New analytics relate usage, cost, tasks and results in ChatGPT Work and Codex. See how to transform telemetry into business decisions.
Direct answer
On September 16, 2026, OpenAI introduced an approach to connecting AI usage to business value in ChatGPT Work and Codex. Analytics combine adoption, spend, task classification, and outcome indicators, including Codex contributions to integrated commits. The tool improves visibility, but the company still needs to provide process context, baseline, and quality criteria to distinguish proven impact activity.
Activity is not value
Number of messages, active users and tokens shows adoption but does not prove benefit. The dashboard needs to be linked to time saved, quality, revenue, risk or capacity delivered.
Classifying tasks reveals where AI works
Grouping interactions by work type helps you find processes with volume and potential. The taxonomy must reflect the actual operation and be revised when flows change.
Baselining makes ROI defensible
Before piloting, record time, cost, rework rate, and quality without AI. Then, compare equivalent samples and include review, integration, and infrastructure in the total cost.
Code requires measurements beyond lines
Built-in commits and code contributions indicate Codex adoption, but do not capture maintenance, incidents, or product value. Testing, review, and delivery metrics complete the reading.
Nexus Reading
The best panel starts with the business question and ends with a decision. Metrics must show where to expand, correct or stop use, with auditable evidence and defined responsibility.
FAQ
Does high usage mean positive return?
No. Usage measures activity; return depends on proven improvement in the result and the total cost of the process.
What metrics to start with?
Time per task, acceptance rate, rework, cost per result, quality and error incidence.
How to avoid an artificial ROI?
Use previous baseline, comparable samples, complete costs, and validation by those responsible for the process.
Essential guides to delve deeper into the decision
This editorial analysis was produced by Nexus from the official sources below, consulted on September 17, 2026. The text is original and interprets practical implications for companies.
- OpenAI — How to connect AI usage to business value: announcement and official methodology for relating usage analytics, tasks, costs and results.
- OpenAI — A scorecard for the AI age: official proposal for metrics aimed at useful intelligence, total cost and results per task.
Date reported by the main source: September 16, 2026.