HiveAI access, capabilities and navigation¶
HiveAI is Zerohive's governed assistance capability. Its access layer determines which supported interactions a person may use in a particular organization. It combines the organization's subscription and policy with the person's permissions, current membership, approved purposes and access to the underlying work. A request to open a manual, read information or prepare an action remains subject to that information's own controls.
This page explains entry and availability. It does not establish that every assistant workflow is delivered in the browser. Conversations, knowledge publication, work assistance and action execution have additional rules that must be documented with their own features.
Documentation status: Source-backed draft awaiting domain and reader review. Backend capability discovery and knowledge-navigation preparation are implemented. In the current account application, a listed HiveAI product has a disabled Not ready yet choice. The frontend runtime catalog does not contain a standalone HiveAI application. These pages do not claim a deployed, working assistant screen. The non-private backend feature families are enabled by default subject to commercial, security, governance, provider and budget controls; Private Cloud remains disabled.
What access means¶
There are several separate decisions between having an account and successfully using assistance:
| Decision | What it establishes | What it does not establish |
|---|---|---|
| Account and workspace access | You are currently recognized as an active person in an active organization | Permission to use every product or read every document |
| Subscription entitlement | The organization has the required Basic or Advanced HiveAI entitlement | Your personal permissions, provider connectivity or remaining spending capacity |
| Workspace policy | The requested capability and feature are enabled for that organization | Access to a specific record, knowledge source or protected action |
| User permissions | Your current permissions allow the requested kind of assistance | Permission to bypass the source product's rules |
| Governance approval | The applicable capability, purpose, provider and jurisdiction context can be checked against active policy | A legal certification or a guarantee that every possible use is approved |
| Provider configuration | The configured model profile passes the checks used by the capability reader | A successful connection to the provider; this reader does not make a live model request |
| Client and environment availability | A suitable application has been implemented and delivered | Approval for every request made from that application |
These decisions can change independently. A person can have a valid account but lack HiveAI permission. An organization can have the appropriate entitlement while its approved provider is unavailable. A backend feature can be enabled while its browser entry point is unfinished.
For account entry and workspace selection, see profile, workspaces and product access. Selecting a workspace establishes the organization context; it does not transfer permissions from another organization.
Basic and Advanced access¶
Basic access is the prerequisite for the capability-discovery entry point and the knowledge-navigation operations described here. The access checks require a Basic or Advanced entitlement, enabled Basic policy and the person's Basic permission. Advanced entitlement alone does not supply a missing user permission.
Advanced operations additionally require the Advanced entitlement, enabled Advanced policy and Advanced permission. The current capability summary treats organization intelligence, action history and reusable-workflow reading as Advanced capabilities. Each operation still has its own feature, role and source checks.
For example, the organization-intelligence dashboard summary requires permission to read tenant metrics and either organization-head status or permission to manage HiveAI policy. Being an organization head does not replace the explicit metrics permission. A manager's job title alone does not establish either permission.
Access is evaluated using the current person, organization and permission context. Expired or changed access can prevent a result from being returned even if the request began while access was valid. A previously displayed option is not a permanent grant.
Enabled by default still has conditions¶
The current non-private feature families are Work Assistant, Organization Intelligence and Governed Operations. Their feature defaults are enabled, with explicit organization overrides and dependencies between related features. Turning off a parent capability can also disable its child features. A feature flag is separate from the subscription and authorization decisions above.
The platform does not require a release-certification document to activate these non-private defaults. It does require the runtime information needed to authorize the request. Missing workspace policy, inactive governance information, an unapproved provider, an unsupported purpose or incomplete jurisdiction information can make a request unavailable.
The checked-in configuration enables the non-private conversation and Azure search paths. That configuration does not prove that an environment has the required credentials, model deployment, indexes or live connectivity. Private provider profiles, private embeddings and the private pgvector path remain disabled; their presence in the code does not make Private Cloud available.
Discovering supported interactions¶
The backend provides a current catalog of registered interactions. It filters that catalog by Basic or Advanced access and by the permissions required by each interaction's owning service. It also asks the source-specific authority whether that interaction is available. Knowledge navigation, for example, checks current organization membership before offering a library or manual destination.
The catalog distinguishes reading information, preparing a handoff and navigating to a destination. Some operations require an explicit preview or a secure handoff. Listing an operation does not perform it, approve it or confirm that its frontend screen exists. Platform-staff and account-authentication journeys are excluded from this tenant interaction catalog.
The current readiness summary describes four areas: basic conversations, the organization-intelligence dashboard, action history and reusable-workflow reading. It is a concise availability view, not a complete list of HiveAI behavior. Individual operation checks remain authoritative.
Understanding readiness messages¶
| Meaning | What a reader should conclude |
|---|---|
| Access cannot be evaluated or is denied | Refresh expired access or ask the workspace administrator to investigate the stated restriction |
| Feature disabled | The person may have general HiveAI access, but this feature is currently disabled |
| Provider configured, not probed | Configuration and approval checks passed; connectivity, source access and credits are checked when a conversation is sent |
| Provider not approved or configuration unavailable | An administrator or operator must resolve the provider configuration or approval issue |
| Model profile suspended | An operational restriction currently prevents use of that model profile |
| Provider not required for this read | The readiness check for this read does not need a model connection; the individual read still rechecks access |
An access problem should not automatically be treated as a need to buy more credits. Entitlement, permission, governance, provider configuration and spending capacity are different questions. This feature's discovery and navigation-preparation paths do not themselves call a language model or reserve conversation credits. Paid execution and settlement belong to the relevant conversation or action workflow.
Preparing a knowledge destination¶
The backend can prepare two knowledge destinations: the organization's published knowledge library and a particular currently accessible manual. Preparation is tied to the requesting person, organization and conversation. It verifies source access and returns a short-lived destination for a compatible client to open.
A prepared destination expires after at most 15 minutes and can expire sooner with the person's permission context. Before returning or acknowledging it, the service checks whether the conversation context, permissions, registered interaction or source has changed. A withdrawn or changed manual can make the saved destination unavailable. The library's contents can also change; preparing its location does not freeze its search results.
The service distinguishes a destination being prepared from the client reporting that it opened. A client may report opened, failed or cancelled. Even an accepted opened report is the client's report of rendering, not proof that a person read or understood the material. No connected browser implementation of this handoff has been established in the inspected frontend.
What managers and executives can rely on¶
This access model is intended to keep assistance within organizational authority. It supports separate decisions about commercial entitlement, personal permission, approved use and source access. It also permits a safety restriction to stop a capability without treating every user as an account problem.
For rollout planning, confirm an actual user journey in the intended environment: the right person can enter, the client can display the feature, the provider can serve an authorized request and the user can recover from denial or interruption. A feature flag, product card or backend readiness response alone is insufficient evidence for that rollout decision.
Business review should check whether people understand the difference between available assistance and completed work. Opening a destination does not submit an approval, sign a document or execute a protected operation. Those outcomes need their own workflow and evidence.
See the access and troubleshooting guide for the checks a user, administrator or support person can make.
Continue with conversations and saved-request recovery or reviewed voice instructions and spoken answers. These guides describe the intended final-screen journey with current frontend and environment limits.
Check supporting knowledge¶
The grounded-answer guide explains source disclosure, missing evidence and changed access. Reading published knowledge explains current member publications and protected document recovery. Their backend capabilities do not change the current frontend or environment availability described here.
Personal preferences¶
Personal memory explains preferences you deliberately choose to save for later HiveAI use. The task guide covers the intended review-and-save steps and the limits of correction, pause, resume, deletion and export. A saved preference does not grant permissions or schedule work.
My Day and work status¶
My Day and work assistance helps you identify workflow tasks to inspect and understand the limits of current record or correspondence status checks. The task guide follows the intended conversation screens and explains why a bounded answer, sample due time or empty result cannot replace a current WorkOS check.
Understanding organization intelligence¶
Organization intelligence explains adoption, quality indicators, credits, budgets, knowledge-source health and governance. The reader guide follows the intended conversation and distinguishes observed use from workforce adoption, safety markers from answer accuracy, and current credit balances from verified historical spending.
Forecasts and executive briefings¶
Briefings and forecasts combine aggregate historical information with a future weekly trend and optional advice. The task guide explains the intended forecast, preview and produced-briefing conversations, schedule rules and safe recovery. These designed journeys remain subject to frontend delivery and current access.
Credits and spending controls¶
Credits, budgets and usage explains prepaid workspace funding and user/team/department limits. The task guide follows the intended Tenant Admin screens and explains warnings, held work, uncertain costs, expiry and recovery.
Review and approve a business change¶
Governed actions and approvals separates a proposal, your approval, source execution and a confirmed result. The task guide follows the intended exact-review, spoken-decision and recovery screens and explains why stopping a conversation or preparing a restoration does not establish that a task has been undone.
Repeat work with an assisted workflow¶
Assisted workflows save an ordered plan and use selected inputs for each run. The task guide follows creation, version changes, retirement, starting, separate action approvals, saved progress and recovery, with current availability limits.
Use approved organizational guidance¶
Shared work memory provides reviewed guidance for relevant answers. Its task guide explains Tenant Admin inspection and decisions, use dates and audience, unavailable authoring, removal and uncertain-result recovery. Approval makes guidance eligible; it does not guarantee inclusion in an answer.
Review limited access to Secret information¶
Secret-access exceptions bind a named recipient, exact source, purpose and finite access window. The decision guide explains inspection, conditional review, revocation and uncertain outcomes. A saved grant does not bypass the remaining source and policy checks.
Review governance and raise concerns¶
The governance and incident guide explains how to interpret audit evidence, report a concern and distinguish an incident from a restriction or recovery decision. Recorded evidence and approved policies have defined limits; a successful answer alone does not establish complete evidence.
Manage shared manuals¶
The workspace knowledge guide explains how an administrator admits a manual, approves a processing maximum and deliberately publishes a ready version. Current members consume published content; preparation and preview alone do not grant that access.
Understand provider availability¶
The provider guide explains configuration and approval messages, interrupted answers, administrator responsibilities and governed model changes. A configuration check does not promise connectivity or access to every source.
Understand the disabled private capability¶
The Private Cloud overview explains the private processing boundary, current disabled status, accountable owners and remaining qualification work. The preparation and support guide distinguishes commercial legacy review from AI activation and explains safe recovery and data responsibilities.
Global Support guidance has its own publishing lifecycle. A processed document, a published source and searchable guidance are separate states.
Zerohive's own staff use a separate Platform staff assistant. Organization roles do not grant staff authority, and its saved requests and platform-internal cost accounting have different boundaries.