Answering Investment Questions in the Meeting: How Benefit Street Partners Speeds Up Underwriting with Codex

Hi everyone, it's Shii-chan!

OpenAI News openai.com

Today I found an OpenAI customer story about how Codex is used on an investment team, so let me walk you through it!

What was announced?

OpenAI News published a customer story about Benefit Street Partners (BSP), a credit-focused investment manager.

It features portfolio manager Anant Kumar, who uses Codex to build his own analysis tools and to answer investment committee questions much faster.

Why it matters

Answering an investment committee's follow-up questions used to take days or weeks.

Now analysts can turn a round of comments in a day, and often answer questions during the meeting itself.

The key point is that BSP isn't using this to do more deals. It's using it to add more review capacity and rigor to underwriting within the same timeframe.

What changes

Here's the before and after:

• Before: building a shareable dashboard needed someone with software expertise
• Now: Kumar builds tools himself, using AI to draft code and review it for errors
• Before: credit agreements of 300 pages or more went to outside counsel to be read page by page
• Now: an uploaded agreement gets a first-pass analysis in about 20 minutes, followed by an hour or more of human review

Kumar sums it up like this (paraphrased):

He says he has gone from being an analyst and portfolio manager to being a developer as well, and can now build things himself.

Dive Deep

Let's look at the two tools from the story!

BDC tracker

• Uses public SEC filings from business development companies (BDCs)
• Compares how different firms value the same loan
• Shows how the team's marks compare with other holders' valuations
• Codex was used for code review
• The first version took a day; total development time, including later features and interface refinements, is about 10-12 hours

Credit agreement analyzer

• Targets credit agreements, often 300 pages or longer, to find clauses that could shift value away from lenders
• Built on memos from leading law firms about clauses and risks to watch for
• Kumar calls it a "second brain within our organization"
• BSP refactored it with GPT-6 Astra, and it is now live in the company-hosted portal

The line is clear: the final investment decision is made by a human. AI helps with collating data, ingesting it, and synthesizing analysis.

Looking ahead, BSP is working to connect information across deals. Its longer-term aspiration is continuous underwriting: monitoring loans as news happens and examining the effects across the portfolio.

Wrap-up

• At BSP, follow-up answers for the investment committee went from days or weeks to about a day, sometimes during the meeting
• The first BDC tracker took a day, and about 10-12 hours in total
• The credit agreement analyzer produces a first pass in about 20 minutes, followed by an hour or more of human review
• The final decision stays with a human, and AI supports the organizing and analysis

If you're a finance or legal professional who wants to build your own tools, this story is for you!