The AI Control Center

The AI Control Center is where an administrator manages every AI Assist capability from one place. It’s organized as a tab bar with nine areas.

Overview

A snapshot of the whole console: a plugin-status chart, a skills-status breakdown (active / inactive / draft / not started), a “Needs attention” count linking to what’s incomplete, a setup journey checklist, and a skill-usage chart of actions over time split by channel (the AI Assist panel, in-product actions, and the Virtual Assistant). It also shows the current AI routing status — whether native model providers or an external gateway currently own outbound AI traffic.

Skills

Browse the AI skill catalog by category (Employee, Technology, Customer, and more), search or filter by feature and status, and Activate/Deactivate individual skills. A details drawer shows a skill’s category, activation state, which model service it runs on, and when it was last updated, with a link to its performance detail.

Experiences

Manages where AI Assist itself shows up — the AI Assist panel and the AI Assist context menu — each with a license status, a summary of what it is and where it appears, a turn-on/turn-off control, and any per-experience settings (for example, voice input for the panel).

Model providers

Where an admin chooses which model actually answers AI Assist requests:

• Pick a primary provider and model, and optionally a fallback provider/model used if the primary is unavailable. Available providers are the platform-native LLM, OpenAI, Azure OpenAI, and Anthropic Claude (via AWS Bedrock), each with its own model list.
• Credentials are encrypted. For OpenAI and Anthropic, the admin pastes the API key directly into this tab; it’s stored in an encrypted vault and decrypted only server-side — the key is never shown back in the browser, never logged, and never handled by an AI agent. A Test connection action confirms the key actually works, and an active runtime banner reports which provider is really answering requests right now, so a stored preference can never silently drift from reality.
• At the top of this tab is the AI Gateway card — the master switch for routing all AI traffic through an organization’s own gateway instead. See Governance for what changes when it’s on.

Guardrails

The same enforced guardrail policy used at runtime, edited from one place (the Guardrails tab and the separate AI Assist Governance page both write to it — there’s no second, divergent copy). Below the enforced policy sits a clearly separated “Advisory settings (display only)” panel for presentation preferences — like a “require human review” posture or a suggested response-length cap — that are not enforced by the guardrail engine itself. See Governance for the categories and their block/flag behavior.

Conversations

A conversational-analytics view over real chat data: activity KPIs (active users, conversation count, resolved-chat rate) plus a master/detail browser over recorded sessions — a table of conversations (created time, duration, user, channel, language, end state) with export and a transcript detail panel for any selected session.

Performance

A left-rail set of analytics dashboards: Usage and Adoption, Self-Service Performance, Skill Performance (with a per-skill drill-down), Custom Skill Details, AI Assist Guardian (blocked/flagged interactions), User Search Analyser, and AI Assist Context Menu usage.

Budget

Month-to-date AI spend against a configured monthly limit, one or more alert-threshold percentages, and a toggle for whether crossing a threshold notifies admins. A limit of 0 tracks spend without capping it; once a hard limit is set and month-to-date spend meets or exceeds it, further AI Assist requests stop calling the live model and return a fixed message asking an admin to raise the limit, instead of continuing to spend.

Settings

A left-rail group covering Plugins (browse and install available AI Assist plugins), Multilingual service, Data sharing and processing, AI Assist Guardian settings, Manage Integration, and Account.

Next

• AI skills — the catalog the Skills tab manages.
• Governance — guardrails, budgets, usage, and the external gateway in depth.