Recommendations and plans

Recommendations are suggested actions raised by rules and AI insights on the dispatch board, each with a plan of steps that you review in plain words before anything runs.

Written By Chris Scaminaci

Last updated About 1 hour ago

Recommendations turn what the board sees into something you can do. A set of rules, and the AI insights, watch the board. When one of them finds a situation worth acting on, it raises a recommendation with a plan: a short list of steps such as moving a ticket to another technician or posting a notice. You read every step in plain words, then accept the plan or put it aside. Nothing runs until you accept.

Before you start: you need a role that can open the dispatch board: Administrator, Service Desk Manager, Dispatcher or Customer Success. Reading recommendations works for all of them. Accepting, ignoring and snoozing also need a sign-in that is linked to a HaloPSA agent, because the steps are written under your name. Steps that move tickets are written to HaloPSA first, so your HaloPSA connection must be working (see Your HaloPSA connection).

Where recommendations appear

In the Recommendations panel of the Intel drawer (see Intel: alerts, capacity and AI insights). The panel count shows how many need action, and the Intel button in the toolbar carries a count of the recommendations waiting for you.
In a banner across the top of the board, for the most severe organisation-wide recommendation that is critical.
In the cockpit and the agent lens, which list the recommendations about the ticket or agent you have open.

If your server does not have the recommendation engine, the panel reads "This server has no recommendation engine yet." and Rebalance from this agent is greyed out.

What raises a recommendation

The rules check the board about every two minutes. The panel says "The rules run every couple of minutes." when it has nothing to show.

Where a rule names service desk managers, it means the agents whose Agent role is SDM Primary or SDM Secondary. Where it names dispatchers, it means the agents whose Agent role is Dispatch. You set the Agent role in Team Management under Edit details (see Inviting people, roles and agent links).

Rule	It is raised when	The plan
Breach imminent	An unassigned ticket has used 85 percent or more of its HaloPSA SLA and an available technician fits it. A ticket with only an estimated SLA does not raise one unless your organisation counts estimated SLAs.	Assign the ticket to the best-fitting available technician.
Load imbalance	An online technician is over capacity (20 open tickets each) while an available technician on the same team has less than 40 percent of theirs.	Move up to three low-risk tickets from the busy technician to the free one.
Agent slowdown	A technician on shift, holding four or more open tickets, has recorded no action for two hours and normally acts more often.	Move up to two breaching tickets to other technicians, post a Q-Notice to the service desk managers (to the whole organisation when there is no service desk manager), and message the technicians who receive the work.
Appointment at risk	An appointment starts within 15 minutes and its technician is still on a call or holds a ticket that has breached.	Post Q-Notices to the technician and to the service desk managers (to the whole organisation when there is no service desk manager).
No dispatcher coverage	Every dispatcher is on a call or away.	Message the dispatchers who are on shift and free. If there are none, post a Q-Notice to the whole organisation.
Shift ending with critical work	A technician's shift ends within 30 minutes and they still hold critical or breaching tickets.	Move up to three of them to technicians who are still on shift, and post a Q-Notice to each receiver.
Recurring issue	The board detects a cluster of related tickets of the same category or ticket type within the last hour.	Post a Q-Notice to the whole organisation.
AI insight	An AI insight that names tickets recommends assigning them or escalating.	Assign the tickets, or post a Q-Notice to the whole organisation.

A card marked macro concerns the whole organisation rather than one ticket or agent. A card with no steps reads "Awareness only — nothing to run".

Review and run a plan

In the Recommendations panel, read the card. It shows the rule, its age, a title, an explanation, and chips for the ticket, client, agent and team it concerns.
Select Accept. The button shows the number of steps. You can also select the card's title to open the same details.
Read the list of steps. Each one is written in plain words, and above them the dialog says what will happen:
- how many ticket assignments are written to HaloPSA first, with the board updating only when HaloPSA accepts them;
- how many Q-Notices are posted, attributed to you;
- how many direct messages are sent on Slack or Teams (a message to someone without a linked account is skipped, not failed);
- that the plan runs step by step and stops at the first failure.
Select Accept and run. Each step is marked as done or failed, with its result.
When every step succeeds, the dialog is titled "Plan ran". Select Close. The card moves to the Recent group of the panel.

If you do not want to run it, select Ignore instead in the dialog (see below).

When a step fails

A failed step stops the plan. The steps after it do not run and are marked "not run". The dialog is titled "Plan stopped", and the card reads "Failed at step" followed by the number and the reason.

Read the reason shown with the failed step and fix its cause.
Select Retry. The plan starts again from the first step. A step that already worked is repeated only if HaloPSA still accepts it.

A message step that finds no linked Slack or Teams account is reported but does not stop the plan.

Ignore or snooze a recommendation

Select Ignore on the card, or Ignore instead in the plan dialog.
Under For how long, choose Snooze 30 minutes, Snooze 1 hour (the default), Snooze 4 hours or Don't show this again.
Optional: type a reason. It goes in the activity log.
Select Snooze or Ignore.

A snoozed recommendation moves to the Snoozed group of the panel and comes back after the time you chose if it still applies. A recommendation you ignore does not come back unless the situation changes enough to raise a new one. The activity log records that you snoozed or ignored it. Recommendations you have acted on stay visible in the Recent group for about an hour.

Run the rules now

Select the sparkle button in the panel header (Evaluate now) to check the rules straight away. The board tells you how many recommendations are new, updated or resolved, or that nothing new came up.

Show recommendations about one agent

Right-click an agent and choose Rebalance from this agent. The panel filters to the recommendations about that agent and the rules run for them at once. A chip at the top of the panel names the agent. Select its cross to show every recommendation again.

The banner names the rule, the title, a button to review the steps, Ignore, and a cross labelled Hide banner. Hide banner hides it on your board for now. The recommendation stays in the panel and for the other dispatchers.

Choose when the browser notifies you

Select Settings → Alerts. In the list Recommendations, choose All, Critical only or Off. This limits which recommendations may raise a browser notification. A notification also needs Browser notifications to be on, and it only appears while the board tab is in the background. See Board configuration.

Was this helpful?

Still need help? Ask the team