1
Observe
Twin monitors streams of new tickets, status changes, SLA alerts, and incident pages arriving from connected tools.
2
Orient
It maps each signal to entities (Customer, Ticket, Incident) and identifies urgency, source, and owner context.
3
Context Assembly
Twin loads related items: past tickets for this customer, linked Slack threads, knowledge articles, Sentry errors, and open Linear issues.
4
Reason
Twin evaluates classification, next action, and risk: is this a known issue? Does it breach SLA soon? Should it escalate?
5
Discover
If knowledge is missing, Twin queries for similar tickets, new documentation, or recent engineering changes that may relate.
6
Delegate
Specialized tasks are sent to domain agents: Incident Response for on-call, Knowledge for article suggestions, or Engineering for Jira creation.
7
Collaborate
Twin exchanges context with agents, Slack bots, and on-call systems to keep every participant aligned.
8
Decide
A recommendation is surfaced: draft reply, change status, escalate, link knowledge, or schedule follow-up.
9
Act
You approve or override. Spine executes the capability, writes to the source system, and generates a receipt.
10
Continue
Context persists for the next ticket, next session, or next responder who picks up the queue.