OpenAI Admin Console Analytics: Connecting AI Usage to Business Value

TL;DR

OpenAI introduced analytics tools in the ChatGPT Admin Console that merge usage, spend, task classification, and outcome metrics for ChatGPT Work and Codex, enabling admins to measure AI adoption, identify training needs, and connect AI activity to concrete business outcomes.

Understand AI Usage and Spend

The Usage view aggregates active users, credit consumption, and token usage across ChatGPT Work and Codex. By filtering on groups or individual users, admins can spot low‑adoption areas, justify workflow reviews, and prioritize training or capacity planning.

The Usage overview shows active users and credit trends across ChatGPT Work and Codex.

See What Work Teams Are Doing with AI

The Insights tab includes a task classifier that groups a sample of messages into use‑case categories (e.g., software engineering, sales & revenue). The Overview tab presents a high‑level mix of work, while the Use cases tab provides a detailed table with credits, messages, and active users per task. Admins can filter by group to surface the dominant AI‑supported workflows for each team.

The Insights overview shows how credits are distributed across tasks, from implementing features to account research.

Identify Training and Support Gaps

Task‑level breakdowns—Models, Reasoning, and Speed—show the share of credits each setting consumes, helping admins evaluate whether the chosen model configuration aligns with the task. The Plugin leaderboard and Skills view reveal which integrations are under‑ or over‑used, flagging potential access issues or training opportunities.

Account research and planning analytics display credit shares by model, reasoning level, and speed, alongside plugin and skill usage.

Track Codex Contributions to Engineering Outcomes

The Outcomes view reports Codex‑generated contributions to merged commits and lines of code, as well as code‑review activity. Filters for groups, users, or repositories let engineering leaders gauge adoption trends and decide where to expand access or provide additional support.

The Outcomes view tracks the share of merged commits and lines of code with Codex contributions over time.

Automate Reporting with the Admin Plugin and API

The Admin plugin can generate ready‑to‑share reports—charts, key findings, and next‑step recommendations—directly within ChatGPT Work. The Admin API enables teams to pull analytics into custom dashboards and combine them with external business‑system data (e.g., ticket resolution times).

The Admin plugin compares team credit trends and summarizes changes for a monthly rollout review.

Connecting Analytics to Business Outcomes

Usage and task data provide a baseline; business owners add context about workflow changes, quality improvements, and monetary impact. A structured evaluation follows five questions:

  1. What would you like to improve? (e.g., faster preparation, higher quality, lower cost)
  2. How does the process look today? (baseline frequency, duration, quality criteria)
  3. What changes with AI? (measure AI‑augmented results, including review effort)
  4. What does this make possible? (additional customer interactions, reduced rework, etc.)
  5. Is the benefit worth the investment? (compare gains to AI spend and support costs)

Illustrative ROI Example – Sales Account Research

Assumptions are hypothetical and for illustration only.

  • Time saved: 5,520 hours per year (20 sellers × 2 briefs/week × 3 h saved × 46 weeks).
  • Capacity value: 5,520 h × 50 % utilization × $75/h = $207,000.
  • First‑year AI cost: $60,000.
  • Illustrative ROI: (207,000 – 60,000) ÷ 60,000 ≈ 245 %.

All figures are hypothetical. ROI reflects estimated capacity value and excludes potential benefits from higher win rates, larger deal sizes, or other sales outcomes.

Real‑World Customer Impact

  • 1Password uses Codex for code generation, review, and testing, reporting a 553 % ROI and $0.8 M annual engineering capacity value.
  • ATV Big Air Tour leverages ChatGPT Work for event listing checks and inventory planning, cutting listing review time from eight hours to one hour per week and inventory work from days to hours.
  • Playco employs GPT‑6 Astra via the OpenAI API to prototype games, achieving 50 % fewer manual fixes compared with the prior model.

Getting Started

  1. Open Insights in the Admin Console.
  2. Choose a high‑impact task and review its classification with the relevant business owner.
  3. Agree on a baseline metric and the outcome to measure.
  4. Set a review date, collect data, and decide whether to expand, refine, or replace the workflow.

Learn more about enterprise analytics and the Admin API.

Sources