Skip to content

AdminIdea

The record of AI use

The record of AI use counts every call your organisation’s models answer. The record keeps who each call was for and what it cost. On Usage, you see what your organisation spent each month, and what the models were used for.

Your organisation pays its providers directly, and their bill is one number. It doesn’t say how much went on support’s customer questions, and how much on an agent reading the code. The record does. Whoever pays can see what a month went on. They see which people and initiatives used the most, and whether moving routine work to a cheaper model paid off.

Emargen records every call. People’s chats with the agent count. So does the work nobody watches, like an agent task’s weekly run, an analysis of your code, or the title the app writes for a new topic. Usage groups them by kind of work, like Chats, Coding agents and Background writing.

Each line names who it was for, or nobody when Emargen worked on its own. It also names the initiative, the topic and the price. Answers that stopped or failed leave a line too.

Cost is counted at list prices. Those are the prices someone who manages models saved for the model, or an open catalogue’s prices. We left discounts out for now. Many organisations have them, but list prices are enough for an indication.

If a model has no price, you see its tokens and no cost. When prices change, the calls already recorded keep the prices they were charged at, so a month’s cost doesn’t change after the fact.

Usage for October 2026 at ACME: Cost at list prices, Tokens and Calls at the top; By activity with Chats, Coding agents, Background writing, Agent tasks and Judgments; By model with GPT-6 Sol, GPT-6 Astra and Jev; By person and By initiative with the top five of each, and under them the lines Background work not tied to a person and Not tied to an initiative (topics, the Inbox, settings).Usage for October 2026 at ACME: Cost at list prices, Tokens and Calls at the top; By activity with Chats, Coding agents, Background writing, Agent tasks and Judgments; By model with GPT-6 Sol, GPT-6 Astra and Jev; By person and By initiative with the top five of each, and under them the lines Background work not tied to a person and Not tied to an initiative (topics, the Inbox, settings).
A month at ACME: what it cost, what it went on, and the part that wasn't about any initiative.
  1. Cost at list prices, with no discounts counted.
  2. By activity: what the month went on.
  3. By model: which models it ran on.
  4. By person: who used the most.
  5. By initiative: which initiatives the work was for.
  6. What wasn’t tied to an initiative: work in topics, the Inbox and settings. Here, about half the month.

Say your organisation handles most customer questions in topics. When the bill comes, you open Usage and find most of the month under Chats, outside any initiative. You move the chat default to a cheaper model. Next month’s Usage shows the difference.

Usage is a page under Models, for people who may manage models. Everyone sees their own month on their profile, with their share of the organisation’s. An agent task’s runs show the tokens each one used, to its owner and the people who may edit it. Nobody edits the record.

Cost by topic isn’t shown, though every line keeps its topic. We left it out because private and shared topics are hard to tell apart there, and there are far more topics than initiatives. An initiative’s cost isn’t shown to its people.

The record limits nothing. Only flows can have a monthly limit on their model work, set under Models. Usage also leaves out what runs on Emargen’s own models, like the admin agent. People’s own assistants, connected through agent access, run on their own accounts and aren’t in it.

How to

Guides

Page by page