Intel: alerts, capacity and AI insights

The Intel drawer is the right-hand side of the dispatch board. It collects what needs a decision beyond the tickets themselves: alerts, recommendations, spare capacity, a leaderboard, AI insights, Q-Notices and a live feed of what is happening. Dispatchers, service desk managers and customer success staff use it to see the whole picture without leaving the board.

Before you start: you need a role that can open the dispatch board: Administrator, Service Desk Manager, Dispatcher or Customer Success (see Opening the dispatch board). Reading every panel works for all of them. Investigating or dismissing an AI insight, and posting or dismissing a Q-Notice, also need a sign-in that is linked to a HaloPSA agent. Without one, the board tells you to sign in as a linked agent.

Open the Intel drawer

1. Select Intel in the toolbar. The button carries small counts for alerts, recommendations waiting for you and active notices.
2. Read the panels from the top. In their usual order they are Alerts, Recommendations, Capacity, next 4 h, Leaderboard, AI insights, Q-Notices and Activity, unless you have reordered them. A panel whose information is not available yet does not appear.
3. To keep the drawer beside the board, select Pin Intel. Select Unpin Intel to let it overlay the board again, or Close Intel to close it. In a window narrower than 1024 pixels the drawer is a full-width sheet and has no pin.

The drawer holds one card at a time. Opening a ticket or an agent replaces Intel with that card. When you close the card, an unpinned drawer closes and a pinned drawer shows Intel again. The setting Keep the Intel drawer open in Settings → Layout does the same as the pin.

Read and act on alerts

The Alerts panel is the board's own list. The alerts for sentiment, tone and stale tickets are worked in the Customer Success Center instead (see Working the alert feed). The panel lists what needs attention now, most urgent first: alerts about a ticket that is pinging you come first, in ping order (see Needs you: pings and the focus card), then the rest by severity. It shows at most 12 alerts, and its count shows how many are active. Each row shows the alert type, a message and Act. The alerts are re-evaluated about every 30 seconds. The list updates when the set of alerts changes, including when the last one clears and the panel reads "No active alerts."

Three alert types are on by default:

Three more alert types are available when enabled for your organisation: agent burnout, a capacity forecast, and a routing anomaly. They appear in the same panel when they are on.

To act on an alert, select its row or Act:

• An alert that names a ticket on the board opens that ticket in the cockpit.
• A client cluster alert whose ticket is not on the board filters the board to that client instead.
• An agent alert focuses that agent.

An alert with no matching ticket, client or agent on the board shows a message saying it has no board action.

Check capacity for the next four hours

Capacity, next 4 h compares the demand with the people who can take it. The count in its header is the number of unassigned tickets. Below it you see three figures:

• unassigned: tickets nobody owns.
• breach in 4 h: tickets that will breach their SLA within four hours.
• free slots: how many more tickets the available agents can take.

A line underneath splits the unassigned tickets by priority and shows how many agents are available and how many tickets are open against total capacity. Each team has a utilisation bar. The count is marked critical when anything will breach within four hours, or when tickets are waiting and there are no free slots. It is marked as a warning when the unassigned tickets outnumber half of the free slots, or when open tickets reach 80 percent of total capacity.

Read the leaderboard

Leaderboard ranks up to eight technicians by tickets closed, then by the faster average time from creation to resolution. It always covers the last 7 days. The period buttons in the footer (1h, 4h, 8h, 24h, 7d) change the opened and closed counts there, not the leaderboard.

Work with AI insights

AI insights lists recent observations about the board, up to 12 at a time, each with its source and age. They come from three tiers: statistics, machine learning and an AI synthesis. Statistics and machine learning refresh in the background. The AI synthesis reads the state of the board. It runs about every 30 minutes and skips a run when nothing on the board has changed. Where faster refresh is enabled for your organisation, it runs about every 10 minutes while someone has the board open.

The AI tier's insights stay on the list for about 30 minutes unless the tier raises them again. If AI processing is paused for your organisation, the AI tier does not run (see Pausing and resuming AI processing). An insight that names a ticket can also raise a recommendation (see Recommendations and plans).

Investigate an insight

An investigation asks the AI to look into one insight. It only reads and advises; it does not change any ticket.

1. Select Investigate on the insight.
2. Read the confirmation. It says that the investigation reads related tickets and "semantic memory" (support memory, see Support memory), then asks the model for findings and recommendations, that this spends model tokens and that it takes a few seconds. Select Investigate to continue.
3. The AI investigation dialog shows the insight and a progress message while it works. When it finishes, it shows a summary, Findings (each with its source when it has one) and Recommendations (an action and a description). If there is nothing to report, the dialog says the model returned nothing actionable.
4. Select Close.

Dismiss an insight

1. Select Dismiss on the insight. Its tooltip reads Dismiss for everyone.
2. Confirm. The insight is hidden for every dispatcher until the intelligence tiers raise it again.

Post and dismiss Q-Notices

Q-Notices lists the active notices for your organisation, each with its severity, type, scope, who posted it, its age and when it expires. For what a Q-Notice is and who else sees it, see Q-Notices.

To post a notice:

1. Select the plus button in the panel header (New Q-Notice). The dialog is titled "Post a Q-Notice for the whole organisation": every technician in your organisation sees the notice.
2. Fill in Message, Type, Severity and Expires. Tick Include in AI processing for this scope or Broadcast to Slack / Teams as well if you want them.
3. Select Post. The notice is attributed to you in the audit log.

This notice has the Organization target, so it shows on the indicator of every record and as a ribbon on the board. With Include in AI processing for this scope ticked, it also reaches ticket analysis: see How notices reach AI analysis.

To post a notice to one agent instead, right-click the agent and choose the option that starts "Post a Q-Notice to".

To remove a notice, select the cross on it and confirm. It expires for everyone straight away with the reason "Dismissed", and the audit log attributes the dismissal to you.

Follow the activity feed

Activity shows live events as they happen, newest first, with ticket numbers in bold. Select a row to open its ticket when that ticket is on the board.

Choose which panels the drawer holds

Select Settings → Layout. Under Intel drawer you can move each panel up or down and hide it. A hidden panel waits under Hidden panels, where one click brings it back. See Board configuration.

Related pages

• The dispatch board at a glance
• Recommendations and plans
• Pulse, live calls and notice ribbons
• Q-Notices