In a case study published October 2, 2026, OpenAI describes how Chatham Financial, a firm that advises clients on capital markets decisions, is using OpenAI's Codex and GPT-5.6 models to build internal and client-facing tools. According to the announcement, Chatham used Codex to build a trade validation application that gathers transaction evidence, compares key terms, and flags discrepancies for review. OpenAI states this reengineering, part of a service Chatham calls Process Zero, reduced trade validation time from 30 minutes to under 4. Chatham says it is comparing the app's results against those of experienced human reviewers and plans to extend it to additional trade types while maintaining professional oversight. The case study also describes Chatham Vibes, an internal platform letting employees build applications, with AI features running on GPT-5.6 Terra by default and GPT-5.6 Sol as an optional upgrade. Chatham's capital markets operating system, Chatham Onyx, uses multiple models including GPT-5.6 Sol, GPT-5.6 Terra, GPT-5.4, and GPT-4.1, routing simpler tasks to cheaper models. The account is published by OpenAI about a customer of its own products. No independent verification of the stated time savings was provided.
- OpenAI published the case study on October 2, 2026
- Chatham reports cutting trade validation from 30 minutes to under 4 using a Codex-built app
- Chatham uses multiple OpenAI models (GPT-5.6 Sol/Terra, GPT-5.4, GPT-4.1), routing simpler tasks to cheaper ones
- Employees build their own apps via an internal platform called Chatham Vibes
What it means for you
A financial advisory firm automated a routine checking task and says it went from 30 minutes to a few minutes. The useful lesson here is not the headline number — it's the method: they broke each workflow into 'what's the minimum evidence needed, where does human judgment actually matter, and what can a tool handle,' then kept experts reviewing the output. That approach works for small businesses too, and you don't need custom software to try it.
Try this
Pick one repetitive checking or comparison task in your business (invoice matching, confirming details across two documents, cross-referencing records). Spend an hour testing whether an AI tool can draft the comparison and flag discrepancies — while you still make the final call. Keep a human reviewing it until you trust it.
Who should care
Small businesses with repetitive, rules-based verification or data-comparison work, especially where accuracy and an audit trail matter (bookkeeping, order processing, contract review).
Skip this if
You don't have high-volume routine checking tasks, or your work is mostly judgment and relationship-based with little repeatable process to automate. This is a large-firm case study; treat the specific numbers as marketing, not a promise.
Sources: OpenAI — read the original