The AI layer on Yres: run your data warehouse in a conversation
The wholesaler runs its data warehouse on Yres, Plainwater's data warehouse automation platform on Azure. Purchasing wants the supplier portal added. That means adding a source, picking tables, deciding keys, loading, checking and building a report.
With Plainwater AI Noor does that in one conversation. The agent works with Yres through the Yres MCP server and knows the rules of Yres: load types, history and changes. It does nothing outside the change process.
Updated .
What is the AI layer on Yres?
The AI layer on Yres is Plainwater AI connected to Yres through the Yres MCP server, and to the Yres warehouse as a data source. The agent manages the organisation in Yres and reads the warehouse under the same rules as any other source.
Yres itself stays what it is: the platform that loads sources into a data warehouse in your own Azure environment, with history and a fixed path from development to production. The AI layer adds the conversation, not a second way of loading.
What can the agent do in Yres?
The agent can do whatever the Yres MCP server offers: view sources, set up tables, follow loads and manage changes. In Noor's conversation it looks like this:
- Look at the source. The agent fetches the organisation's data sources and refreshes the supplier portal's metadata.
- Pick tables. It shows which tables exist and which are not loaded yet.
- Propose keys. Yres predicts the key columns; the agent presents the proposal and Noor confirms.
- Set up inside a change. The agent creates the used tables within an open change in the development environment.
- Load and check. It follows the load history and pipeline runs, and explains what went wrong when something fails.
Deploying to test and production then goes through the normal Yres process. The agent does not put a change into production itself.
Why does the agent need to know the rules of Yres?
Because the right load type, key and history differ per table, and a wrong choice only shows weeks later. An agent that only knows SQL will not see it.
The agent works with the concepts of Yres: load types such as full and delta, history per change, changes and projects. Every table change belongs to an open change, and changes only happen in the development environment.
| Noor asks | What the agent uses |
|---|---|
| Which tables are we not loading yet? | List of unused tables per source |
| What is the key here? | Yres key prediction, presented to Noor |
| Why did last night's load fail? | Load history and pipeline runs |
| Add this table | Create a used table inside an open change |
How do you get from a loaded table to a report?
The Yres warehouse is an Azure SQL database, so for Plainwater AI it is a regular source. Everything that works on a source works here: quality checks, performance and dashboards.
Once the supplier portal is loaded, Bram asks for lead time per supplier. The agent builds a dataset on the warehouse and a dashboard that refreshes on a schedule, without running an AI model each time.
Data access to the warehouse follows the platform's modes. With the source on metrics only, the model sees counts and metadata and no supplier names.
Where it stops
The AI layer works through the Yres MCP server and can therefore do what that server offers. Closing, releasing and installing changes happens outside the tools, in Yres itself.
The link between Plainwater AI and Yres is made of building blocks: connecting an external MCP server and the warehouse as a source. Ask in a demo what that looks like for your environment.
If you do not use Yres, you miss nothing of the rest of the platform. This page describes an extra layer for Yres customers.
Frequently asked questions
Do I need Yres to use Plainwater AI?
No. Plainwater AI works on every supported source. The AI layer on Yres is an extra option for organisations that already run their warehouse on Yres.
Can the agent put changes into production?
No. The agent sets up tables inside an open change in the development environment. Closing, releasing and installing go through the normal Yres process.
How does Plainwater AI connect to Yres?
Through the Yres MCP server, as an external MCP server in the platform, and through the Yres warehouse as an Azure SQL source.
Does the AI model see my warehouse data?
The mode of the source decides. In the default mode the model gets metadata and counts, no rows or business values.