Skip to main content
Dashboards and ticket work

Copilot suggestions and reply drafts

Qubit Copilot proposes actions on a ticket and drafts replies for you to approve, edit or reject, and approved actions run in HaloPSA as you.

Written By Chris Scaminaci

Last updated About 2 hours ago

Qubit Copilot is the panel on the Ticket Dashboard that proposes actions for the ticket you are working and drafts replies for you. The panel is titled Qubit Copilot - Suggested Actions. QuantumOps only proposes. Nothing is written to HaloPSA until you approve it, and then it is written as you. Technicians use it from the HaloPSA ticket tab. The browser extension has the same suggestions in its Ticket tab, described in Ticket tab.

Before you start: your administrator must have added the Ticket Dashboard to HaloPSA as a tab, and for the tab to accept actions it must carry your organisation's tenant tab key (see Adding QuantumOps tabs to HaloPSA). To approve anything you also verify your agent identity once on each browser, as described below. Your HaloPSA agent record needs an e-mail address, because that is where the verification code goes.

Know what you can do here

A badge at the top right of the panel shows how much trust your session has. Hover over it for a short explanation.

BadgeWhat it meansWhat you can do
View onlyThe tab does not carry the tenant tab key.Read suggestions and the What this will do preview. Approve is greyed out.
Tier 1 · feedbackThe tab carries the tenant tab key.Everything above, plus feedback: rate research findings, mark memories, and start a deep investigation. See Sherlock research and live context.
Tier 2 · verifiedYou verified your agent identity on this browser.Approve, edit and reject suggestions, draft and send replies, and promote research findings. Actions run in HaloPSA as you.

You may also see a Tier 3 · tools badge. It adds tool execution on top of Tier 2 when your own StackJack connection is live. It does not change anything described on this page.

Verify your agent identity

You verify once per browser. QuantumOps then recognises the browser for that organisation for up to 90 days.

  1. On any suggestion card, select Verify to act. It sits next to a greyed-out Approve. The button appears only once the panel lists a suggestion. On a ticket with no suggestions yet, wait for the analysis to finish, or verify from another ticket. Verification is per browser, not per ticket.
  2. In Verify your agent identity, select Send code. QuantumOps sends a code to the e-mail address on your HaloPSA agent record. The e-mail is titled "Your QuantumOps verification code".
  3. Enter the 6-digit code and select Verify.

The badge changes to Tier 2 · verified and the action buttons appear without reloading the tab. The code works for 10 minutes and you have 5 attempts to enter it. You can ask for a new code after 60 seconds.

If something goes wrong, use this table.

What you seeWhat to do
"Could not send a verification code." or "No agent id is available on this page to verify."Check the badge. If it says View only, the tab has no tenant tab key. Ask your administrator to use the tab address from the setup wizard.
"Incorrect code."Check the code and try again.
"Too many attempts. Request a new code."Select Send code again to get a new one.
"A code was just sent. Please wait before requesting another."Wait a minute, then try again.
"Too many verification requests. Please try again later."Wait a few minutes, then try again.
"We could not send the verification email. Please try again shortly."Try again. If it keeps failing, ask your administrator to contact TechPulse support. QuantumOps cannot send e-mail from your instance until e-mail sending is set up. E-mails QuantumOps sends lists what depends on it.
The panel says a code was sent but nothing arrives.Check your junk folder. Then ask your administrator to confirm that your HaloPSA agent record has an e-mail address, and to contact TechPulse support if it does.

Your administrator can reach TechPulse support as described in Getting help.

Review a suggestion

The panel lists suggestions as cards. Each card shows:

  • a short title and a percentage that is the model's confidence in the suggestion, in green, amber or grey from high to low;
  • Why?, which opens the reasoning behind the suggestion and becomes Hide reasoning;
  • What this will do, a preview of exactly what approving will do.

Until the ticket has been analysed the panel reads "No suggestions yet. They appear as the ticket's analysis completes." and fills in on its own when the analysis finishes. A suggestion that your PSA cannot carry out is not shown.

QuantumOps can propose these suggestions:

SuggestionWhen it appearsWhat approving does in HaloPSA
Draft a reply to the requesterThe ticket has research or a summary.Opens the reply draft editor. See Draft and send a reply.
Consider pre-emptive escalationThe ticket has a sentiment alert, a tonality alert or an escalation type.Adds a private note that records the escalation risk and asks for the account manager to be told. It does not change the ticket's priority.
Nudge this stale ticketThe ticket is stale.Adds a private nudge note.
Merge into a matched ticketDuplicate detection matched the ticket to another one.Merges this ticket into the match. A Ticket Merged note that is not marked private, so the customer can see it, is added to this ticket, and a private note naming who merged it is added to the match. This changes the structure of the tickets, so read the preview first.
Draft a KB article from this resolutionThe ticket is closed and has research. A closed ticket gets only this suggestion.Adds a private note that summarises the resolution for a knowledge-base article.
Resolution replayA closely matching closed ticket was resolved cleanly.Adds a private note with a numbered checklist of the steps that fixed the matching ticket, with the tickets it came from.

The research cards can add suggestions too: promoting a finding creates a reply draft or a note here. See Sherlock research and live context.

Approve, edit or reject a suggestion

You need the Tier 2 · verified badge for these buttons.

  1. To run a suggestion as proposed, select Approve. The card shows the result, for example "Nudge note posted.", and then its status and who acted on it.
  2. To change it first, select Edit. An Edit before sending box opens with the suggestion in a structured form rather than as plain text. The wording sits between quotation marks, such as the sentence after "Note". Change only that wording and leave the rest exactly as it is, then select Approve & send edited, or Cancel to leave it unchanged. Reply suggestions have no Approve or Edit button. You review and edit them in the reply draft editor, described in Draft and send a reply.
  3. To decline it, select Reject. In the Reject suggestion dialog choose a Reason and add Notes (optional), then select Reject. The reasons are Not relevant, Wrong recommendation, Client policy, Already handled, Insufficient info, Would cause churn and Other. The suggestion is removed from the panel.

A suggestion runs only once. If someone else already ran it, QuantumOps answers "Already executed." and writes nothing more. The escalation, nudge, KB-draft and replay notes are private in HaloPSA and are written as you.

Your choices are recorded as feedback: approved, edited with how much you changed it, or rejected with your reason. They feed Copilot Analytics and the feedback review in Override and feedback review.

The session of a tab lasts up to 15 minutes. If the tab has been open that long, approving may answer "A valid embed token is required to execute an action." and rejecting may answer "A valid embed token is required." Reload the tab and try again. The browser stays verified, so you do not need a new code.

Draft and send a reply

A reply suggestion shows Draft & send reply and Reject instead of Approve. Selecting Draft & send reply opens Grounded reply draft under the card.

  1. Wait for Generating grounded draft.... QuantumOps writes a draft from the ticket, its recent conversation and similar past tickets and memories, shaped by your organisation's communication policy.
  2. Choose where the reply goes with Customer email. Left unticked, the reply is a private note on the ticket. Ticked, the reply is e-mailed to the customer with the subject Re: Ticket # followed by the ticket number, and the draft ends with a list of sources. Changing the box writes a new draft and replaces your edits, so decide before you edit.
  3. Read the draft. The sources it relied on are listed under Citations as [1], [2] and so on, with a link marked "open" for those that have a web address.
  4. To change the draft, edit the text directly. To have QuantumOps change it, type an instruction such as "shorter, add an ETA" in the box and select Refine. Select Regenerate for a fresh draft.
  5. Select Approve & send. The reply is posted as you, and the card says "Reply sent." Select Cancel to close the editor without sending.

When you send, QuantumOps compares the text you sent with the draft it wrote, and the size of your edit is recorded for Copilot Analytics.

See what the communication policy checks

Every draft is checked against your organisation's communication policy. The check lists its findings under the draft.

  • A banned phrase in the draft, or a required disclaimer that is missing, blocks the send. Approve & send is greyed out, the editor says Send blocked by policy and the list under the draft names the phrase or disclaimer. QuantumOps refreshes this list each time it writes or rewrites the draft, not as you type. To clear a block, select Refine with an instruction such as "remove the banned phrase" or "add the required disclaimer", or select Regenerate.
  • A missing signature, or a tone that does not match the policy, is only a warning. You can still send.

QuantumOps checks the text again when you select Approve & send. If it breaks the policy, the editor says "Resolve the blocking policy issues before sending." If the policy changed while the editor was open, the send can instead stop with a message that begins "Blocked by communication policy".

Select the sliders icon, whose tooltip is Tenant tone / guardrails, to read the policy in Tenant communication policy: Tone, Style guide, Banned phrases (one per line), Required disclaimers (one per line) and Signature ({AgentName}, {ClientName} tokens allowed). From here the policy is view only. Only an Administrator can change it. See Communication policy for AI replies.