Konnie Control Center
The Control Center is the organization settings and evidence area for Konnie. Use it to understand what Konnie can do, which services it uses, who can approve actions, and what happened. Settings affect other users; check the active organization before making a change.
Access and navigation
Open Konnie Control Center from the administration/settings navigation available to your account. Access requires the appropriate Konnie control permission. Legacy role-based deployments and deployments using scoped capabilities can expose different access rules; a title such as administrator is not a substitute for current permission. Reading settings, managing them, approving actions, and managing company memory/learning can require different rights.
The navigation groups Home, Capabilities, and Evidence organize the pages. Use settings search or select a section directly. A section hidden by a deployment feature flag is unavailable in that deployment.
| Section | What to inspect or configure |
|---|---|
| Overview | Posture, readiness warnings, pending approvals, and recent activity. |
| Autonomy & access | Explanation of personal autonomy, Access rights administration and external approval settings. |
| Providers & storage | Reasoning provider/model, video analysis, attachment analysis, storage, and Live prerequisites. |
| Sandbox | Region and which users may use sandbox analysis. |
| Background work | Scheduled tasks, scoped learning/proactive workers, saved mandates and durable action reviews. |
| Integrations | Connected mailboxes and related integration/access settings. |
| Pronunciation | Company terms and pronunciation previews. |
| Learning | Organization learning, sources, personal participation overview, and skill improvements. |
| Memory | Inspect and manage learned knowledge within your authority. |
| Audit log | Action decisions, approvals, details, and CSV export. |
| Usage & limits | Credits, request/token reporting, breakdowns, warnings, and limits. |
| Retention | Audit retention classes, raw-payload policy, and legal hold. |
Change settings deliberately
Record the current values and intended effect before changing a shared policy or provider. Provider/model setup uses Save. Other dropdowns and switches generally save immediately; individual text/limit editors use their displayed save control. Approval-policy changes have a confirmation step. Watch the save confirmation or error, and refresh the section to confirm the persisted result.
A failed save can leave the previous configuration in place. Do not infer success from a selected value alone. Restore the previous saved values if you need to revert a configuration change. Reverting settings does not roll back actions, files, or records created while those settings were active.
Providers, models, and storage
The organization has a set of enabled providers. It does not have one selected AI provider. The model chosen for a message determines which enabled provider handles it.
- Open Providers & storage → AI providers and models.
- Enable the providers your organization is approved to use, then choose the available models in each provider's card.
- Review the processing region. Under Connection settings, authorized administrators can supply or rotate credentials. Blank fields keep existing credentials; stored secrets are never displayed.
- Choose an organization default model and reasoning effort across the enabled models, then select Save. Provider enablement and model settings save together. Disabling every provider prevents new AI work.
- Under Administration, optionally restrict models and set defaults for groups or individual users. Organization limits always apply. User overrides take precedence; otherwise configured groups combine by intersection. Conflicting defaults resolve predictably, and groups with no common model block use.
Users choose their model and reasoning effort with two separate controls in the composer. Their choice is remembered while allowed; unavailable choices fall back to a permitted default. Only valid reasoning levels appear for each model. A provider configuration check is not proof of a successful request, regional account entitlement or completed business workflow.
Video analysis has its own profile and limits. Live speech has separate prerequisites and can use a dedicated approved OpenAI speech connection while reasoning continues with another provider. Do not change the reasoning provider just to troubleshoot Live. See voice and Live.
Storage can be Ennea-managed S3, customer S3, or S3-compatible storage. Customer storage requires the configured bucket and region; compatible storage additionally exposes endpoint, prefix, and path-style settings. Coordinate storage changes with the deployment administrator: selecting a new profile does not migrate existing files.
Autonomy, access and background work
Internal Konnie actions use the acting identity's current platform permissions and native business rules. Under Access rights, administrators choose a person's or group's allowed autonomy levels and whether unattended work is permitted in the selected workspace. Users select Ask before changes, Save drafts or Full autonomy in chat. The former internal action bundles and global action matrix are replaced by this model.
Use Background work to create and inspect personal scheduled tasks and shared learning/proactive workers. Each assignment displays its purpose, exact site/environment, selected level, status and authorization revision. Personal tasks retain their owner's current rights; shared workers use a dedicated scoped execution identity. Activation records explicit authorization. Changing a chat level does not change an existing background assignment. Inspect saved task settings before activating or resuming work.
Background planning also keeps the level and permissions authorized when the run starts. Shared schedule drafts can be saved under Save drafts or Full autonomy; Ask before changes creates an exact review. Changing the chat later does not expand an existing run. Production commitment still follows the shared scheduler’s human review.
Pending background reviews show the exact proposed action. Approve or reject under current authority; a changed mandate invalidates old reviews. Pause and revoke controls stop subsequent effects. An unknown dispatched outcome remains unconfirmed until reconciled and cannot be blindly retried. Existing background tasks without a saved mandate require review and activation before they run.
After revoking a shared worker, an authorized administrator can explicitly create a replacement with fresh ownership, current permissions and a new mandate. Old identity handles and pending reviews remain invalid. Active assignments cannot be silently taken over.
Connections retain their own external-communication permissions and eligible approvers. Providers, usage limits, learning-source participation, retention and native business approvals remain separate controls. See chat and governance.
Sandbox, attachment search, and learning
Sandbox settings select the available region and user audience. Disabled sandbox access can prevent spreadsheet or archive analysis even when ordinary chat works.
Attachment settings control automatic analysis, its model, maximum automatic file size, and monthly credit budget. To index an existing collection, request an estimate, inspect eligible file counts and estimated credits, and explicitly confirm a bounded backfill. Monitor completed/failed counts, actual credits, and cancellation status. An estimate is not an invoice or proof all files will process successfully.
Learning settings control organization learning, permitted sources, and whether skill improvements are disabled, reviewed, or eligible for trusted automatic execution. Users can separately set conversation-learning participation in personal Konnie settings. Review company memories and pending skill versions regularly. See skills and memory.
Active Mode, integrations, and pronunciation
Proactive work reviews selected platform events under an explicitly activated scoped worker. Its cadence and event settings determine what is observed; the worker's saved mandate and autonomy level determine what may happen. Existing Active Mode feature availability still applies. Pausing a proactive worker does not disable spoken Live conversations.
Under Integrations, inspect connected mailboxes and their configured access, then follow the connection-specific setup and review controls. Authorize only the intended mailbox and users. Provider readiness does not prove that a mailbox or external action is connected. See email integrations.
Under Pronunciation, add a term, language, and pronunciation, play the preview, and save the dictionary. Avoid duplicate entries for the same term/language. If another editor changed the dictionary, refresh and reapply your intended changes to the current version. Preview can use the configured speech service and is subject to its access and usage rules.
Usage and retention
Usage shows the reporting period, total credits/events, AI requests/tokens when available, and breakdowns by provider, entry point, status, and action. The view identifies whether evidence comes from metered AI events or an audit estimate. Missing provider usage is not proof of zero external cost, and Ennea credits are not a provider invoice.
Set organization/user limits, specific Active Mode and sandbox-analysis budgets, warning thresholds, and the offered limit behavior. A hard stop governs further work at its enforcement points; in-flight requests, delayed final usage, and external provider accounting mean it should not be treated as an exact invoice cap.
Retention settings govern standard, critical, debug, and raw-payload audit evidence, plus legal hold. These are not a universal deletion timer for every operational artifact, browser draft, file, backup, or provider account. Use your approved organization retention policy and security/deployment documentation.
Troubleshooting
If access is denied, check current control permissions and selected scope. If settings do not save, read the validation message and check whether additional authentication is required. For provider failures, inspect the selected model, region, credentials, feature availability, and readiness messages. For missing approvals, verify eligible groups and inspect whether the item can resume a live action or only records a decision.
A green Live prerequisite check confirms current login and connection configuration; microphone, conversation access, budgets, and a real audio connection are still checked when the call starts.
AI-assisted report translation
Translation uses the provider, model and region selected by the governed Konnie provider configuration. It does not silently select a separate translation provider. Standalone Form authoring/publication and new Form capture are retired; new App labels and input text are maintained in App source/settings in English and German.
Authorized historical Form reports can retain prepared language copies. Where report controls offer generation of a missing language copy, the request can send translatable text-field answers and attachment titles/descriptions to the configured provider. This can include personal or confidential free text. Attachment/image bytes, signatures, non-text answer values and actor identities are excluded from that report-translation payload. Review what the report contains before requesting a copy.
Downloading an existing prepared PDF does not itself request new translation. Generated language copies retain provider/AI provenance and derive from the immutable submitted revision; they do not replace canonical answers or extend a signature to translated free text. Review important wording and correct copies through the offered controls. They are not certified translations.
Provider contract, region, transfer safeguards, retention and permitted use remain governed by the active profile and deployment policy. A visible legacy source or saved language copy does not restore retired Form editing or submission. For current reusable document translation see Reports.