随着更多团队使用 AI，管理员和业务负责人需要了解 AI 在哪些方面创造价值，以及下一步应在哪里投入。使用量和支出能说明部分情况，但管理员还需要了解人们用 AI 做什么，以及它帮助他们完成了什么。
ChatGPT 管理控制台中的分析功能汇集了 ChatGPT Work 和 Codex 的使用量与成本数据、任务洞察和成果指标。
以下是管理员如何使用这些工具来了解采用情况、支持团队并评估业务价值。
了解 AI 使用情况和支出
使用量分析显示采用率在哪些方面增长、支出集中在哪些方面，帮助管理员聚焦支持、审查成本并评估容量请求。使用量视图汇集了 ChatGPT Work 和 Codex 的活跃用户、额度和 token 使用量。例如，按群组或用户筛选可以揭示采用率较低的地方，让管理员有理由与团队负责人一起审查入门工作流和培训需求。
使用量概览显示 ChatGPT Work 和 Codex 的活跃用户和额度趋势。所有截图均使用示意性演示数据。
了解团队用 AI 做什么工作
洞察中的任务分类器通过将消息样本归入用例和任务，帮助管理员了解 AI 支持哪些工作。例如，软件工程包括功能开发和代码维护，而销售与营收包括客户研究和规划。概览标签页一目了然地展示工作构成；用例标签页提供包含任务明细的详细表格。管理员可以按群组筛选，了解团队如何使用 AI，然后与业务负责人一起决定评估哪些工作流和成果。
洞察概览显示额度如何分布在各任务中，从实现功能到客户研究。
对于下方显示的销售团队，客户研究和规划是额度使用最多的用途——这是讨论 AI 如何改变客户准备工作的起点。
用例表格按任务细分团队工作，并显示额度、消息和活跃用户。
确定团队需要培训和支持的地方
在任务详情中，模型、推理和速度明细显示每项设置在该任务额度中的占比。这有助于管理员评估设置是否适合该工作，并针对模型选择开展培训。例如，对于例行简报，可能值得用更快或成本更低的设置进行测试，并比较质量以及审查和纠正所花的时间。
插件排行榜和技能视图显示哪些工具支持某项任务，帮助管理员聚焦培训，并决定维护或分享哪些工作流。相关插件使用率低可能意味着存在访问或培训需求；频繁使用的技能可能需要明确负责人并定期更新。
跟踪 Codex 对工程成果的贡献
成果视图显示 Codex 对已合并提交和代码行数的贡献，以及代码审查活动。趋势以及可用的群组、用户或代码仓库筛选器，帮助管理员和工程负责人了解采用情况，并决定在哪里扩大访问权限或支持团队。
如果 Codex 对已合并代码的贡献占比不断增长，工程负责人可以将这一趋势与审查时间、缺陷和返工进行比较，以评估它是否帮助团队更有效地交付软件。
成果视图跟踪一段时间内 Codex 贡献的已合并提交和代码行数占比。
ChatGPT Work 中的 Admin 插件让管理员可以比较采用情况、支出和任务，然后将发现转化为用于预算和推广决策的报告。它还可以创建成品，例如包含图表、关键发现和建议后续步骤的领导层演示文稿，随时可与跨职能利益相关者分享。
借助 Admin API，团队可以在自己的仪表板中自动生成报告，并将分析与业务系统数据相结合。例如，支持仪表板可以同时显示额度使用情况和工单解决时间。
Admin 插件比较团队额度趋势，并为月度推广审查总结变化。
将分析与业务成果联系起来
使用量和任务数据是管理员与业务负责人一起调查价值的起点。业务负责人补充评估价值所需的背景信息：工作流发生了什么变化、结果是否改善，以及这种改善价值多少。双方可以共同将产品活动与交付时间、质量或盈利能力等指标联系起来。
假设管理控制台中的任务分类器显示，客户研究在某个销售群组中很常见。管理员检查模型选择是否适合该任务，在 CRM 插件使用不足的地方提供培训，并分享一项技能以实现一致的客户计划。随后，销售负责人将准备时间和计划质量与团队基线进行比较。如果节省了时间，他们会跟踪有多少时间投入到客户对话中、哪些对话成为合格机会，以及哪些机会转化为销售。该业务的收入和贡献利润率有助于说明额外容量是否产生了财务收益。
要探索某个工作流的价值，可以先问几个问题：
你想改进什么？ 选择一个对团队重要的成果，例如更快的准备、更高质量的工作、更低的成本或更多销售。
今天这个过程是怎样的？ 建立一个起点：团队执行该任务的频率、耗时多久，以及好的结果是什么样的。
有了 AI 会有什么变化？ 在确定的时间段内比较结果。包括审查和纠正工作所花费的时间，这样即使完成得更快，也仍然符合团队的质量标准。
这为团队带来了哪些可能？ 节省的时间可能意味着与客户进行更多对话。更高质量的工作可能意味着更少的纠正。与团队沟通，了解这些改进在哪些方面最重要。
收益是否值得投入？ 将团队获得的收益与你在 AI、设置和持续支持上的支出进行比较。这可以帮助你决定扩展哪些方面，以及团队在哪些方面可能需要更多帮助。
示例性年度 ROI：销售客户研究
设想一个由销售人员组成的团队，每人每周准备两份客户简报。每份简报都汇集了客户历史、行业研究以及为下一次对话准备的观点。
有了 AI 会有什么变化？
这加起来是多少？
年度节省时间
20 名销售人员 × 每周 2 份简报 × 节省 3 小时 × 46 周 = 5,520 小时
估计年度产能价值
假设团队将其中 50% 的时间投入到有价值的工作中，按员工全负荷成本每小时 75 美元计算。销售人员可以利用这段时间联系新潜在客户、跟进机会，并花更多时间与客户相处。
5,520 小时 × 50% × 75 美元 = 207,000 美元
第一年总成本
假设 60,000 美元用于 AI、设置、培训和持续支持。
（207,000 美元 − 60,000 美元）÷ 60,000 美元 = 245% 示例性 ROI。
工作流的价值可以超越生产力。在销售中，更好的客户研究和客户共识可以支持更深入的对话、更牢固的关系和新的机会。团队随后可以衡量这些改进是否会带来额外的业务成果，例如加快交易周期或提高赢单率。
所有数字均为假设。ROI 反映的是估计产能价值，不包括更高赢单率、更大交易规模或其他销售成果带来的潜在收益。
客户通过 AI 看到的业务影响
从交付软件到举办现场活动，客户正在将 AI 用于对其业务至关重要的任务：
1Password使用 Codex 构建、审查和测试软件，以便工程师更快地交付功能。它估计 ROI 为 553%，年度工程产能价值为 80 万美元，使用 Codex 更快地构建生产就绪的功能和内部工具，同时保持严格的安全策略。
ATV Big Air Tour使用 ChatGPT Work 检查活动列表、规划库存并提高其网站的可见性。列表审查从每周八小时降至一小时，库存工作从两到三天降至两到三小时，让团队有更多时间专注于巡演及其客户。
Playco通过 OpenAI 的 API 使用 GPT‑6 Astra 构建和测试可玩的游戏原型。该团队从一个基础创建了三个主题原型，并报告手动修复比使用前一个模型减少了 50%，帮助开发者测试和比较更多想法。
如何开始
打开 Admin Console 中的 Insights，选择一个支持业务优先事项的常见任务。与业务负责人一起审查，就基线和要衡量的结果达成一致，并设定审查进展的日期。利用结果来决定是扩展工作流、改进团队使用方式，还是测试不同的方法。请在我们的文档中了解更多（在新窗口中打开）。
As more teams use AI, admins and business leaders need to understand where it creates value and where to invest next. Usage and spend tell part of the story, but admins also need to see what people use AI for and what it helps them accomplish.
Analytics in the ChatGPT Admin Console bring together usage and cost data, task insights, and outcome metrics across ChatGPT Work and Codex.
Here’s how admins can use these tools to understand adoption, support teams, and assess business value.
Understand AI usage and spend
Usage analytics show where adoption is growing and spend is concentrated, helping admins focus support, review costs, and assess capacity requests. The Usage view brings together active users, credits, and token usage across ChatGPT Work and Codex. For example, filtering by group or user can reveal where adoption is low, giving admins a reason to review starting workflows and training needs with the team owner.
The Usage overview shows active users and credit trends across ChatGPT Work and Codex. All screenshots use illustrative demo data.
See what work teams are doing with AI
The task classifier in Insights helps admins understand what work AI supports by grouping a sample of messages into use cases and tasks. Software engineering, for example, includes feature development and code maintenance, while sales & revenue include account research and planning. The Overview tab shows the mix of work at a glance; the Use cases tab provides a detailed table with task breakdowns. Admins can filter by group to see how teams use AI, then work with business owners to decide which workflows and outcomes to evaluate.
The Insights overview shows how credits are distributed across tasks, from implementing features to account research.
For the sales team shown below, account research and planning is the largest use of credits—a starting point for discussing how AI changes account preparation.
The Use cases table breaks down a team’s work by task, with credits, messages, and active users.
Identify where teams need training and support
In task details, the Models, Reasoning, and Speed breakdowns show each setting’s share of credits for a task. This helps admins assess whether the setup fits the work and target training on model selection. A routine brief, for example, may be worth testing with a faster or lower-cost setup, comparing quality and the time spent reviewing and correcting it.
The Plugin leaderboard and Skills view show which tools support a task, helping admins focus training and decide which workflows to maintain or share. Low use of a relevant plugin may point to an access or training need; a frequently used skill may need a clear owner and regular updates.
Track Codex contributions to engineering outcomes
The Outcomes view shows Codex contributions to merged commits and lines of code, alongside code-review activity. Trends and available group, user, or repository filters help admins and engineering leaders understand adoption and decide where to expand access or support teams.
If Codex contributes to a growing share of merged code, engineering leaders can compare that trend with review time, defects, and rework to assess whether it is helping the team ship software more effectively.
The Outcomes view tracks the share of merged commits and lines of code with Codex contributions over time.
The Admin plugin in ChatGPT Work lets admins compare adoption, spend, and tasks, then turn findings into reports for budget and rollout decisions. It can also create finished work, such as a leadership deck with charts, key findings, and recommended next steps, ready to share with cross-functional stakeholders.
With the Admin API, teams can automate reports in their own dashboards and combine analytics with business-system data. For example, a support dashboard could show credit use alongside ticket resolution time.
The Admin plugin compares team credit trends and summarizes changes for a monthly rollout review.
Connect analytics to business outcomes
Usage and task data are a starting point for admins to investigate value with business owners. Business owners add the context needed to assess it: what changed in the workflow, whether results improved, and what that improvement is worth. Together, they can connect product activity to measures such as delivery time, quality, or profitability.
Suppose the task classifier in the Admin Console shows that account research is common in a sales group. The admin checks whether model choices fit the task, provides training where a CRM plugin is underused, and shares a skill for consistent account plans. The sales owner then compares preparation time and plan quality with the team’s baseline. If time is saved, they track how much goes into customer conversations, which conversations become qualified opportunities, and which opportunities become sales. Revenue and contribution margin on that business help show whether the extra capacity produces a financial benefit.
To explore the value of a workflow, start with a few questions:
What would you like to improve? Choose an outcome that matters to the team, such as faster preparation, better-quality work, lower costs, or more sales.
How does the process look today? Establish a starting point: how often the team does the task, how long it takes, and what a good result looks like.
What changes with AI? Compare results over a defined period. Include the time spent reviewing and correcting the work so that faster completion still meets the team’s quality standards.
What does this make possible for the team? Time saved might mean more conversations with customers. Better-quality work might mean fewer corrections. Talk with the team to understand where those improvements matter most.
Is the benefit worth the investment? Compare what the team gains with what you’re spending on AI, setup, and ongoing support. That can help you decide what to expand and where the team could use more help.
Illustrative annual ROI: sales account research
Imagine a team of sellers, each preparing two account briefs per week. Every brief brings together account history, industry research, and a point of view for the next conversation.
What changes with AI?
What does that add up to?
Annual time saved
20 sellers × 2 briefs per week × 3 hours saved × 46 weeks = 5,520 hours
Estimated annual capacity value
Assume the team puts 50% of that time into productive work, valued at a fully loaded employee cost of $75 per hour. Sellers could use that time to connect with new prospects, follow up on opportunities, and spend more time with customers.
5,520 hours × 50% × $75 = $207,000
Total first-year costs
Assume $60,000 for AI, setup, training, and ongoing support.
($207,000 − $60,000) ÷ $60,000 = 245% illustrative ROI.
The value of a workflow can extend beyond productivity. In sales, better account research and a shared understanding of the customer can support deeper conversations, stronger relationships, and new opportunities. Teams can then measure whether those improvements lead to additional business outcomes, like accelerated deal cycles or increased win rate.
All figures are hypothetical. ROI reflects estimated capacity value and excludes potential benefits from higher win rates, larger deal sizes, or other sales outcomes.
The business impact customers are seeing with AI
From shipping software to running live events, customers are putting AI to work on tasks that matter to their businesses:
1Passworduses Codex to build, review, and test software so engineers can ship features faster. It estimates 553% ROI and $0.8M in annual engineering capacity value, using Codex to build production-ready features and internal tools faster while maintaining rigorous security policies.
ATV Big Air Touruses ChatGPT Work to check event listings, plan inventory, and improve its website’s visibility. Listing reviews fell from eight hours to one hour a week, and inventory work from two to three days to two to three hours, giving the team more time to focus on the tour and its customers.
Playcouses GPT‑6 Astra through OpenAI’s API to build and test playable game prototypes. The team created three themed prototypes from one foundation and reported 50% fewer manual fixes than with the previous model, helping developers test and compare more ideas.
How to get started
Open Insights in the Admin Console and choose a common task that supports a business priority. Review it with a business owner, agree on a baseline and the outcome to measure, and set a date to review progress. Use the results to decide whether to expand the workflow, improve how teams use it, or test a different approach. Learn more(opens in a new window) in our docs.
本文内容采集自官方网站，排版和翻译可能与原页面存在差异。
阅读官方全文