How Benefit Street Partners accelerated its credit analysis using OpenAI's Codex

In the competitive and fast-paced investment management market, the ability to respond quickly to critical committee questions can be the difference between closing a profitable deal or losing an opportunity. Benefit Street Partners (BSP), a renowned investment manager focused on credit in North America, found an innovative solution to this challenge. By using Codex, OpenAI's artificial intelligence model, the company was able to drastically reduce the time required to respond to its investment committee queries. What once took days or even weeks can now be resolved in a few hours or even during the meetings themselves, ushering in a new era of agility and analytical rigor in the financial sector.

Context and background: the bottleneck of traditional analysis

Historically, the credit underwriting process and portfolio analysis require a colossal effort to review lengthy documents and cross-reference financial data. In firms focused on alternative credit, like BSP, analysts routinely deal with credit agreements that often exceed 300 pages.

Before the adoption of generative AI technologies, answering follow-up questions from an investment committee required mobilizing multiple departments. Often, this meant tasking outside counsel to leaf through agreements page by page in search of risk clauses, or relying entirely on IT teams to build data dashboards that would allow for the comparison of market metrics. This reliance on technical and legal intermediaries created a natural bottleneck in deal timelines, which could range from a few days to several weeks.

The adoption of Codex by BSP marks a paradigm shift. The tool is not necessarily being used to close a higher volume of deals, but rather to add greater review capacity and rigor to underwriting processes within the same timeframe.

From analyst to developer: the internal revolution at BSP

The protagonist of this transformation at Benefit Street Partners is Anant Kumar, portfolio manager, global investment strategist, and the firm’s leader of artificial intelligence. With the support of Codex, Kumar was able to transcend the traditional technical barriers of his role.

To create an interactive dashboard that could be shared with his colleagues, Kumar no longer needed to open tickets for the software engineering team or hire developers. He used AI to write and review the code for his own tracking and analysis tools, mitigating errors and accelerating the development cycle.

BDC tracking dashboards

The first major tool developed by Kumar was a tracker focused on Business Development Companies (BDCs). This dashboard uses public data from US Securities and Exchange Commission (SEC) filings to compare how different companies value the same loan.

The results in terms of productivity were impressive:

• Initial development: The first version of the tracker was built in just one day.
• Total estimated time: After adding new features and refining the interface over the following weeks, Kumar estimates that the total development time was between 10 and 12 hours.
• Practical impact: The BSP team can now instantly see how their valuations compare to those of other market holders.

The credit contract analyzer and the "Second Brain"

Another crucial innovation was the creation of a credit contract analyzer. These long, complex documents need to be scrutinized for clauses that could divert value from creditors. Kumar gathered memos from major law firms containing guidelines on clauses and risks to monitor, creating what he describes as a "second brain within the organization."

• Agile processing: Uploading a credit agreement now takes about 20 minutes for the AI to generate a first-pass analysis.
• Essential human review: This automated processing is followed by one hour or more of human review, ensuring accuracy and contextual interpretation.
• Updated technology: The tool proved so vital that BSP refactored it using the newly released GPT-6 Astra, and the solution is already operating live on the company's corporate portal.

The role of human intelligence in the age of AI

Despite the massive leaps in efficiency, BSP leadership is keen to emphasize that artificial intelligence does not replace the financial analyst. According to Kumar, “AI is very useful in data collection, ingestion, and synthesis of analysis, but the final investment decision must be made by a human.”

This approach, where the machine acts as a powerful co-pilot for subject matter experts, shows that the real value of large language models (LLMs) in the financial sector depends on them being incorporated into workflows with clear data provenance, controlled permissions, and auditable processes. The result is an investment committee better equipped with evidence: “We feel more confident in our decisions because we have more information at our disposal. The quality of my analysis has become much better and much faster,” adds Kumar.

Impact for the sector, users, and the Latin American market

Benefit Street Partners' initiative resonates across the global financial sector, signaling an era of "citizen developers" at high-caliber institutions. Financial market professionals who do not know how to program professionally can now build bespoke data tools.

Impact for users (Analysts and Managers):

• Freedom from routine tasks, allowing for a focus on high-level analysis and strategy.
• Autonomy in creating technological solutions without exclusive dependence on IT departments.
• Real-time responses during critical stakeholder meetings.

Perspectives for the Latin American market: Latin America has an ecosystem of fintechs and credit managers in rapid expansion (especially in countries like Brazil, Mexico, and Colombia). The adoption of tools like Codex or GPT-6 Astra by local credit funds can help democratize credit risk analysis, which is often limited by a lack of extensive legal resources. Latin American managers struggling with bureaucracy and slow document processes could replicate the BSP model, building internal tools that translate and process large volumes of credit agreements in Spanish or Portuguese in minutes. This could reduce transaction costs and improve regional competitiveness in the global market.

What to expect next

The future of Benefit Street Partners' operations points toward interconnectivity. The company is working to connect and integrate information generated across multiple agreements and past deals.

The long-term aspiration is what the industry calls "continuous underwriting." Instead of re-analyzing a credit or rewriting a memo only after a specific corporate event (such as a company's earnings release), the team will be able to monitor loans in real-time, instantly evaluating the systemic effects of news as they happen across the entire investment portfolio. With Codex "unlocking" the day of professionals like Kumar, there is time left to build the algorithmic future of fundamentalist analysis.

Frequently Asked Questions

1. Is OpenAI's AI making investment decisions at Benefit Street Partners? No. The AI is used strictly to collect data, synthesize files (such as long contracts), and build dashboards. Benefit Street Partners emphasizes that the final decision on credit allocation and investment continues to be, invariably, made by human experts after careful review of the AI's outputs.

2. How much time does BSP save in developing internal tools using Codex? The impact is massive. The first version of a complex market tracker was created in about a day, consuming a total of only 10 to 12 hours of development and refinement, something that would traditionally require weeks of work and coordination with external software engineers.

3. Does the credit contract analyzer ignore the role of lawyers or compliance? The tool does not replace legal guidelines. In fact, it was fed by previous memos from reputable law firms. The system performs the "first pass" in about 20 minutes, highlighting potential risks, which then accelerates the human review (by lawyers and analysts) that follows for another hour or more.