OpenAI FDE: Closing the deployment gap with enterprise case studies

Details

OpenAI Forward Deployed Engineering (FDE) addresses the "deployment gap" by embedding with customers to solve complex problems, aiming to make their own role obsolete by raising the baseline for AI reliability.
In a semiconductor engagement, FDEs achieved ~48% time savings in directed test generation and compressed weeks of debugging into hours. Later, by migrating to Codex-based plugins and skills, they removed roughly 90% of the custom code from each project.
For Thrive Holdings, the team built Tax AI that processed 7,000 returns and increased throughput by ~50%, achieving 90% precision and recall on rental-property returns after six weeks of refinement.
When model capabilities fell short, such as in a telecommunications voice project where initial evals scored ~60%, FDE collaborated with Research to improve models, raising performance to 90%+ over six months.
The approach prioritizes starting with the outcome to get value into users' hands, then scaling what works by converting custom solutions into reusable product components like OpenAI Presence.

Key point