OpenAI adds workspace analytics

OpenAI rolled out the Workspace analytics refresh on March 12, 2026, formally replacing the prior "User analytics" dashboard. (help.openai.com) The feature set includes aggregated Task Insights, workspace-level benchmarks, and user‑distributed impact surveys (enabled by default), plus a new "analytics viewer" role and cohort filters to drill into top tools and GPT usage. (help.openai.com) Workspace analytics access is limited to owners, admins, and the new analytics viewer role, and the Academy guide explicitly calls out read-only sharing and cohort comparisons as the primary admin workflows for adoption diagnostics. (help.openai.com) Because the dashboard surfaces aggregate adoption and task patterns rather than request-level traces, teams must correlate those signals with infrastructure telemetry — request traces, latencies, error rates, and cost attribution — to investigate reliability and SRE incidents. (academy.openai.com) Enterprise operators running multi‑brand platforms will find Workspace analytics a low‑friction front‑door signal for adoption, but production-grade agent orchestration requires linking those signals into an OpenTelemetry trace+metrics pipeline and an LLM observability layer (examples include Langfuse/LangSmith/Arize) for cost, grounding, and hallucination attribution. (academy.openai.com) Task Insights and impact surveys are enabled by default while workspace owners and admins retain opt‑out controls, and OpenAI’s admin playbooks (updated March 10–13, 2026) recommend using filters and cohort comparisons to diagnose activation → weekly usage → depth drop‑offs before escalating to infra tracing. (help.openai.com)