Chatham Financial says Codex-built tool cut early trade review time to under four minutes

Chatham Financial said on October 2 that a trade-validation application built with OpenAI’s Codex reduced review time from about 30 minutes to under four in early measurement. The capital markets adviser says the tool gathers transaction evidence, compares key terms and flags discrepancies for review. The reported speed gain could free staff time, but Chatham has not disclosed the measurement sample or accuracy results.

How Chatham's trade-validation application works

Chatham’s Controls and Data Integrity team checks whether each transaction record reflects what a client authorized and what was executed, according to the announcement published by OpenAI. The Codex-built application collects supporting evidence and compares transaction terms, then draws attention to differences for a reviewer. That puts the reported time saving in the part of the process where staff assemble and check information, rather than establishing that a machine makes the final judgment on a trade.

The work is part of Chatham’s Process Zero service, which the firm describes as redesigning workflows around their intended outcomes. For each workflow, Chatham says it identifies the inputs and evidence needed, where human judgment is essential and how AI-built tools should support the work. Chief executive Matt Henry said the aim is to identify where judgment matters and design a way to deliver the desired outcome using available technology.

What the under-four-minute figure shows

Alex Nordlinger, co-head of Chatham’s AI Advisory practice, described the change as an early measurement: review that took approximately 30 minutes took under four with the application. He said Chatham is validating its performance against real transactions and experienced reviewers before expanding automation. The announcement does not specify the number or types of trades measured, the comparison protocol, error rates or measured accuracy.

Why accurate trade records matter

The stakes of checking trade terms extend beyond a single firm’s workflow. In a 2007 report, the U.S. Government Accountability Office examined confirmation backlogs among major credit-derivatives dealers. It found that 14 dealers had more than 150,000 unconfirmed trades in September 2005, partly because manual processes had not kept pace with trading volumes. The agency said unconfirmed terms could allow errors to go undetected and increase operational risk.

The GAO later reported that a joint regulatory initiative and greater automation helped those dealers cut confirmations outstanding for more than 30 days by 94%, to 5,500 by October 2006. That episode concerned a different market problem and period. It illustrates why reliable records, detection of discrepancies and oversight matter when financial firms try to speed up trade processing; it does not test Chatham’s new application.

Where the tool fits in Chatham's wider AI work

Chatham says it also uses Codex to build internal and client-facing tools and GPT-5.6 to power AI features. Its Onyx platform is designed to bring information on assets, debt and derivatives into a connected environment while retaining links to underlying source material. The firm says its teams use Codex to help plan, build, test, document and review software for that platform, with different OpenAI models assigned to different tasks.

One Onyx feature, ChatFIN, is described as summarizing patterns in historical market data, helping users understand portfolios and locating documents concerning debt, derivative and lease terms. Chatham says its advisers interpret that information for clients. Separately, employees use an internal application-building platform called Chatham Vibes for work that includes reviewing maturing-cap trades, preparing pricing workbooks and drafting client communications. The firm says its professionals evaluate and refine work before it reaches clients.

For trade validation, the next question is whether the faster review can be repeated while preserving the accuracy Chatham’s controls team is meant to protect. Chatham says it is comparing the application with experienced reviewers and plans to expand it to other trade types. Until it releases the comparison method and results, the under-four-minute figure remains an early company-reported measure, rather than an independently verified performance benchmark.