Speckle Unveils Major AI-Powered Update for AEC Teams

Summary

Speckle has released a comprehensive platform update. The update includes a read-only AI assistant, a fully rebuilt WebGPU-based 3D viewer, and chat-based report generation designed for AEC teams managing large, multi-disciplinary model data.

Why it matters

AEC coordination remains bottlenecked by fragmented tools, manual model reviews, and slow cross-disciplinary data extraction. A central hub that can answer natural language questions about published geometry, join data warehouse tables, and generate shareable, live-updating reports could significantly reduce back-and-forth traffic between architects, engineers, and contractors. The WebGPU rebuild also signals an investment in browser-native performance for models that previously required desktop software.

Highlights

• The AI assistant is read-only and limited to published model data. It can execute SQL queries on attached CSV files and Snowflake/Databricks data warehouse tables.
• Any chat response can be saved as a fully editable report with interactive inputs that recalculate charts and KPIs in real time for viewers.
• Viewer 3 (WebGPU) has been rebuilt to handle the loading of large models. The in-viewer AI can directly manipulate the scene: isolating objects, adjusting section planes, and preparing camera tours.

Critical take

The article is essentially a feature list dressed up as journalism: there are no independent benchmarks, no user case studies, and no discussion of failure scenarios (such as the AI misinterpreting complex BIM semantics, data privacy concerns with third-party warehouse connections, or what happens if a model is not published). Without at least one skeptical counterpoint or performance comparison, the text reads like a press release rather than an editorial assessment.