Credit Saison rolls out ChatGPT Enterprise to all employees, reports 954% ROI in pilot

According to a public OpenAI case study, Credit Saison rolled out ChatGPT Enterprise to all employees in September 2025. A pilot involving 315 people from late June through early August 2025 recorded 954% ROI and monthly time savings of 14 hours per participant, and the company said it decided to roll out the product companywide one month after the pilot ended.

The public OpenAI case study says Credit Saison based its decision to roll out ChatGPT Enterprise companywide on time savings and other results from the 315-person pilot. It is a case in which the basis for a companywide rollout decision and subsequent usage results can be assessed separately; the key points are how ROI was calculated and what effects followed the rollout.

The focus of this case is the decision to move generative AI beyond a pilot and make it available to all employees, linking that decision to the results of a short-term trial. OpenAI’s published account cites monthly time savings and ROI for the 315 participants, and says the company decided on a companywide rollout one month after the trial ended. What can be established here, therefore, is an operational approach that used measured results to inform the rollout decision—not that all employees achieved the same results in the same tasks. The ROI and time savings from the ChatGPT Enterprise pilot should be distinguished from the average monthly savings of 850 hours from Codex use by the engineering team. OpenAI’s case study lists these as separate metrics, but the text provided does not specify the tasks covered, calculation methods or basis for the ROI calculation. The companywide rollout at a business with about 31.16 million cardholders indicates its scale, but the account does not say that AI was applied to tasks such as customer service or screening. The information also does not establish the scope of use in financial-related operations or workflows that include human review. OpenAI reportedly indicates timing for public release of humanoid robots, covered by this newspaper, focused on reports about OpenAI’s humanoid robots. This case should be treated separately as an example of expanding generative AI use across a company, not as evidence of robot development or deployment of physical systems. Even when the same company name appears in discussions of physical AI, a record of expanding software use cannot be recast as progress in robotics technology or business. The next points to examine are the costs and benefits included in the 954% calculation, whether usage rates and time savings continued after the companywide rollout, and which tasks account for the time saved through Codex. The case study provided does not show these measurement conditions or results by task after implementation. See the original source for details.