自动化交易对账Automating trade reconciliation

Jack McDonald：另一个例子是仓储贷款的购买活动。我们利用银行融资，也可以随时在固定收益市场购买贷款。过去，我们每周都会核对买入和卖出情况：买了什么、卖了什么，以及实际完成结算的交易。Jack McDonald: Another example is warehouse purchase activity. We use bank financing and can buy loans in the fixed-income market at any point. Historically, we produced a weekly reconciliation of purchases and sales: what we bought, what we sold, and what actually settled.

这件事非常依赖人工操作。团队需要查看银行账户和对账单，核对交易记录，然后通过邮件向管理层发送报告。我们用 Rowspace 重现了同一份报告。现在大约一分钟就能生成报告，并再次执行这一流程。It was very manual. A team went into bank accounts and statements, reconciled the activity, and emailed management a report. We took that same report and reproduced it in Rowspace. It can now be generated and repeated in about a minute.

整个流程是自动化的，我们的团队每天都会检查，确保结果看起来正确。过去需要两三个人每周耗费许多小时的工作，如今已成为一项例行的质量保证流程。The process is automated, and our team checks it each day to ensure it looks right. Work that used to consume many hours a week for two or three people has become a routine quality-assurance process.

建立超越通用型聊天机器人的信任Building trust beyond a general-purpose chatbot

Michael Manapat：这些例子已经不只是阅读聊天机器人的输出以获取信息了。你们团队需要做出交易和投资组合管理决策，因此准确性至关重要。为什么公司不能直接把这些信息输入 Claude 或 ChatGPT？你们团队如何才能对 Rowspace 这样的产品建立足够的信心，并据此做出业务决策？Michael Manapat: These examples go beyond reading chatbot output for information. Your team makes trading and portfolio-management decisions, so accuracy matters. Why can’t a firm simply put this information into Claude or ChatGPT? How does your team gain enough confidence in a product such as Rowspace to rely on it for business decisions?

Jack McDonald：Claude、ChatGPT 和其他大语言模型有助于开展一般性研究、探索想法，或快速了解某款产品或一项投资。但这并不意味着它们掌握了贵公司的内部数据。Jack McDonald: Claude, ChatGPT, and other LLMs are useful for general research, exploring ideas, or quickly understanding a product or investment. But that does not mean they have your firm’s own data.

我们在 Redding Ridge 的做法是掌握自己的数据，并在此基础上采用我们确信使用准确信息的技术。Rowspace 与我们合作过的其他科技公司不同，因为他们的团队会与我们并肩工作。Michael 每天都和我们的团队坐在一起，了解我们的信息流。一条数据可能来自 20 个来源，但其中只有一个是正确的。Our approach at Redding Ridge is to own our data and have technology on top of it that we know is using accurate information. Rowspace is different from other technology companies we’ve worked with because its team is on the ground with us. Michael sits with our team daily, learning our information flows. A single data point might come from 20 sources, with only one being correct.

团队与我们的投资组合管理团队和财务团队合作，了解计算方式、确定正确的数据、必要时覆盖其他数据源，并汇总结果。我们的团队会验证这些信息，从而建立对系统的信任。The team works with our portfolio-management and finance teams to understand calculations, identify the right data, override other sources when necessary, and aggregate the results. Our own teams validate the information, which creates trust in the system.

我们可以知道这些信息来自我们的交易订单管理系统、绩效系统、Bloomberg，还是其他定价指数。在许多现有系统中，包括我们公司内部，信息都是碎片化的：我们必须从一个地方取出一部分，再手动与另一处的信息整合。Rowspace 会自动完成这项工作。We know whether the information comes from our trade-order-management system, performance system, Bloomberg, or another pricing index. In many existing systems, including within our company, information is fragmented: we have to take a piece from one place and manually combine it with another. Rowspace automates that work.

Apollo 的 AI 应用方法Apollo’s approach to AI adoption

迈克尔·马纳帕特：早在去年年初，Claude Code 尚未问世、Cursor 还专注于 Tab 补全时，Apollo 在 AI 领域的做法似乎就已领先于许多公司。这包括如何在 AI 时代进行投资，以及如何围绕 AI 原生平台重建运营，而不是采用单点解决方案。这种理念在 Apollo 的企业文化中源自何处？Michael Manapat: Even early last year, before Claude Code and while Cursor was still focused on tab completion, Apollo seemed ahead of many firms in its approach to AI. That included how to invest in an AI world and how to rebuild operations around an AI-native platform rather than point solutions. Where did that come from in Apollo’s culture?

杰克·麦克唐纳：阿波罗始终致力于在另类资产管理和金融服务领域引领行业。我们以创新为荣，并始终走在变革前沿。Jack McDonald: Apollo has always aimed to lead in alternative asset management and financial services. We pride ourselves on innovation and being at the forefront of change.

我们的管理团队致力于提升员工能力，并在市场变化时保持灵活应变。我们意识到人工智能已经到来，也认识到我们需要适应并采用它。我们一直积极将新技术融入工作流程，提高工作效率，并取代手动流程，让员工有更多精力进行思考。Our management team focuses on improving people’s capabilities and remaining nimble as markets change. We recognized that AI was here and that we needed to adapt and adopt it. We have been open to bringing new technology into workflows, making work more efficient, and replacing manual processes so people have more capacity to think.

从本质上说，Apollo 是一家创新型公司，拥有扁平、开放且乐于接受变革的文化。At its core, Apollo is an innovative firm with a flat, open culture that is receptive to change.

金融业在采用 AI 方面处于多早期的阶段？How early is finance in AI adoption?

Michael Manapat：软件工程已经发生了翻天覆地的变化，而法律和客户服务领域可能也即将迎来类似的转变。金融业在这一进程中处于什么位置？人工智能是否已经带来了重大影响，还是说我们仍处于起步阶段？如果还处于早期阶段，要改变金融业的格局，还需要发生什么？Michael Manapat: Software engineering has changed completely, and law and customer service may be approaching a similar point. Where is finance on that spectrum? Has AI already made a major difference, or are we still at the beginning? If it is early, what needs to happen to change the shape of finance?

Jack McDonald：银行和大型金融机构规模庞大，拥有大量中台和后台支持人员。这些工作还有很大的改进空间。正如你在与我们合作时所看到的，我们的运营十分复杂，因此问题很少是简单明了的。Jack McDonald: Banks and large financial organizations are massive, with substantial middle- and back-office support. There is a lot of room to improve that work. As you’ve seen in working with us, our operations are complicated, so the problems are rarely straightforward.

在资产管理、投资银行和机构金融领域，我认为目前仍处于非常早期的阶段。AI有助于建模、快速了解一家公司以及汇总研究资料，但在我们看来，这些用途目前的价值相对较低。In asset management, investment banking, and institutional finance, I think it is still very early. AI helps with model creation, quickly understanding a company, and aggregating research, but from our perspective those uses are relatively lower-value applications today.

这种情况会随着时间推移而变化，有些职能对初级员工支持的需求可能会逐渐减少。目前，我们正在引入这项技术，尽可能提高日常任务的效率。这样就能腾出人手投入其他项目；我们有几乎无穷无尽的改进事项。这也让我们能更灵活地开展这些工作。That will evolve, and some functions may require less junior support over time. For now, we are adopting the technology to make routine tasks as efficient as possible. That frees capacity for other projects; we have an almost endless list of things to improve. It gives us more flexibility to pursue that work.

生产力工具，还是持久的竞争优势？Productivity tool or durable competitive advantage?

Michael Manapat：AI 可能成为一种加快任务处理、降低运营开支的工具，就像 Excel 一样实用，但本身并不能让一家公司脱颖而出。它也可能创造持久的竞争优势，进一步巩固阿波罗在交易结构设计和创新交易方面的优势。你认为哪种结果更有可能？Michael Manapat: AI could become a tool that speeds up tasks and reduces operating expense, much as Excel is useful but does not by itself differentiate a firm. It could also create a lasting competitive advantage, reinforcing Apollo’s strengths in structuring and creative deals. Which outcome do you expect?

杰克·麦克唐纳：如果得到妥善采纳和使用，AI 将提升公司的竞争力。在资产管理和投资领域，尤其是在交易活跃的市场中，流程和信息获取至关重要。Jack McDonald: If adopted and used properly, AI will make a firm more competitive. In asset management and investing, especially in actively traded markets, process and access to information matter enormously.

你需要拥有强大的交易执行流程，并清楚了解任何时点可获得的流动性、定价和价值。这会让你占据优势。这些能力在活跃交易的市场中已经具备，但在私募市场中仍在不断发展，而我们一直致力于推动 Apollo 在这方面向前迈进。You need a strong execution process and a clear view of available liquidity, pricing, and value at any given time. That gives you an edge. These capabilities exist in actively traded markets, but they are still evolving in private markets, where we have been working to move Apollo forward.

随着时间推移，更便捷地获取信息有助于企业进一步扩大领先优势。团队还必须确信这些信息准确且一致。我们最不愿看到的，就是决策建立在错误信息之上，继而在整个企业内部传播，最终影响投资者。Over time, better access to information can help firms pull further ahead. Teams must also know the information is accurate and consistent. The last thing we want is a decision based on misinformation that then flows through the enterprise and ultimately to investors.

接下来需要改进的地方What needs to improve next

Michael Manapat：要预测五年后的情况似乎有失公平，毕竟就连预测我们下周会用哪些编程工具都很难。不妨把目光放在接下来的几个月：你希望自己的工作、团队以及 Apollo 对 AI 的使用有哪些变化？具备哪些条件，这项技术才能真正发挥价值？Michael Manapat: Predicting five years ahead seems unfair when it is difficult even to predict which coding tools we will use next week. Looking at the next few months instead, what do you hope to see in your job, team, and Apollo’s use of AI? What conditions would make the technology truly valuable?

Jack McDonald：首先，我们完全确信提供给投资者的信息准确无误，无需花费数小时手动核查。我们的投资者遍布全球，人数超过 400，因此确保分享的信息可靠且没有错误至关重要。Jack McDonald: First, full confidence that information going to our investors is accurate, without spending hours checking it manually. We have more than 400 investors globally, so it matters that we share reliable information without mistakes.

其次，我希望我们拥有更具技术赋能意识的思维方式。我们的团队并不缺乏这种意识，但有时会把技术放在次要位置。随着日常任务的效率不断提高，人们就有更多余力进行创造性思考：流程为什么存在？目前是如何运作的？我们能否改进或改变它，或者引入不同的技术？Second, I want a more technology-enabled mindset. Our team has not lacked that mindset, but technology has sometimes been an afterthought. As routine tasks become more efficient, people have more capacity to think creatively: Why does a process exist? How does it work today? Could we improve it, change it, or bring in a different technology?

理想情况下，随着我们 Redding Ridge 团队解决问题并扩大组织规模，这应成为首要考量，而不是次要考虑。Ideally, that becomes a primary consideration as our team at Redding Ridge tackles problems and scales the organization, rather than remaining a secondary thought.

为将 AI 应用于金融领域的开发者提供建议Advice for builders applying AI to finance

Michael Manapat：与你共事让我了解到，这项工作既复杂又需要创造力，也让我看到要让一切顺利运转，需要投入多少人力和 AI 的努力。对于想从事技术或金融领域工作的人来说，在像你这样的领域应用 AI 时，他们应该牢记些什么？Michael Manapat: Working with you has shown me how complex and creative this work is, and how much human and AI effort it takes to make things work. For people who want to work on either the technology or finance side, what should they keep in mind when applying AI to a domain like yours?

杰克·麦克唐纳：这些组织极其复杂。看起来奇怪或格格不入的事物，可能是出于某种原因才这样安排的。我们的运营情况十分复杂，因此，花时间与团队成员或你正在为其开发系统的前台专业人员坐下来交流，是很值得的。Jack McDonald: These organizations are extremely complex. Something that looks strange or out of place may be structured that way for a reason. Our operations are nuanced, so time spent sitting with the team or front-office professional you are building for is time well spent.

我们在 Rowspace 上取得的成功，以及进度超前于路线图和预期，都得益于 Michael 和团队几乎每天都到现场与我们共事。了解我们的工作流程至关重要。The success we have had with Rowspace, and the fact that we are ahead of our road map and expectations, comes from Michael and the team spending time with us on the ground almost every day. Understanding our workflows is crucial.

Michael Manapat：Rowspace 和 Apollo Technology 都在招聘希望将 AI 应用于金融领域的人才。如果你感兴趣，欢迎联系我或 Jack。Jack，谢谢你，和你聊天很愉快。Michael Manapat: Rowspace and Apollo Technology are both hiring people who want to apply AI to finance. Reach out to either of us if you are interested. Jack, thank you; it has been great chatting.

杰克·麦克唐纳：这次交流很愉快。谢谢。Jack McDonald: This has been great. Thank you.