Speckle Launches AI Assistant and WebGPU Viewer Update

Speckle has updated its platform with a built-in AI assistant designed specifically for architectural and civil engineering data. According to the company, the assistant operates in read-only mode using data already published to the platform. Rather than offering open-ended chat functions, it extracts information regarding model versions, user update histories, metadata, and project comments, delivering responses as plain text, on-demand charts, key performance indicators, or structured tables.

To support broader project analysis, the platform now integrates external files and enterprise databases. Workspaces include a data library where uploaded CSV and Excel documents are converted into queryable tables. Speckle states that users can also connect Snowflake and Databricks cloud warehouses, allowing the AI assistant to join live enterprise datasets with spatial model data. Furthermore, the viewer now condenses dense software-specific data, such as Revit material quantity logs, into cleaner rows.

The update allows project teams to convert assistant answers directly into editable, living reports. These saved pages can combine key indicators, text blocks, embedded 3D views, and dynamic numeric controls such as currency or percentage inputs. Speckle claims that modifying these input parameters immediately updates all associated calculations. Modifications to reports can be guided by AI prompts on working copies, allowing users to review or undo edits before publishing changes across the project workspace.

Inside the viewer interface, the AI assistant can execute camera and visual operations based on live workspace context. Speckle says the tool can isolate components, apply property-based color schemes, slice section planes, or generate background visual tours without disrupting active user sessions. For repetitive analytical tasks, teams can save custom prompt workflows as reusable commands, known as Skills. Typing a slash command triggers these saved instructions, which can be shared across the entire workspace.

Alongside the artificial intelligence features, Speckle released Viewer 3, a redesigned WebGPU rendering engine that acts as the platform's new default viewer. The software replaces older stream formats with compact version artifacts and employs a hybrid streaming method, displaying geometry over a dedicated line while downloading full files in the background. Subsequent views draw from local cache, while the underlying scene tree uses lazy rendering to keep large infrastructure models responsive during navigation.

To maintain oversight, Speckle has introduced workspace-level governance controls for all artificial intelligence capabilities. Workspace administrators must explicitly activate the feature set and accept platform terms before users can access it. Usage is metered through a shared workspace credit pool, supported by an activity dashboard. Furthermore, administrators can enforce standing system rules—such as fixed unit systems or regional standards—that automatically inject into every AI query across the organisation.

Why it matters for AI for Construction & Civil Engineering Excellence

Civil engineers and construction managers handling complex federated models often struggle to query object data and share progress without manual exports. By combining contextual AI queries with enterprise database links and an upgraded WebGPU viewer, Speckle enables teams to inspect heavy models, verify properties, and automate reporting directly within a browser without exporting raw files to external software.

What to do about it

• Review workspace permissions and opt into the AI capabilities through the admin panel.
• Upload project CSVs or link Snowflake warehouses to test joint data queries.
• Create standardized analytical prompts and save them as reusable workspace Skills.

Tools mentioned

• Numeric
• Beyond

Source: Speckle Unveils Major AI-Powered Update for AEC Collaboration – parametric-architecture.com (parametric-architecture.com)