Ask your data
A setup is a fixed scope at a fixed price, agreed with you after the AI access review. Ask your data combines three parts: data pipelines that load your systems into one data warehouse in your own cloud, an MCP server that only reads that warehouse, and the AI tool your people already use. They ask about the numbers and get the answer with the records behind it.
This is for you if
- A question about costs waits until someone has exported data from several systems.
- The same person builds the spreadsheet every time someone asks.
- Your people already use Claude or another AI tool, and it cannot see your own figures.
What you leave with
- The result
- Answers about your figures, with the records behind them. An employee asks in the AI tool they already use, for example "Why did costs rise in Sweden last quarter?", and gets the answer with the figures and the records it rests on. You ask, the agent helps.
- One warehouse
- Data pipelines load your systems into one data warehouse in your own cloud on a schedule, so every answer rests on the same data as your reports.
- Read only
- The MCP server only reads. Each person signs in as themselves and sees only the datasets their role may see.
- Traceable
- Every answer is saved with its query and the tables it read, and every call is written to the logbook.
What we do, what you do
| We | You |
|---|---|
| Agree the scope with you after the AI access review: which systems, which questions, who may see what, and the fixed price. | Bring the questions your team asks most about your figures. |
| Build the data pipelines and the data warehouse in your own cloud. | Give read access to each source, through someone with admin rights, and check the first figures against a report you trust. |
| Build the MCP server, read only, with sign-in per person and a tier per role, and set reading allowed and writing blocked in your AI tool. | Decide which datasets each role may see. In Claude Team, an owner of the Claude account adds the server once as a connector. |
| Hand over the code and the settings. Gathena's agents build and maintain it, and every change waits for your approval. | Own all of it, in your own cloud. |
What it does not do
It answers questions and writes nothing back, to the warehouse or to your systems. Before any figure leaves the company, a person approves it. When the same questions come back in the same way, a regular report or an AI agent can take them over, built on the same warehouse.
Your questions
- Which AI tool do we need?
- Claude Team or Enterprise, where an owner of the Claude account adds the MCP server as a custom connector. Other AI applications that speak MCP, such as ChatGPT, can reach the same server. Your people keep working in the tool they already use.
- Can it see everything?
- No. It sees only the datasets you open to it, and it only reads. Each person signs in as themselves, each role gets its own datasets and limits, and a person's access can be withdrawn in one place.
- What does it take from us?
- The AI access review first: two weeks at a fixed price, credited against the build if you continue. Then read access to your sources from someone with admin rights, the questions your team asks most, a check of the first figures, and an owner of your Claude account to add the connector.
- What happens after?
- Your people keep asking, and Gathena's agents maintain the pipelines and the server for a fixed monthly fee, with every change waiting for your approval. When questions repeat the same way, a regular report or an AI agent can take them over.