Benefit Street Partners (BSP), a credit-focused investment manager, is using OpenAI Codex to build internal tools that accelerate credit research, loan valuation comparisons and investment committee follow-ups. In an OpenAI customer story dated October 16, 2026, portfolio manager Anant Kumar explains how AI-assisted development is helping analysts work through complex financial documents and answer questions faster without delegating final investment decisions to AI.
From analyst to tool builder with Codex
Kumar previously depended on software specialists to create dashboards and connect data. Now he builds tools himself and uses Codex to review generated code for errors. One example is a business development company (BDC) mark tracker that draws on public SEC filings to compare how different firms value the same loan. This lets the team compare its own marks with other holders' valuations.
The first version was built in about a day. Including later interface and feature refinements, Kumar estimates roughly 10–12 total development hours. These are figures reported in the customer case study, not independent performance benchmarks.
Credit agreements analyzed in 20 minutes, then reviewed by humans
Credit agreements can exceed 300 pages and include provisions that materially affect lender protections. Kumar assembled guidance from law-firm memoranda to help build an analyzer that flags contractual issues and risks. The system processes an uploaded agreement into a first-pass analysis in around 20 minutes, after which a human reviewer spends at least another hour checking the findings.
BSP subsequently refactored the tool with GPT-6 Astra, and OpenAI reports that it is live in the company's hosted portal. The workflow illustrates AI-assisted document synthesis rather than fully automated legal or investment advice.
Faster investment committee follow-ups
According to Kumar, answering follow-up questions from investment committees could previously take days or weeks. With the new tools, analysts can often turn around a round of comments within a day and sometimes answer questions during the meeting itself. BSP says the aim is to increase underwriting rigor and review capacity within existing timeframes, not simply to increase deal volume.
Human judgment remains central
The company emphasizes that AI can collect and synthesize evidence, but humans remain accountable for evaluating it and making investment decisions. This distinction matters in financial services, where auditability, interpretation of contractual clauses and risk assessment require expert oversight.
What comes next: continuous underwriting
BSP is exploring how to connect information across deals and potentially monitor loan portfolios as new information emerges. Continuous underwriting remains a longer-term aspiration, not a capability that the source says has been fully deployed.
Why this matters for enterprise AI
The case highlights a shift in who can create internal software: subject-matter experts can increasingly build specialized applications, with coding agents supporting development and review. The reported benefits are faster iteration, quicker access to relevant evidence and more time for high-value judgment.
NetContentSEO assessment: This is an enterprise AI and workflow automation story, not a confirmed change to search engines, SEO rankings, crawling, indexing or AI search citations. The most transferable lesson is the combination of domain expertise, AI-assisted coding and mandatory human verification.
Primary source: OpenAI — Benefit Street Partners answers questions faster with Codex, October 16, 2026.