Agent Performance Hub settings

Hub settings choose which agents the Agent Performance Hub covers, how profile runs and automatic AI narratives behave and where the workload thresholds sit, and show the collector's health and AI narrative spend.

Written By Chris Scaminaci

Last updated About 3 hours ago

The Settings panel of the Agent Performance Hub decides which agents the hub covers, how profile runs and automatic AI narratives behave, and where the workload thresholds sit. It also shows whether the metrics collector is healthy and what the hub's AI narratives have used. It is for whoever manages the service desk: the settings apply to your whole organisation.

Before you start:

The Agent Performance module must be on for your organisation.
Everyone who can open the Agent Performance Hub can open Settings: the Administrator, Service Desk Manager, Dispatcher and Customer Success roles. Only an Administrator can change them. For the other roles the panel is read-only: a note says Read-only. Only an administrator can change the Agent Performance Hub settings., every field is disabled and there is no Save Changes button, only Close.

Open and save settings

In the hub header, select Settings. The Agent configuration panel opens in place of the view.
Change what you need (Administrator only). Nothing is saved yet.
Select Save Changes. One save stores the agent settings, the profile generation settings and the workload thresholds together.
Select Close to go back to the figures.

Settings is a panel, not a view. Choosing a view or selecting Settings again also closes it.

Choose which agents are included

Under Agents, one card per active agent lists them by team, then by name. Each card has:

Control	What it does
The checkbox beside the name	Ticked means the agent is included. Clear it to exclude them.
Profiled or Not profiled	Whether the agent has been analysed. Hover over Profiled to see the date of the last analysis.
Role	The agent's role: Agent, Senior Agent, Team Lead, SDM (Primary), SDM (Secondary), Dispatch, Manager or Executive. This is the same Agent role that Team Management sets under Edit details, so a change in one place shows in the other. Team Management offers only Agent, SDM Primary, SDM Secondary and Dispatch. It is not an access role: see Inviting people, roles and agent links.
Assessment	Which measures the analysis emphasises: Technical, Non-Technical, Manager (Technical), Manager (Non-Technical), Dispatcher or Account Manager. Technical roles focus on problem-solving accuracy, non-technical roles on communication and process adherence. Non-Technical and Dispatcher are profiled from the actions the agent performed in HaloPSA rather than from assigned tickets.
Team	The team the hub groups the agent under. The list holds the teams already assigned to your agents, and No Team. Importing the agent again from HaloPSA, in Team Management or the initial data import, replaces this with their primary HaloPSA team when HaloPSA has one. Role and Assessment are kept.

An excluded agent does not appear in the hub, is left out of the daily figures and of profile generation, and does not appear on the Agent workload card on Home. Use it for people who do not work tickets.

Set how many tickets a profile run reads

Under Profile generation, Tickets is the number of each agent's most recent tickets that a profile run reads when you generate by ticket count. Enter a whole number from 50 to 10000 in steps of 50. The default is 500. A higher number gives the AI more to read and takes longer, and values above 2000 may time out. The daily refresh of each agent's snapshot figures also reads this many recent tickets per agent. For an agent assessed as Non-Technical or Dispatcher, a profile run counts their recent actions instead.

See Generate profiles for how a run uses the number.

Turn on the automatic AI narrative

Auto-generate narrative is off by default. The hub always collects its metrics, whether it is on or off. The setting controls only the AI-written narrative in agent profiles, and two nightly jobs.

When it is on, the hub looks at every included agent once on weekdays between 03:00 and 05:00, in your organisation's time zone, and does one of three things:

It writes a new AI narrative when the agent has no profile yet, has no AI narrative on record, or has a narrative older than 30 days. It also does this when the agent has new tickets and their 30-day figures have moved materially, which means volume, speed, rates or backlog have moved out of their bands, or the DiSC primary style has changed.
It applies a rule-based update, with no AI, when there are new tickets but the figures have stayed within their bands.
It does nothing when there are no new tickets.

Turning it on also turns on the nightly updates of the client and user profiles shown on the Client Dashboard and the User Dashboard.

Limit scheduled narratives

Monthly narrative cap is optional. It is off by default, shown as 0. Enter a number from 1 to 10000 to limit how many scheduled AI narratives run in a calendar month. When the cap is reached, an agent who would have had a new narrative gets the rule-based update instead if they have new tickets, and otherwise waits. A manual run is never blocked by the cap.

Set the workload thresholds

Under Workload thresholds, set when an agent counts as busy. Workable tickets are the open tickets that count toward an agent's load.

Setting	What it does
Overloaded	Agents with this many workable tickets or more are flagged as overloaded. The default is 15.
High	Agents with this many workable tickets or more are flagged as high workload. The default is 10.
Exclude SLA hold	When ticked, tickets on hold or SLA hold are not counted as workable, because nobody can act on them. It is ticked by default.

Both thresholds take whole numbers from 1 to 100. With the defaults, the workload level of an agent is:

Workable tickets	Level
None	Fully available
1 to 3	Available
4 to 9	Moderate load
10 to 14	High load
15 or more	Overloaded

The levels, the Overloaded count in the Executive view and the overloaded concern in Team View all use these settings. So does the Agent workload card on Home.

Check the metrics collector

The Metrics collector block shows whether the background collector is keeping the hub's figures current. The collector runs around the clock and costs no AI use. A pill beside the title reads Running, Error or Waiting for first run.

Line	What it shows
Cadence	How often the collector runs, between every 15 and 60 minutes.
Last cycle	When the last cycle ran, and how long it took.
Today refreshed	When today's figures were last refreshed.
Last finalized day	The most recent day whose figures are complete. The collector closes each day once it has ended.
Agents	How many agents the last cycle covered.
Rollup days	How many days of daily summaries are stored.
Halo reports	Whether the collector could read HaloPSA's reports. Unavailable means first response, SLA and logged-hours figures are incomplete until HaloPSA answers again. The collector fills those days in on later cycles.
Last error	The last problem the collector hit. It appears only when there is one.

Until the collector has stored its first daily summary, the hub shows Live figures. See Where the figures come from.

Review AI narrative spend and activity

The AI narrative and spend block shows what the hub's AI narratives are doing. A pill reads Auto-generate on or Auto-generate off.

Line	What it shows
Window	When scheduled narratives run: weekdays 03:00 to 05:00 in your organisation's time zone.
Last narrative pass	When the hub last looked at its agents for narratives.
Batch API	Whether your organisation's scheduled narratives run as a batch (Eligible) or one at a time (Not enabled).
Monthly cap	How many scheduled narratives have run this month, and the cap if one is set, or Off.

A Spend table shows the hub's AI use for Today, Last 7 days and Last 30 days: the number of Calls, the Input and Output sizes in tokens, and the estimated cost in USD. Narratives run as a batch are not in this table. The block counts them separately. For AI use across the whole product, see AI Usage Analytics.

Recent narrative activity lists the latest 12 times the hub acted on an agent. Each row shows When, the Agent, the Mode, the Reason and Tokens in / out. The mode is Batch when a scheduled narrative is queued, BatchApplied when its reply arrives (the tokens are on that row), Sync for a scheduled narrative written straight away, Manual for a run someone started, RuleBased for an update that used no AI, and Skipped when the monthly cap deferred an agent or the pass hit an error. The reason says why, for example that it was the agent's first profile or that the narrative was older than 30 days.

Agent Performance Hub: the views these settings feed.
Agent detail and profiles: how to generate profiles.
AI Usage Analytics: AI use across the whole product.

Was this helpful?

Still need help? Ask the team