AI Agents Cut 3-Day Portfolio Builds to Minutes — Jack McDonald, Apollo & Michael Manapat, Rowspace
Loading slides…
Read the talk
AI Agents Cut 3-Day Portfolio Builds to Minutes — Jack McDonald, Apollo & Michael Manapat, Rowspace
Redding Ridge uses AI to bring fragmented credit data into portfolio construction and trade reconciliation. The harder work is deciding which information to trust and keeping people responsible for the decisions.
From a talk by Michael Manapat and Jack McDonald
At a glance
Ideas worth remembering
McDonald says a 200–400-name mock CLO portfolio that once took two to three days in Excel can now be prepared in minutes for a manager to refine.
Automated trade reconciliation recreates a report in about a minute, while the team checks its output daily.
Trusted financial data requires resolving competing sources and validating calculations with the people who use them.
Time saved on routine tasks can give teams room to question and improve the processes themselves.
Credit decisions start with fragmented data
A portfolio manager at Redding Ridge works with loan information scattered across private documents and systems while making decisions in an active market. Jack McDonald, an Apollo principal and Redding Ridge’s COO, says the structured credit and collateralized loan obligation (CLO) manager oversees about $50 billion in assets and receives thousands of investment data points daily.
Michael Manapat, co-founder of Rowspace, describes its platform as using agents to reconcile information across systems. The immediate challenge is to make the data accurate, consistent and usable. That requires understanding Redding Ridge’s calculations and which source should supply each piece of information.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Build a mock portfolio in minutes
A CLO portfolio must satisfy specific tests and follow a defined waterfall for distributing money. Those rules live in lengthy indentures and other legal agreements. Before using Rowspace, a portfolio manager reviewed the documents, built Excel models and assembled a mock portfolio of roughly 200 to 400 individual credit names. McDonald says that manual work typically took two to three days.
The newer workflow brings together information from document libraries, trade and compliance systems, and real-time pricing. McDonald says it can produce a portfolio in minutes, at roughly 80–90% completion, leaving the manager to apply judgment and refine it. These are his operational estimates; the discussion does not provide an accuracy measurement or a controlled comparison.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Turn a weekly reconciliation into a repeatable report
Redding Ridge also uses bank financing to buy loans. Its team historically produced a weekly recap of purchases and sales by checking bank accounts and statements to see where trades had actually settled, reconciling the activity, and emailing management. McDonald says the process consumed many hours a week across two or three people.
The same report is now represented in Rowspace and can be recreated in about a minute, according to McDonald. The team checks the automated output daily to see whether it looks right. The change removes repeated manual assembly while retaining a review of the result.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Choose the right source before relying on the answer
A faster report matters only if its inputs support trading and portfolio decisions. McDonald finds general-purpose models useful for research and getting up to speed on an investment, but describes Redding Ridge’s need as owning its data and building a trusted platform around it.
One data point might appear in 20 sources, with only one right for the task. Rowspace works alongside portfolio management and finance staff to understand calculations, identify the correct information and override competing sources. Redding Ridge’s team validates the information it expects to rely on.
Those sources include trade order management and performance systems, Bloomberg and pricing indexes. Previously, staff often had to lift pieces from separate systems and join them manually; Rowspace automates that aggregation. The discussion describes the source-selection work and human validation, but does not specify the implementation used to resolve each conflict.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
From routine work to a possible competitive edge
Manapat describes Apollo’s early interest in an AI-native platform rather than isolated tools. McDonald connects that approach to an openness to changing workflows, improving staff capabilities and replacing manual processes so people have more time to think.
McDonald sees adoption across asset management, investment banking and large financial institutions as early. Research aggregation and model creation already have value, and he expects some functions may need less junior support over time. At Redding Ridge, the present focus is more immediate: make routine tasks efficient and free capacity for a long list of other projects.
Could AI become as commonplace as Excel, or provide a lasting edge? McDonald expects an advantage when a firm combines good execution processes with information its team trusts: available liquidity, current pricing and value. In his view, better information can help firms move ahead, especially as private markets evolve. A misinformed decision, meanwhile, can carry consequences through the enterprise to investors. The competitive advantage is his forecast, not a measured result presented in the talk.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Build with the people who know the workflow
McDonald’s near-term goal is confidence in information sent to Redding Ridge’s more than 400 global investors without spending hours manually checking it. He also wants technology to become an earlier consideration when teams encounter a problem: ask why a process exists, how it works and whether it should be improved or changed.
For builders, the warning is that an unusual step in a financial workflow may be there for a reason. McDonald credits Rowspace’s progress against its roadmap to spending time almost every day with the people doing the work. Sitting with front-office professionals helps a builder learn the calculations, exceptions and information flows that a useful tool must accommodate.
Suggest correction
This note stays in this page until you copy or download it. Nothing is submitted; reloading clears the draft.
Full transcript
Loading transcript…