OpenAI Cuts Trade Checks From 30 Minutes To Under 4

OpenAI has published a case study showing how Chatham Financial, a US capital markets advisor, used its Codex tool and GPT-5.6 model to redesign internal workflows. The result: trade validation checks that used to take 30 minutes now take under 4 minutes, according to OpenAI's write-up published on 2 October 2026.

Chatham isn't a UK trades firm, but the maths translates. Trade validation is really just document checking, comparing numbers against a reference and flagging mismatches. Every UK business with paperwork like that, supplier invoices against purchase orders, RAMS documents against site requirements, quote line items against material lists, is doing the same job by hand.

What this means in hours

Say a building firm checks 20 supplier invoices a week, each taking 30 minutes to cross-reference against purchase orders and delivery notes. That is 10 hours a week on one task alone. Apply the same roughly 87% time cut Chatham reported and that drops to about 80 minutes a week, a saving of nearly 9 hours. Over a year, that is more than 400 hours, the equivalent of ten working weeks, freed up without hiring anyone.

These are illustrative figures based on the ratio OpenAI reported for Chatham's specific workflow, not a guaranteed result for every business. The real saving depends on how structured your documents are and how much setup work the automation needs. A firm with messy, handwritten delivery notes will see less of this benefit than one working from digital POs.

What to do this month

• Pick one document-checking task you do weekly, invoices, RAMS, quotes.
• Time how long it actually takes you or your team over a full week.
• Test whether a tool like ChatGPT or Codex can draft the first pass of that check.
• Keep a human doing the final sign-off. The saving comes from cutting the first read-through, not removing oversight.