A financial workflow was redesigned around the review outcome
Chatham Financial says its Process Zero team built a trade-validation application with Codex. The tool gathers supporting transaction evidence, compares key terms and flags discrepancies for a reviewer. In early measurement, Chatham reports that review time fell from about 30 minutes to under four.
The number is company-reported and the published case study does not specify the sample size. Chatham says it is comparing the application’s results with real transactions and experienced reviewers before expanding automation. The claim is therefore an early workflow result, not a published independent accuracy benchmark.
The useful part is how the work is divided
The application handles evidence assembly and structured comparison, while professionals retain responsibility for judging exceptions and client context. That division is especially important in capital markets, where a record needs to match both what a client authorized and what was executed.
OpenAI’s case study says Chatham is using Codex to build tools and GPT-5.6 to power AI features. The trade-validation application is one example within a broader effort that also includes employee-built applications and Chatham Onyx.
A faster first pass is not the same as an automatic decision
A minutes-long review could free specialists to focus on mismatches and difficult cases, but only if the workflow catches the right exceptions and preserves traceable evidence. Chatham says it plans to extend the tool to more trade types while maintaining professional oversight.
The next useful evidence would be performance across a disclosed sample: agreement with experienced reviewers, error rates, the kinds of discrepancies flagged and how often a human must correct the tool. For now, the concrete news is that a Codex-built application is being evaluated in a real finance workflow, with experts still checking its output.
Source published October 2, 2026. Coverage is based on the maker’s announcement and demonstration.
