The News, Explained

OpenAI describes a way to connect ChatGPT Work and Codex usage to business results through the ChatGPT Admin Console. The approach combines Usage, Insights, and Codex Outcomes data with workflow baselines and outcome metrics such as time, quality, review effort, and cost. Source

1. Check adoption in Usage

The Usage view shows active users, credits, and token consumption. Admins can filter by group or user to find teams with lower adoption and use that information to investigate suitable starting workflows or training needs. Source

It also shows the credit mix across model, reasoning, and speed settings and the use of plugins and skills. OpenAI gives an example of testing a faster or lower-cost setting for routine brief work while comparing output quality and the time spent reviewing and editing the result. Source

2. Pick a workflow in Insights

Insights groups sampled messages into use cases and tasks to show what work AI is supporting. Admins can look at the task mix by team, choose a common workflow tied to a business priority, and agree with the workflow owner on the result to measure and when progress will be reviewed. Source

3. Establish a baseline before the comparison

For the selected workflow, record how often it occurs, how long it normally takes, and what quality standard the output must meet. Those pre-AI measurements form the baseline used for comparison. After AI is used for a defined period, compare task time, review and editing time, quality, and cost against the same baseline. Source

For software work, Codex Outcomes can show the share of merged commits and lines of code with Codex contributions, plus code-review activity, with filters for groups, users, and repositories. OpenAI recommends looking at those measures together with review time, defects, and rework. Source

4. Combine AI data with operating metrics

Admins can use the Admin plugin to compare adoption, spend, and task data and generate reports. The Admin API can automate analysis in an internal dashboard and combine it with data from business systems. An API is a defined interface that lets one system request data or functions from another. OpenAI gives the example of viewing credit usage alongside ticket-resolution time in a support dashboard. Source

5. Calculate capacity and cost from local assumptions

OpenAI’s sales-account research example assumes 20 sellers save three hours on two briefs per week for 46 working weeks, producing 5,520 hours of annual time savings. Valuing 50% of that time at a fully loaded labor rate of $75 per hour produces an estimated capacity value of $207,000. A fully loaded labor rate includes pay plus employer-paid benefits and related employment costs. With $60,000 in assumed first-year costs, the example ROI is 245%. ROI, or return on investment, expresses the net benefit relative to the cost of the investment. The example is based on assumptions and excludes changes in win rate or deal size. Source

OpenAI also says organizations’ business data is not used for model training by default. Source

OYOPICK’s Take

A baseline shared by administrators and workflow owners can make the case for AI spending more concrete. The example’s capacity value assumes saved time is redirected to productive work. A team can record where shorter preparation time actually goes and whether output quality is maintained.

A dashboard becomes more useful when that comparison informs the next adoption decision. We hope teams find workflows delivering good results at modest cost and share those verified practices.