The data foundation, explained
What does a data foundation for AI agents look like? Pipelines bring data from the systems you already run into one warehouse in your own cloud, and every figure traces back to its record. Every knowledge worker reads from that warehouse. Building it is the first thing worth doing, whatever you decide about AI.
Start with the data foundation
Get your data in order. That is the data foundation.
A five-year AI plan is impossible; the technology will not look the same in five years. The useful question is: what can we do now that helps us today and prepares us for a future we do not yet know? The answer is almost always the same: get your data in order.
AI models are rocket engines. Your data is the fuel, and no AI solution is ever better than the data under it. The data foundation covers master data, CRM, ERP, the accounting data and the files. Putting it in order helps you twice: you get better reporting and better decisions now, with or without AI, and you are ready for whatever comes next.
We build the data pipelines first. We connect the systems you already run and bring their data into one warehouse in your own cloud, Google Cloud or Microsoft Azure and Fabric, where every figure traces back to the record it came from. The pipelines run on a schedule, keep the raw files as evidence, and tell someone when a source changes. Knowledge workers read from that warehouse, so an answer to "why did costs rise" uses the same figures as the month-end report. You control access, the data stays in your cloud, and nothing moves onto a platform we own.
If you have watched a data project overrun, the cause was usually the manual work: people structured, cleaned, and enriched the data by hand. AI does that work now: it reads the unstructured text in emails and documents, standardises and de-duplicates records across systems, and adds detail where records are thin. We use AI to build the data foundation, so if you turned down a data project before AI, that verdict no longer holds.
The stacks we build in
See where we can help in the stack you already have.
Google Cloud
- Google Cloud
- BigQuery
- Cloud Run
- Workflows
Microsoft
- Microsoft 365
- Azure
- Fabric
- Business Central
- Power BI
- Copilot
AI tools and connections
- Claude
- Gemini
- OpenAI
- Model Context Protocol
- GitHub
- Cloudflare
Product and company names are trademarks of their respective owners, used to identify the products referred to. No endorsement or partnership is implied. Microsoft, Microsoft 365, Azure, Fabric, Dynamics 365 Business Central, Power BI, SharePoint and Copilot are trademarks of the Microsoft group of companies. Google, Google Cloud, Google Workspace, Gemini, BigQuery and Cloud Run are trademarks of Google LLC. Model Context Protocol is a trademark of LF Projects, LLC.
Cases
Cases: what we built and how.
DuffBeauty on Google Cloud
An accounting agent that prepares the entries; the bookkeeper approves them.
Before: To answer a question about costs, someone exported data from three systems, and it took two days. After: The employee asks in Claude Team and gets the answer, with the records behind it.
Our own tool · Google Cloud or Azure
An invoice reader that extracts the invoice lines and checks them.
Before: Every invoice line was typed by hand, or read by a service priced per page. After: The reader extracts the lines; the bookkeeper approves a batch and reviews the flags.