Skip to main content

Next Best Actions

A ranked list of actions the agent could take next, ordered by how confident the AI is that each one will help. It's a suggestion engine, not an automation - every action still happens because the agent does it.

Next Best Actions panel showing three action cards stacked vertically, each with an action title (e.g. "Assign to L2 Hardware Support"), a confidence badge (92%), and an expand chevron

What the agent sees​

A list of action cards. The most confident suggestion is on top. Each card has:

  • An action title - e.g. "Assign to L2 Hardware Support"
  • A confidence badge - e.g. "92% confidence"
  • An expand chevron that opens the body

When expanded, the card shows:

  • A short reasoning paragraph explaining why the AI suggested this. A copy icon next to it copies the reasoning to the clipboard.
  • A list of supporting evidence - bullets pointing at specific notes, fields, or past tickets the AI drew on.
  • A specific recommendation - what to do, in one sentence.

What it reads​

The current ticket (subject, description, notes), the tenant's ingested knowledge, and past tickets data. Past resolved tickets are the strongest signal - if 14 tickets with a similar fingerprint were all closed by routing to Hardware L2, that's where the suggestion comes from.

When it refreshes​

On ticket creation and on every ticket update (new note, field change, status change). The list is always current to the ticket's latest state.

Why "informational" and not "automated"​

The cards never click themselves. The agent reads, decides, then takes the action through the usual ticket UI. This keeps the agent in control - and means a bad suggestion doesn't become a wrong reassignment.

How it differs from Troubleshoot​

QuestionFeature
What are all the things I could do, ranked?Next Best Actions
Walk me through one fix at a time and adapt to my feedbackTroubleshoot

Next Best Actions is a menu. Troubleshoot is a conversation.