Model access
Approve model A for the team
The team can use model A. The gateway rejects requests to models that have not been approved.
AI platform
Use AI together. In your own cloud account.
We develop and operate your platform for chat, development tools and existing automations. You decide on model access, data routes and budgets.
Use
For people and automated workflows
Three ways to use it, one set of rules: a shared model gateway connects everyday use with control by your IT team.
Does the platform fit your situation?
You provide AI for teams or processes and want to attribute usage and costs. We assess what your tools cover and where the platform adds something.
What you can decide
Each request passes through the gateway. These three examples show how a setting affects what happens next.
Team budget
Set a monthly budget for the team
Once the budget is exhausted, the gateway stops further requests until the next budget period.
Process key
Give the inbox process its own key
Model costs for this process appear under its own key. They can be distinguished from chat and other workflows.
Data route
Where each kind of data goes
Usage data, chat histories and model processing have different locations and access rules. Defining these is part of setup.
How requests and responses travel
Chat, development tools and existing automations share one model gateway. This diagram shows the data route when using an external model provider.
Your applications
Chat, development tool or existing automation
Request
Response
Your company’s cloud account
The gateway checks model access and budgets. Configuration and usage data stay in your account.
Approved request
Model response
Approved model provider
The provider processes the approved request and returns the model’s response.
Together we define which data may be sent to a model provider. Owning the account does not mean that every processing step happens in your cloud account.
- Usage data
- Gateway in your cloud account
Key, timestamp, model, tokens and cost. Administrators see usage across all teams; users see their own.
- Chat histories
- Chat in your cloud account
Conversation histories for each user. Access permissions for chat are defined separately.
- Requests and model responses
- Approved model provider
The provider processes the approved content. Together we define the provider, the data sent and the processing terms.
The gateway does not store prompts or responses. Reporting focuses on teams, processes and models. You set the rules for any reporting about individuals.
Cost attribution
Model costs per team over 14 days
In this example data, Sales makes more requests than Engineering but incurs lower model costs. Automation appears as a separate item.
The bars show one measure at a time. Requests and USD have separate scales.
Values as a table
| Team | Model cost in USD | Requests |
|---|---|---|
| IT | 105.87 | 1,113 |
| Engineering | 28.15 | 677 |
| Sales | 4.99 | 889 |
| Automation (bot) | 2.21 | 182 |
What your team receives in the pilot
We start with one unit and support its use of the platform. Before starting, we agree on tasks, access and acceptance criteria.
- net
- 8,900 EUR
- months
- 3
- unit
- 1
Three months, one unit with up to 25 approved users, 8,900 EUR net at a fixed price. A unit has its own model permissions, budgets and access rules. It can be a department, a team or an automation team with its processes.
Prepare
An agreed unit with tasks, model permissions and a budget.
Your contributionYou name business and technical contacts and provide the necessary access.
Set up
The gateway, chat and agreed connections run in your cloud account.
Your contributionYour IT team checks sign-in, permissions and data routes with us.
Use and review
Your team uses the platform. We handle technical questions and adjust the agreed rules.
Your contributionYou give feedback on the agreed tasks and assess whether the answers are fit for their purpose.
Hand over
Source code, configuration and documentation go to your IT team.
Your contributionYou decide on an extension and agree with us who will operate the platform next.
Included
the gateway in your cloud account, chat for the unit, connections to agreed development tools (where they support their own API endpoint), model access for existing automations through their own keys, usage and cost tracking per key and team, a rule set for model use, and the handover with source code, configuration and documentation.
Not included
cloud infrastructure and model usage (paid directly to the respective providers), connections to documents and knowledge sources (available separately), new agents and connections to business applications (separate proposal after scoping).
Afterwards
For a full rollout, we credit the full pilot price against setup. Without a rollout, the platform stays in your account; cloud infrastructure and model usage continue to accrue, and before the pilot ends we agree who takes over operations.
Rollout and pricing
Rollout and ongoing operations
A rollout adds further units or knowledge sources and the agreed setup for ongoing operation. The pilot price is included in setup and credited against it.
Platform
up to 3 units, no knowledge sources
- Total setup
- 24,000 EUR
- Operations per month
- 2,200 EUR
Platform + Knowledge
up to 5 units, 2 knowledge sources
- Total setup
- 46,500 EUR
- Operations per month
- 3,000 EUR
Company-wide
up to 10 units, 4 knowledge sources
- Total setup
- 66,500 EUR
- Operations per month
- 4,250 EUR
Within the included units, we do not charge per person. A knowledge source is a defined repository, such as a SharePoint site collection with one permission structure.
All prices net. Cloud infrastructure and model usage are additional.
| Additional unit | Setup, one-time | Operations per month |
|---|---|---|
| Additional unit | 2,000 EUR | 150 EUR |
| First knowledge source | 13,500 EUR | 250 EUR |
| Each further knowledge source | 5,000 EUR | 250 EUR |
Prices for further knowledge sources apply when they use the same integration method. Business systems and new integration methods are priced separately.
Operations
Who takes responsibility for each part
We maintain the installation and agree technical changes with your IT team. Your company keeps the accounts and decides what is approved.
- monthly updates for the gateway, chat and security
- keeping the model catalogue and prices up to date
- data backups with a restore test
- changes to budgets and usage rules
- a monthly report on usage by team and model
Your company
- Hold the cloud account and provider contracts
- Approve models, data routes and budgets
- Define business tasks and name contacts
bitvaria
- Set up and maintain the gateway and chat
- Carry out updates, backups and restoration tests
- Adjust agreed rules and report on usage
Model provider
- Process approved model requests
- Bill for model usage based on consumption
Questions about the platform
Does the platform run in our cloud account?
bitvaria develops and operates AI platforms in the customer’s cloud account, not as SaaS in a provider’s account. The gateway and chat run in your account. Configuration and usage data are stored there. Source code, configuration and accounts belong to you, even if the engagement ends.
Can we start with just one unit?
bitvaria delivers this as the platform pilot: three months, one unit, a fixed price of 8,900 EUR net. Cloud infrastructure and model usage are not included. The pilot price is credited in full against the setup fee for a full rollout. If you decide against a full rollout, you keep the platform.
Which team would use the platform?
Briefly describe your current tools and the tasks you want to run through them. We work out together whether the pilot fits.
Discuss your project