OpenAI and Ironclad Team Up to Advance Computer-Use Agents for Contracting Work

OpenAI and Ironclad are collaborating to train and evaluate AI agents on complex contracting workflows, with the goal of pushing computer use out of demos and into professional work.

According to OpenAI, the partnership centers on having agents carry out tasks inside real, multi-step contracting processes and measuring how reliably they do so. Contracting workflows typically span multiple documents, clause-by-clause comparison and cross-system actions, tasks that place a premium on accuracy and consistency.

Computer use refers to giving a model the ability to operate software the way a person does: clicking, typing and moving between windows instead of relying on a bespoke integration for every system. Dropped into a professional process like contracting, that means an agent has to follow context, honor company rules and fail in ways a human can catch.

For OpenAI, teaming up with an industry software vendor is a practical way to obtain real business processes and evaluation criteria. Ironclad's contracting focus can supply task samples and judging dimensions that generic benchmarks struggle to cover.

The collaboration reflects a broader shift: the agent race is moving from whether a model can operate a computer at all to whether it can deliver reliably inside a company's critical workflows. That makes evaluation methods and industry settings the next contested ground.

Details remain limited for now. The specific training approach, the evaluation metrics and whether the work will become a customer-facing product have yet to be spelled out, and those are the things to watch next.