Organization intelligence¶
Organization intelligence helps an authorized leader understand how HiveAI is being used, where information is missing, and which operational or financial controls need attention. It brings together usage, quality indicators, credits, budgets, knowledge-source health and governance.
The purpose is to support decisions about the service. These figures do not measure individual productivity, establish employee performance, prove answer correctness or demonstrate a financial return.
Availability and intended experience¶
The backend contains a six-section organization dashboard and separate tools for selected aggregate work metrics. The inspected frontend still has HiveAI entry disabled and no standalone HiveAI application. Final Screens shows an intended conversation about organizational use with some measures unavailable. It does not establish that the full dashboard or its navigation is deployed.
The task guide follows that intended conversation and explains the information an authorized reader should check. Current delivery gaps affect whether aggregate numbers can reach an assistant answer. Treat designed sample numbers as examples.
Who can use this information¶
The dashboard requires an organization head or a person with the relevant governance authority, together with advanced HiveAI and tenant-metrics permissions and current organizational policy. Being a manager, administrator or executive by job title does not automatically grant those permissions.
The separate organization work-metric tools currently require organization-head visibility. Their richer analytics path additionally depends on HiveOrbit access. Access to the dashboard therefore does not guarantee access to every metric that can be requested in conversation.
Ordinary members and external participants cannot acquire organization-wide visibility by asking for it. See HiveAI access and availability and analytics scope.
Basic activity totals¶
A separate basic read summarizes recorded HiveAI audit activity for a chosen period. Organization-wide totals require the organization head and the tenant-metrics permission; delegated authority that permits the advanced dashboard does not automatically permit this basic organization-wide read.
These totals count saved audit records, their recorded success/refusal labels, tokens, citations and incident references. They do not count unique people, complete business outcomes, distinct incidents or every attempted request. A successful label does not establish answer correctness. Missing records cannot be recovered by interpreting a zero as proof of no activity.
The period includes its start and excludes its end. The ordinary request defaults to the preceding 30 days, accepts a positive period up to 366 days, and rejects an end in the future. Its freshness label uses the requested end; it is not an independently measured last-update time. The basic read has no minimum-group suppression rule of its own, unlike the richer dashboard and work-metric controls described below. Its organization-head boundary still applies.
This is an explanation of implemented behavior. The existing task guide describes inspected intended screens; it does not establish a separate connected basic-metrics page. Staff-wide summaries have different populations and accounting periods; see governance and operational metrics.
What the six sections mean¶
| Section | What it helps you inspect | Main interpretation limit |
|---|---|---|
| Adoption | Recorded executions, participating subjects, successful use and capability codes used | The population comes from audit records, not the full licensed workforce |
| Quality | Citation presence, refusals and recorded tool-safety outcomes | Grounded-answer evaluation is unavailable; safety is not accuracy |
| Cost | Tenant credit balances and a calculated settled-credit difference | The underlying balance reads do not establish historical period spending |
| Budgets | Applicable configured budget scopes, warnings and limits | The scopes are resolved for the current reader, not necessarily every organization budget |
| Knowledge-source health | Processing backlog, quarantines, retries, active sources and removals | This is the latest received snapshot, not a historical view for the selected period |
| Governance | Audit incident references, policy identity and current suspended controls | Counts and timestamps combine different kinds of evidence; they are not a compliance certification |
A section can be unavailable while another contains data. Missing, suppressed and zero values have different meanings. Preserve each section's explanation and freshness when sharing a summary.
Business rules and privacy¶
Small cohorts can be suppressed to reduce the risk of identifying people. The minimum is at least five and organizational policy can raise it. Further checks reject individual-ranking or employment-decision requests and unsafe result content.
Suppression is an intentional information boundary. Do not try to reveal an individual by changing filters or subtracting related totals. An aggregate figure remains subject to its definition, coverage and time basis.
The work-metric source reads a bounded set of contribution records. Its counts are not a complete inventory of current tasks or a count of unique completed business outcomes. Use the owning WorkOS and HiveOrbit views for their documented purposes.
What happens behind the scenes¶
The dashboard reads local HiveAI audit and policy information, obtains governed budget and credit information, and reads the latest knowledge-processing health projection. Source-health updates arrive asynchronously after ingestion changes. A dashboard read does not ingest documents, change a budget, resolve an incident or update a work item.
Conversation work remains subject to ordinary HiveAI permissions, provider availability and credit controls. Do not infer that an organization-intelligence question is free, or convert microcredits into currency without the applicable commercial information.
Decisions this feature can support¶
Use the information to identify questions for follow-up: whether use is increasing within comparable observed populations, whether knowledge processing needs investigation, whether a budget warning needs attention, and whether missing evidence prevents a decision.
Before comparing periods, confirm that scope, metric definition and coverage are comparable. The current cost-period limitation prevents treating the calculated settled-credit difference as a verified period expense. A quality indicator cannot establish that answers are correct.
Related guidance¶
Continue with read and interpret organization intelligence, conversations, governed knowledge and organization analytics. Briefings, forecasts, budget administration and governed operations have distinct ownership; their presence in a summary does not establish those separate actions or journeys.
From organization information to a briefing¶
Executive briefings reuse the organization dashboard's facts and their interpretation limits, adding a separate process forecast and optional recommendation. The briefing guide explains why a preview differs from a produced occurrence and why a ready result can still contain unavailable measures.
Interpret credits and budget rows¶
Use credits, budgets and usage to distinguish wallet funding from monthly scope limits. The guide explains overlapping scopes and why current-balance reads are not historical spending reports.
For individual work items, use case, project, meeting and leave facts. Those bounded reads have different access, date and freshness rules from organization metrics, and their current assistant delivery has explicit limitations.