Getting model data AI-ready: Two early movers in AEC data share what actually moves the needle

Would you trust AI trained on your firm's models today?

Everyone is talking about the potential of AI, but the reality is that most design data is still trapped in models that are full of errors, gaps, and inconsistencies. Like many other AEC firms, you’re likely sitting on mountains of data you’ve never used. Not because it's missing, but because you’re waiting for it to be clean enough and don’t know how to get started on fixing it.

In this session, two firms that invested early in getting their data working—Murat Melek from Suffolk Construction and Ryan Haunfelder from HGA—share what their teams have done to start extracting value from their BIM portfolios: the mistakes they made, the techniques that worked, and why broad, lightweight extraction across a messy portfolio consistently is where to realistically get started.

This session is for you if:

• You're responsible for your AEC firm's AI or data strategy and wondering how to bring in 3D/BIM data in a reliable, scalable way
• You're a practitioner who's tried to wrangle BIM data into something useful and hit walls (or multiday export processes)
• You're afraid your BIM data "isn't ready for AI" and you're not sure if it ever will be

You'll leave with a concrete framework for validating and conditioning design data from both historical and active projects, plus a live look at how Speckle makes that process continuous, automated, and actionable.