Communication policy for AI replies

The communication policy is one set of rules for every reply that QuantumOps drafts with AI. It decides how a draft should sound, which phrases it must never contain, which wording it must always contain, and how it is signed. The same policy applies wherever a technician drafts a reply: in the Ticket Dashboard (see Copilot suggestions and reply drafts), in the browser extension (see Ticket tab) and on the dispatch board (see Needs you: pings and the focus card).

Only Administrators can change the policy. Everyone else can read it but not edit it.

Before you start:

• You need the Administrator role.
• The policy belongs to your whole organisation. There is one policy, not one per client or per technician.

What each field does

Banned phrases and required disclaimers are checked by QuantumOps itself before a reply is sent, not only by the AI that writes the draft. That is why only Administrators can change them: a technician cannot widen the rules for their own replies.

With an empty policy, drafts use a professional, clear and empathetic tone, and nothing is blocked.

Set the policy

1. Select the gear icon in the page header (its tooltip is Settings). The page is titled Tenant Configuration.
2. Scroll to Communication Policy.
3. Fill in the fields above. Put each banned phrase and each disclaimer on its own line. QuantumOps trims the spaces around each line and ignores blank lines.
4. Select Save policy. A message confirms that the policy was saved.

The policy section has its own Save policy button. The Save buttons of the other sections do not save it. The change applies to the next draft and the next send. See Tenant settings for the rest of the page.

How the checks work

The checks apply to every reply draft, including a draft written as an internal note.

What a technician sees

In the Ticket Dashboard, the reply editor lists each problem in a coloured row. A red row marks a banned phrase or a missing disclaimer, and an amber row marks a note that does not block. While a red row is present, Approve & send is off and the editor shows Send blocked by policy.

To fix a blocked draft:

1. Read the red rows. Each one names the phrase to remove or the disclaimer to add.
2. Type an instruction in the refine box, such as "remove the banned wording and add the disclaimer", and select Refine. Or select Regenerate to write a new draft that follows the policy.
3. Read the new draft. QuantumOps checks every new draft against the policy, and it checks the draft again when you select Approve & send.

The slider icon in the editor (its tooltip is Tenant tone / guardrails) opens Tenant communication policy so a technician can read the rules. Inside the Ticket Dashboard as it is embedded in HaloPSA, the policy is always read-only, even for an Administrator. The dialog says it is view only and that an Administrator edits it under Tenant Configuration.

QuantumOps also checks a reply at the moment it is sent, so the rules hold on every surface. On the dispatch board, when you select Send reply and the policy refuses the reply, the message begins "Blocked by communication policy" and names the phrase or disclaimer. Your text stays in the box. Edit it to remove the phrase or add the wording, then send it again. A reply posted from the browser extension goes through the same check.

Write rules that work

Good rules are specific and short:

• Banned phrases such as "per my last email" or "as you should know" keep a harsh phrase out of every reply.
• Keep each banned phrase long enough to be unambiguous. Because a phrase also matches inside longer words, a single common word or a short fragment can block almost every draft, including ones that are perfectly polite.
• Keep each required disclaimer short, and type it exactly as it must appear. The AI has to reproduce it word for word, so a long disclaimer with unusual punctuation fails more often.
• A disclaimer you require is required in every draft, including internal notes. Add one only when every reply really needs it.
• Start with an empty banned list and add phrases as you meet real problems.

Related pages

• Copilot suggestions and reply drafts: where reply drafts are made and approved.
• Ticket tab: drafting in the browser extension.
• Needs you: pings and the focus card: replying from the dispatch board.
• AI providers and models: the AI that writes the drafts.
• Tenant settings: where the policy lives.
• Q-Director Settings: the analysis, triage and research settings that sit beside these AI settings.