AI Consoles

AI Consoles

1. What it is

AI Consoles is the per-solution control surface where admins monitor autonomous AI activity, approve or reject pending actions, and set how independently each AI-First Solution can act. Every registered solution domain gets its own console showing a hero metric, what the AI handled, a success rate, and an approval queue.

• Enablement: AI Consoles is a platform feature, not a licensed app. It is available to any admin whose tenant has at least one AI-First Solution registered. No separate licence or toggle is required.
• What it is not: AI Consoles is not the same as AI Operations (Admin → AI Operations). AI Operations contains the global kill switches (pause/resume all), agent health monitoring, and trust-level configuration. AI Consoles is the per-solution operational surface where you see what a specific solution did and approve its queued work.

2. Standing it up

There is no separate setup for AI Consoles — the consoles appear automatically as soon as a solution domain is registered.

1. Enable the app whose AI-First Solution you want to manage (e.g. Shifts Scheduling, Service Desk, Inspections) from Admin → Apps Marketplace.
2. Navigate to the console via either path:
   • Admin → AI Operations → AI Consoles tab — the portfolio index showing all 20 registered solutions.
   • Inside any app → AI Console sidebar tab — the in-app console for that specific solution, with the full app sidebar intact.
3. Set the AI mode using the autonomy dial (see §3 below). Until you move the dial off “Off”, the AI will not act autonomously for that solution.

The AI Console sidebar tab appears automatically in every app that owns a registered solution. It is visible only to users who pass the console’s permission gate (platform admins, plus the app’s own admin predicate where one is registered). Individual apps can hide the sidebar link via a Show AI Console in the sidebar toggle in their settings.

3. How it fits together

Autonomy dial — The three-position control that governs how independently a solution’s AI can act. Each solution has its own dial; the setting is stored with that app’s own configuration, and it also sets the agent’s trust level.

Mode	Label in the UI	What happens
Off	Off — manual only	The agent will not act unattended. No autonomous actions are created.
Approval	Approve — AI acts, human confirms	The agent proposes actions but every one enters the “Waiting on you” queue for a human to approve or reject before it executes.
Auto	Auto — AI acts when confident	The agent acts on its own when its confidence meets the threshold. Only critical actions still require human approval.

Changing the dial from Off to Approval sets the agent’s trust level to Probation; changing it to Auto sets trust to Trusted. The underlying trust ladder (Probation → Standard → Trusted) can still self-graduate under Auto mode via earned trust.

Hero metric — Each solution provides a single headline number specific to its domain (e.g. shifts auto-filled, tickets resolved, onboarding plans progressed). Each solution supplies and computes its own metric. A 30-day sparkline trend appears once at least two daily snapshots have accumulated.

Approval queue — The “Waiting on you” card, listing every AI action in this domain that is still waiting for a decision. Approving executes the action; rejecting requires a reason. The queue is visible only to users who hold the approver right — platform admins plus the domain’s registered approver predicate (e.g. Service Desk front-line agents can approve actions on tickets they already own).

Activity feed — A unified “What the AI handled” list combining autonomous actions with trigger-fired playbook runs for the domain, capped at the 15 most recent. Completed items show a “Done by AI” badge; failed items show “AI couldn’t complete”.

Two hosting modes — Every console can render in two places. The app-native console at /apps/<slug>/ai-console keeps the app’s full sidebar; the generic console at /ai-console/<domain> renders outside any app context. All 20 registered domains have app-native consoles. The generic URL redirects to the app-native one (302) so bookmarks and links land on the in-app surface.

Default landing redirect — When an admin sets a solution to Auto mode, the app’s primary list view (e.g. the shift list, ticket queue) automatically redirects that admin to the console instead. Non-admin users always see the classic list view. Appending ?view=list to the URL bypasses the redirect.

4. Running it

Reviewing AI activity

1. Open the console for the solution (sidebar tab or AI Operations → AI Consoles).
2. The header shows how many actions the AI handled today and the 7-day success rate.
3. Expand or collapse the “What the AI handled” feed to review recent completed and failed actions.

Approving or rejecting pending actions

1. Open the solution’s console. The “Waiting on you” card lists pending actions.
2. Click Approve to execute the action, optionally adding notes.
3. Click Reject — a reason is required. The action is marked rejected and will not execute.
4. If the queue has more than 15 items, expand the “View all” disclosure to see the rest.

Pending-approval digests are emailed to admins with a direct link to the console.

Changing the AI mode

1. On the app-native console (or the app’s automation/settings page for Scheduling, Service Desk, and Onboarding), locate the AI mode dropdown.
2. Select Off, Approve, or Auto.
3. Click Save. The change takes effect immediately.

On consoles where the dial lives on a separate settings page, the console shows a read-only badge with a “Manage in Settings” link.

Using the portfolio index

Navigate to Admin → AI Operations → AI Consoles to see every registered solution in a card grid. Each card shows the solution name, hero metric, and a pending-approval count badge. Click any card to open that solution’s console.

5. Settings

AI Consoles itself has no dedicated settings page. The settings that affect console behaviour live on each solution’s app configuration:

Setting	Where	Default	What it changes
AI mode (autonomy dial)	Each solution’s console or settings page	Off	Controls whether the AI acts autonomously and whether actions need approval
Confidence threshold	Autonomy dial (submitted with mode change)	95	The minimum confidence score (0–100) an AI action must reach to execute without approval in Auto mode
Show AI Console in the sidebar	Per-app settings (where the app declares it)	On	Hides or shows the AI Console tab in the app’s sidebar for eligible admins

6. More help

• AI Operations — Pausing, Approvals & Agent Health — the global kill switches, trust levels, and agent health monitoring
• AI Operations FAQ — specific questions about pausing, trust graduation, and error thresholds
• Background Tasks & Job Monitoring — progress tracking for imports, exports, and bulk operations
• Ask AI — the assistant answers questions about AI Consoles from these articles.