At Benefit Street Partners (BSP), a credit-focused alternative investment management firm, answering follow-up questions from the investment committee (IC) used to take days or even weeks. Portfolio Manager Anant Kumar notes that analysts using Codex can now turn around a cycle of feedback in a single day—often answering questions directly in the middle of a meeting. BSP is leveraging these tools to increase the depth and rigor of its underwriting reviews within the same timeframe, rather than simply aiming to close more deals.
“I’ve gone from being just an analyst and portfolio manager to also being a developer. I’m now able to build tools myself.” — Anant Kumar, Portfolio Manager, Benefit Street Partners
Building Custom Dashboards
In the past, creating a data dashboard to share with colleagues required people with software development expertise to build a website and ingest data. Now, he can build tools himself, using AI to draft code and review errors.
The tracking tool he built pulls public SEC filings from Business Development Companies (BDCs) to compare how different institutions value the same loan. This allows the team to see where their own valuation marks stand relative to other holders.
Kumar built the tracker and used Codex for code review. The initial version took just a single day. Over the following weeks, after adding features and refining the interface, he estimates the total development time was around 10 to 12 hours.
Building a Credit Agreement Analyzer
Another tool Kumar developed analyzes credit agreements, which often run 300 pages or more. Investors need to identify provisions that could shift value away from lenders—a task that historically required hiring outside counsel to review agreements page by page. Kumar compiled memos from top law firms detailing key terms and risks to watch. He describes the analyzer as “a second brain inside our firm.”
Uploading a credit agreement takes roughly 20 minutes to generate a first-pass analysis, followed by an hour or more of human review. BSP has since refactored the tool using GPT-6 Astra, and it is now live on the firm’s proprietary internal portal.
“AI is incredibly helpful for organizing, ingesting, and synthesizing data, but the final investment decisions must still be made by humans,” he says.
Underwriting is typically an iterative, multi-round process. A round of review involves further due diligence and revised memos addressing questions from the investment committee. Because deal cycles range from days to weeks, the number of rounds varies. “Because we have more information readily at our fingertips, we feel much more confident in our decisions,” Kumar explains.
“The quality of my analysis has improved dramatically, and it’s much faster.” — Anant Kumar, Portfolio Manager, Benefit Street Partners
Results at a Glance
-
Built the initial BDC tracker in just one day using Codex for code review, with cumulative development time totaling only about 10–12 hours including subsequent feature additions.
-
Generates a first-pass analysis of a credit agreement in roughly 20 minutes, followed by an hour or more of human review.
-
Accelerates responses to investment committee follow-up questions using Codex, often answering them in real time during meetings, according to Kumar.
Looking Ahead
BSP is now working on connecting data points across different deals. “Continuous underwriting” is a longer-term objective: rather than re-underwriting only after an event occurs, the team aims to monitor loan conditions in real time as news breaks and evaluate the broader portfolio impact.
“Codex has freed up my day, allowing me to spend more time on higher-level work and less on routine, tedious tasks,” Kumar says.