# AI Agents Cut 3-Day Portfolio Builds to Minutes

Michael Manapat, Rowspace & Jack McDonald, Apollo / Redding Ridge | AI Engineer NYC 2026 | 18:57

Source: https://www.youtube.com/watch?v=BOWYujhIQuo
Channel: AI Engineer (https://www.youtube.com/@aiDotEngineer). Summarised by AIE Talks.
Page: https://aietalks.com/talks/ai-agents-cut-3-day-portfolio-builds-to-minutes
Published: 2026-10-16
Tags: data-pipelines, enterprise, finance

## TL;DR
- Rowspace helps Redding Ridge reconcile fragmented financial data by finding the correct source and applying the firm's own calculations.
- AI reduced mock portfolio construction from two or three days of Excel work to minutes, leaving portfolio managers to refine the final part using their judgment.
- Finance is still early in AI adoption because its workflows contain private data, complex rules, and many sources that must be checked before decisions can rely on them.

## Summary
Jack McDonald describes how Redding Ridge, Apollo's structured credit and CLO manager, uses Rowspace to work with fragmented investment data. CLO portfolio construction once required portfolio managers to read legal documents, build Excel models, and assemble portfolios of 200 to 400 credit names over two or three days. The system now creates a real-time portfolio in minutes, after which the manager refines it. Redding Ridge also automated a weekly trade reconciliation that required two or three people to check bank statements, settlement information, and trading activity. McDonald explains why a general-purpose LLM is insufficient for this work. The firm needs its own data, calculations, and source hierarchy because one fact may appear in 20 systems and only one may be correct. He says finance remains early in AI adoption, especially in complicated asset-management workflows. His advice is practical: engineers need to spend time with front-office teams and understand why unusual processes exist before automating them.

## Key ideas
### AI becomes useful when it can resolve the firm's fragmented data
[01:39](https://www.youtube.com/watch?v=BOWYujhIQuo&t=99s)
Redding Ridge receives thousands of data points each day across many investments, including loan pools and broadly syndicated loans. The information is private and spread across documents, systems, and legal agreements. Jack McDonald says Rowspace works directly with the teams to understand the calculations, identify the golden source of truth, and make the data consistent enough for fast decisions. The work is more than placing a model on top of existing information. It includes learning which source is correct and repeating the firm's own rules.

### Portfolio construction fell from two or three days to minutes
[04:03](https://www.youtube.com/watch?v=BOWYujhIQuo&t=243s)
CLO portfolio construction depends on detailed rules and waterfalls contained in indentures and other legal agreements. Before Rowspace, a portfolio manager reviewed those documents, built Excel models, and assembled a mock portfolio with roughly 200 to 400 individual credit names. That process took two to three days. The system now pulls current information from document libraries, trade and compliance systems, and real-time pricing sources to create a portfolio in minutes. The manager can then apply judgment to refine the remaining part of the portfolio.

### Trade reconciliation became a daily quality check instead of manual weekly work
[05:43](https://www.youtube.com/watch?v=BOWYujhIQuo&t=343s)
Redding Ridge used to reconcile weekly purchases and sales by checking bank accounts and settlement statements, then sending a recap to management. Two or three people spent many hours each week completing the work. The team now drops the same report into Rowspace, where the report is created and repeatable in about a minute. Staff still check the output each day, but their role is quality assurance rather than manually assembling the reconciliation.

### A general LLM does not know which internal source is correct
[07:20](https://www.youtube.com/watch?v=BOWYujhIQuo&t=440s)
McDonald says Claude, ChatGPT, and other LLMs are useful for general research, new ideas, and quickly learning about an investment or product. They do not automatically contain the firm's own data. At Redding Ridge, one data point may come from 20 different sources while only one is right. Rowspace works with portfolio management and finance teams to understand the calculations, validate the correct information, and override conflicting sources from systems such as trade order management, performance, and pricing platforms.

### Apollo adopted AI through a culture of changing workflows
[09:39](https://www.youtube.com/watch?v=BOWYujhIQuo&t=579s)
McDonald says Apollo has long presented itself as an innovative alternative asset manager. Its management approach emphasizes upgrading team capabilities, adopting new technology, and freeing people from manual processes so they can spend more time thinking. He describes the firm as flat and open to change. That attitude led Apollo to treat AI as something to bring into workflows rather than something to dismiss.

### Finance is still early because its work is complicated
[11:59](https://www.youtube.com/watch?v=BOWYujhIQuo&t=719s)
McDonald places finance, especially asset management and investment banking, early in AI adoption. Large institutions have substantial middle- and back-office operations with room for improvement, but their processes are not straightforward. Today, AI is useful for tasks such as creating models, aggregating research, and handling routine work. McDonald expects some junior support work to decline over time, while current deployments mainly free capacity for other projects.

### Accurate internal information can become a lasting advantage
[14:03](https://www.youtube.com/watch?v=BOWYujhIQuo&t=843s)
McDonald believes AI can improve competitiveness when a firm uses it with a strong process and reliable information. In actively traded markets, better access to liquidity, pricing, and value helps teams make decisions. In private markets, he says this advantage is still developing. The risk is that incorrect information can lead to a bad decision and pass errors through the enterprise and to investors. Trust in the data is therefore part of the advantage, not a separate concern.

### Engineers need to work beside the front office
[17:23](https://www.youtube.com/watch?v=BOWYujhIQuo&t=1043s)
McDonald's advice to builders is to spend time on the ground with the team that will use the tool. Finance organizations contain unusual structures and processes that may look unnecessary until their reason becomes clear. He credits Michael Manapat and the Rowspace team with spending time alongside Redding Ridge staff, learning their workflows, and getting ahead of the product roadmap. Understanding the work directly is part of building a useful system.

## Notable quotes
- Jack McDonald: "Today, we're able to kind of get the latest and greatest information that's real time, create a portfolio that would take somebody days to now do it in the matter of minutes." (04:44)
- Jack McDonald: "For one piece of data that might come from 20 different sources and there's only one that's actually right." (08:22)
- Jack McDonald: "I think finance broadly, especially in asset management, investment banking, the bank institution institutional world, I still think it's very early." (11:59)
- Jack McDonald: "Spending the time and really putting boots on the ground to sit with whatever team or front office professional that you're working to build a tool for or enhance is time well spent." (18:02)

## Tools & references mentioned
- Rowspace
- Apollo
- Redding Ridge
- Claude
- ChatGPT
- Bloomberg

## Who should watch
- You are building AI for asset managers, lenders, or other finance teams where the same data appears in many internal systems.
- Your team is automating a workflow that still depends on Excel, email, bank statements, or manual reconciliation.
- You need a practical account of why engineers have to work closely with portfolio managers and other front-office users.

## Related talks

- [AI Agents Will Upend Investing Even More Than Coding](https://aietalks.com/talks/ai-agents-will-upend-investing-even-more-than-coding) (Frank Long, Coatue, 14:05)
- [Proving Alpha: How a Hedge Fund Measures AI's ROI](https://aietalks.com/talks/proving-alpha-how-a-hedge-fund-measures-ais-roi) (Hari Kumar, Contour Asset Management, 18:24)
- [How Agents Cut Hour-Long Data Checks to 5 Minutes](https://aietalks.com/talks/how-agents-cut-hour-long-data-checks-to-5-minutes) (Amol Kapoor & Clifford Ressel, Nori Agentic, 19:34)
- [Why Frontier Models Fail as Buy-Side Analysts](https://aietalks.com/talks/why-frontier-models-fail-as-buy-side-analysts) (Janie Gu, Google DeepMind & Suyeol Yun, LinqAlpha, 18:32)
- [Redesigning How Software Gets Built With AI Agents](https://aietalks.com/talks/redesigning-how-software-gets-built-with-ai-agents) (Tariq Shaukat & Ali-Reza Adl-Tabatabai, Sonar, 19:12)
