How we evaluate spatial AI workflows
A spatial AI workflow is only useful if it helps a team move from evidence to decision without losing context.
We do not evaluate spatial AI workflows by asking whether the demo looks impressive. A polished map, 3D scene or assistant panel is useful only if it helps a team complete real work with less confusion and a clearer audit trail.
Our evaluation starts with a task. Find an asset. Review a hazard. Explain a delay. Route a crew. Draft a report. We then follow the path from evidence to decision: what data is visible, what the model inferred, where a person approves the answer and what record is left behind.
The strongest workflows keep spatial context alive. The user should not have to jump from map to spreadsheet to email to PDF just to understand what happened. The system should carry the place, the evidence and the next action together.
That is why our benchmark criteria include task completion, context switching, latency, explainability, human approval and report quality. A workflow that scores well on those dimensions is closer to operational software than a visual prototype.