You prepare common data environment (CDE) data for AI by making it complete, consistent and owned. That means bringing content from every platform and project into one record, translating metadata into a single schema, preserving version history, making scanned content searchable with OCR, and holding the result somewhere you control. The Utopiaverse™ does this in your own Azure tenancy, ready for AI Insights to search and answer questions.
Why preparation matters
Many AI pilots in construction stall because the data is incomplete, inconsistent or scattered. The model is rarely the problem. A pilot that indexes one project in one platform gives partial answers. A pilot that ignores metadata cannot tell an issued drawing from a superseded one. A pilot that skips scanned content misses much of the record. Preparing the data properly is what turns AI from a demonstration into a tool people trust.
The three conditions
Complete means the record covers every platform and project that matters to the questions you want answered, including completed projects where much of the organisation's knowledge sits. Consistent means metadata from each CDE is translated into one schema, so status, revision, discipline and document type mean the same thing everywhere. ISO 19650 provides a practical basis for that schema. Owned means the record sits in a tenancy you control, in the region you choose, so you decide how it is used and who can query it.
How to prepare CDE data for AI, step by step
Define the questions AI should answer and the teams who will use it.
Inventory the CDEs, projects and repositories that hold the relevant content, including completed work.
Agree a common metadata schema based on the ISO 19650 fields your projects use.
Set up the Azure tenancy and region that will hold the record.
Bring content across: CDE Sync™ for live connections, CDE Backup™ for ongoing copies, CDE Archive™ for completed projects and CDE Migrate™ for platform changes.
Map each platform's metadata to the common schema and preserve version history.
Run OCR on scanned content, index the record and test with real questions.
Common preparation problems
A few issues come up repeatedly. Metadata is missing on older projects, so it has to be derived from folder structures or file names. Different projects use different codes for the same thing, so the mapping needs value translations as well as field mappings. Some content sits outside any CDE, on network drives or in email, and needs to be brought into scope. Permissions vary, so access to the AI record has to reflect who should see what. Working through these early, with a pilot on a representative set of projects, avoids surprises later.
Keeping the record current
Preparation is not a one-off exercise. Live projects keep changing, and an AI record built once will drift out of date. Using CDE Sync™ and CDE Backup™ to keep content flowing from live projects means the record stays current without manual effort. Completed projects can be added with CDE Archive™ as they close. Over time the record becomes the organisation's most complete view of its own work, and AI answers improve as more of the portfolio arrives.
A simple test shows whether data is ready. Take ten questions your teams genuinely ask, such as where a decision was recorded or which projects used a particular product, and check whether the record can answer them with correct citations. Where it cannot, the gap usually points to a missing platform, inconsistent metadata or unprocessed scans. Fixing those gaps improves every later answer, not just the test questions. Repeating the test as more projects are added gives a practical measure of progress that leadership can follow.
Security and governance
Prepared data is valuable, so governance matters. The Utopiaverse™ holds the record in your own Azure tenancy, and AI Insights runs natively on Azure inside it, so content is not sent to a third party to be indexed. Utopia Digital's products stream files without storing them, and Utopia Digital holds SOC 2 Type II and Cyber Essentials. See the Security Trust Portal and the global data region map.
AI-ready data and the Utopiaverse™
Preparing CDE data for AI is exactly what the Utopiaverse™ does. CDE Sync™, CDE Backup™, CDE Migrate™ and CDE Archive™ build a complete, consistent, owned record in your Azure, and AI Insights uses RAG and OCR to answer questions with citations. Related reading: How Do I Sync ISO 19650 Metadata Between CDEs? and How Does RAG Search Work on Construction Documents?
See prepared CDE data in action in the monthly live demo, or talk to the team about your data.
