A virtual agent that answers, deflects, and hands off — the moment it should.
Phyllis builds and extends your entire conversational operation — any topic, deflection rule or messaging channel — grounded in your live instance and staged for your approval.
Built by ServiceNow architects — CMA-grade by default
Grounded in your instance
Approval-gated delivery
CMA-grade documentation
Self-heal on your approval
A quick scroll through what Phyllis can do
Scan chatbot setup
Build a topic
Branching flow
Deflection rate
Deflection tracking
Live-agent handoff
Search-backed answers
Guided intake
Wire web channel
Teams & Slack
Trigger phrases
Conversation reporting
Scan chatbot setup
Build a topic
Branching flow
Deflection rate
Deflection tracking
Live-agent handoff
Search-backed answers
Guided intake
Wire web channel
Teams & Slack
Trigger phrases
Conversation reporting

Questions?
We’re glad you asked.
The configuration that makes a conversation work — the topics and guided flows, the knowledge they answer from, live-agent handoff, and the wiring that puts the assistant on your web portal, Microsoft Teams, or Slack. The live conversations and the tickets they raise stay in your team's hands; Phyllis builds the flows and channels, not the chats.
She wires the ServiceNow side so the assistant can send and receive on those channels, and if a required integration isn't installed yet she tells you rather than pretending to hand-build it. Provider credentials are always set in the console, never by Phyllis.
Phyllis sets up deflection tracking so your analytics actually capture when the assistant resolves a request on its own, and routes handoff to a live-agent queue on the signals you choose. The result is a customer who asks for a person reaching one instead of looping.
Correct. Phyllis discovers, proposes a plan, and only builds on your explicit approval, with every change staged for review and reversible. Making a topic live is a deliberate publish step she flags separately — an authored flow isn't live until you publish it.
Every claim is cited to where it came from in your live instance — the topics that exist, the deflection numbers your analytics report, the channels actually wired. She reads what's really configured rather than assuming out-of-box defaults.































