ROBO is a conversational assistant built into the main workspace — available to agents, analysts, managers, and admins while they work, not something employees see in the self-service portal. It shows up as a small floating button on every page (drag it wherever's convenient; it remembers where you left it), and clicking it opens a chat drawer.
Opening it on a ticket
Open ROBO while viewing an incident, problem, change, CI, asset, or service request, and it automatically scopes the conversation to that record — and immediately posts an opening assessment without you having to ask anything, summarizing what it's found and suggesting a first step. Open it anywhere else and it starts as a general conversation instead, offering a handful of clickable prompt suggestions tailored to your role (a technician sees things like "Draft an RCA for a database connection pool exhaustion"; a manager sees "Draft a major incident communication").
What it can answer
ROBO grounds its answers by actually looking things up rather than guessing — knowledge base articles, runbooks and their steps, known errors, similar past incidents and problems, the CMDB dependency graph, and prior conversations. While it's working, it shows what it's doing ("Searching the knowledge base…," "Tracing CMDB dependencies…"), and its answers carry citations back to whatever it used, so you can check the source instead of taking it on faith. It can also answer "how do I do X in PulseServe" questions directly, walking you through the steps inline with a "Take me there" button that jumps to and highlights the right part of the screen.
What it can do — with your approval
Beyond answering questions, ROBO can propose a small set of concrete actions on the ticket you're discussing: link a knowledge-base article, promote an RCA into a draft runbook, update priority, add a comment, or reassign to a group. Every proposal shows up as an Approve / Dismiss card — nothing happens until you approve it, and approving still respects your normal permissions, so it can't do anything through chat that you couldn't already do yourself through the UI.
If a response misses the mark, a thumbs-down prompts ROBO to take another pass at it right away rather than leaving you to rephrase the question yourself.
Switching models
A model picker lets you choose between hosted models or a locally-run one if your tenant has that configured — available to any user, not just admins, alongside a running token/context-usage indicator and a history panel of your past sessions.
Example
An L2 analyst opens a slow-checkout incident and clicks ROBO. It immediately posts: "This CI has had 2 similar incidents in the last 30 days, both traced to database connection pool exhaustion — the most recent runbook fix was a pool size increase." The analyst asks "did that runbook actually work last time?" and gets a cited answer pulling from the closed incident's resolution notes. Confident it's the same issue, they ask ROBO to link the relevant KB article to the current ticket — a card appears, they click Approve, and it's attached without leaving the chat.