---
title: "AI workspace: an employee chat paid by the tokens used"
description: "The AI workspace is a chat front end for employees on the AI gateway. You pay the provider for the tokens used instead of a seat licence for everyone."
url: https://www.cureman.de/en/modules/ai-workspace
language: en
updated: 2026-11-16
alternate_de: https://www.cureman.de/de/module/ki-arbeitsplatz.md
---

# AI workspace: an employee chat paid by the tokens used

> The AI workspace of Cureman is a chat workspace for employees under the organisation's own name. A chat front end, Open WebUI by default, is connected to the AI gateway, people choose from the models the organisation allows, and usage is attributed to each person. The organisation pays the model provider for the tokens actually used instead of a flat seat for everyone. Each month the module computes who used the workspace, what a seat licence for each of them would have cost and what their tokens cost, and the difference is the saving.

A chat seat costs the same every month, whether its holder sends five messages or five hundred. The AI workspace gives employees a chat front end that runs through the organisation's own gateway. The organisation then pays the model provider for the tokens actually used, and sees per person what that came to.

## What it is

The workspace is a chat front end connected to the [AI gateway](https://www.cureman.de/en/modules/ai-gateway). By default that front end is Open WebUI, deployed as a separate application for each organisation. Its OpenAI-compatible connection points at the gateway and uses a workspace key. Any other front end that speaks the OpenAI chat format against a base URL and a key works the same way.

The front end forwards the signed-in user with every request. That is what lets the gateway attribute usage to a person and apply a budget per person. The provider bills the tokens to the organisation's own API key, as for any other gateway traffic.

## What the organisation controls

| Setting | Effect |
| --- | --- |
| Models | The model list of the workspace key is what people can choose from in the front end |
| Budget per person | A monthly budget per user on the key; once it is used up, further requests are refused until the next month or until an administrator raises it |
| Routing | Routing policies apply as for any other gateway key, so a simple request can be served from a cheaper model |
| Name and address | The name employees see and the host name where they open the workspace |

How far the front end's own branding can be replaced depends on that front end's licence terms.

## What it shows

For the current calendar month the workspace page lists the people who used it, with requests, tokens and provider cost per person, and the same per model, including how many requests a routing policy served from a different model. Names are shown only to roles that hold the permission to view people.

## The comparison with seat licences

1. Everyone who made at least an agreed number of requests in the month counts as one seat that was not needed.
2. That number of people times the agreed seat price is what the seats would have cost, pro rata while the month is still running.
3. The provider cost of all workspace tokens is subtracted, including the calls of a classifier or selector.
4. The difference is the saving. In a month where the tokens cost more than the seats would have, there is no saving and none is counted.

The seat price is not measured. It is the licence agreed as the comparison in the module's terms, which is what the organisation pays, or would pay, for one chat seat. Every step is shown with its formula on the workspace page. How to work out the break-even for one person is described in [AI chat: seat licence or pay per token](https://www.cureman.de/en/articles/ai-chat-seat-licence-or-pay-per-token).

## Moving people from seats to the workspace

Whether the workspace would do for someone who holds a chat seat today is a question [licence optimisation](https://www.cureman.de/en/modules/licence-optimisation) answers, where that module is booked. The playbook of a chat service compares each holder's measured use with what the same use would cost by token. Where the data cannot decide, a campaign asks the holders whether the workspace would cover what they do.

That campaign is drafted and goes out only after a person has approved it, and nobody on the organisation's do-not-contact list is asked. People who say the workspace would do get access, and removing their seat is a step of an action plan that a person approves. Everyone else keeps their seat.

## What it needs and what it measures

The AI workspace requires the AI gateway, with an upstream in the OpenAI format, because that is the format chat front ends speak. [Prompt insights](https://www.cureman.de/en/modules/prompt-insights) are optional and add the task mix per person.

The module meters the people who used the workspace in the month, what a seat licence for each would have cost, and the provider cost of their tokens.

## Related reading

- [AI chat: seat licence or pay per token](https://www.cureman.de/en/articles/ai-chat-seat-licence-or-pay-per-token)
- [AI budgets and limits](https://www.cureman.de/en/articles/ai-budgets-and-limits)
- [Allocating AI costs to teams](https://www.cureman.de/en/articles/allocate-ai-costs-to-teams)

## Questions and answers

### Which chat front end does the workspace use?

Open WebUI by default, deployed as a separate application for each organisation and pointed at the gateway with a workspace key. Any chat front end that speaks the OpenAI chat format against a base URL and a key works the same way, because the gateway does the metering and the limits.

### Who pays for the tokens?

The organisation pays its model provider directly, because the gateway forwards every request with the organisation's own API key. The gateway meters each request, computes its cost from the model price list and attributes it to the person who sent it.

### What does the comparison with seat licences rest on?

On three things. The seat price is a figure agreed in the module's terms, the licence the organisation pays or would pay per chat user, and it is not a measurement. The users are counted from gateway requests, everyone at or above an agreed number of requests in the month. The token cost is metered. The same requests always give the same result.

### Can use per person be limited?

Yes. A workspace key can carry a monthly budget per person and a list of allowed models. When a person's budget for the month is used up, further requests are refused until the next month begins or an administrator raises it. The front end offers only the models on the key's list.

### Does the gateway store the chats?

No. The gateway stores token counts, cost, the task type and the user reference for each request, and no prompt or response text unless the organisation switches prompt samples on. Chat history is kept by the chat front end, which the organisation deploys itself.

## Read next

- [AI chat for employees: seat licence or pay per token](https://www.cureman.de/en/articles/ai-chat-seat-licence-or-pay-per-token): How to compute, per person, whether an AI chat seat licence or paying per token is cheaper, with the break-even formula and a worked example.
- [AI budgets and limits: per key, per user, per model](https://www.cureman.de/en/articles/ai-budgets-and-limits): How monthly budgets per API key and per user and model allow-lists work for LLM spend, what happens at the limit, and how to set the amounts.
- [How to allocate AI costs to teams, projects and people](https://www.cureman.de/en/articles/allocate-ai-costs-to-teams): A method to attribute LLM spend to teams, features and people with keys, tags and user identifiers, price every request and reconcile with the provider.
