Benefit Street Partners answers faster with Codex, accelerating the technical work that underpins underwriting and investment committee follow-ups. Portfolio manager Anant Kumar says analysts who once took days or weeks to return committee comments can often now supply responses within a day — and sometimes address questions during the meeting itself. The firm has integrated Codex into several parts of its workflow to compress routine tasks and free human analysts for higher‑level judgment.
One of the most tangible wins has been in internal tooling. Kumar used Codex to assemble a dashboard that compares loan valuations across holders by drawing on public SEC filings from business development companies (BDCs). The tracker makes it simple for the team to see how BSP’s marks line up with valuations reported by other holders and flags differences that merit closer review. Where BSP once required software specialists to assemble a site and ingest the data, Kumar now drafts and reviews the underlying code with AI assistance. He reports the initial version of the tracker took a day to create; subsequent feature work and interface refinements brought the total development time to roughly 10–12 hours.
Codex has likewise been applied to credit agreement analysis, a traditionally time‑intensive activity when documents can top 300 pages. BSP deployed a credit agreement analyzer that ingests full agreements, collates guidance from leading law firms on relevant clauses, and delivers a first‑pass analysis in roughly 20 minutes. That automated pass is then followed by an hour or more of human review before the findings are used internally. Kumar describes the analyzer as a “second brain” for the firm, helping to surface clauses and risks quickly so human reviewers can focus on judgment and synthesis rather than rote extraction.
The firm has continued to iterate on the toolset: BSP refactored the agreement analyzer with GPT‑6 Astra and now runs it inside a company‑hosted portal. Kumar stresses the AI’s role as a collator and synthesizer of information while making clear that the ultimate investment decision remains with human analysts. The workflow intentionally pairs a rapid automated first pass with deliberate human checks, blending speed with the caution appropriate for credit underwriting.
Operationally, the adoption of Codex and related models is aimed at increasing review capacity and rigor within existing timeframes, not immediately expanding deal volume. According to BSP, the tools have improved both the speed and quality of analysis, allowing the team to complete more high‑level work and reduce routine tasks. Key outcomes cited by the firm include the roughly one‑day build of the initial BDC tracker (about 10–12 hours total development time), the approximately 20‑minute automated first‑pass credit agreement review followed by human checks, and faster answers to investment committee follow‑ups — often delivered in the meeting itself.
Those gains are deliberate: rather than replacing human expertise, BSP is using Codex to shift where human time is spent. Analysts can move away from repetitive extraction and code assembly to concentrate on interpretation, negotiation points and portfolio strategy. Kumar’s example of drafting and reviewing code himself — aided by AI — underscores the firm’s preference for a model in which domain experts, not only software specialists, can iterate on internal tools.
Looking ahead, BSP is working to connect information across deals with the goal of moving toward continuous underwriting. That vision would allow the team to monitor loans as news occurs and assess portfolio impacts without waiting to re‑underwrite only after an event. For now, the immediate focus remains practical: using Codex and related models to speed analyses, strengthen internal review processes and shorten the turnaround on committee inquiries.
By combining automated first passes with careful human review, Benefit Street Partners answers faster with Codex while preserving human accountability in investment decisions. The firm’s approach highlights how targeted AI deployment can increase throughput and analytical rigor without foregoing the judgment that underpins credit investing.
Source: Read the original source
Leave a Reply