Your queue, triaged
Inbound tickets arrive classified by intent and urgency, with a grounded draft answer and its source citations attached — agents clear the routine, humans take the judgment calls.
AI for Service teams
Answers your customers can trust — grounded in policy, not improvised.
The Service pack resolves support requests from your approved knowledge base, routes what it can't, and escalates with full context — while refunds, credits, and account changes gate through the approvals your team already controls.
Connects to Zendesk · Salesforce · Intercom — plus anything with an API
Agent Studio
skills
Search KBCreate ticketSend email add
Publishedv4 · reliability 96
Where is order #4471?
Shipped, arriving Thursday.tools: Search KB · $0.004
studio ▸ build → publish → chat-test
Day one
What service teams see when they log in.
Not a blank canvas — a working role with default goals, allowed tools, and the guardrails already on.
Answers with receipts
Every response cites the policy or article it came from. If the knowledge base doesn't support an answer, the agent says so and routes — it never improvises policy.
Escalations with context
What reaches a human carries the full trail: what was asked, what was tried, which policy applied. No 'please repeat your issue' — for the customer or the rep.
Skills invoked · governed by policy & licensing
knowledge.searchcomms.draft_response
Systems
ZendeskSalesforceIntercom
In the pack
The Service Agent ships ready to work.
Searches approved support knowledge and drafts a response or escalation recommendation; it does not update tickets. Installed from the library, wired to your systems, and run through the same 8-gate chain as everything on Cortex.
- Search approved support knowledge
- Draft a customer response
- Recommend when a person should review
Governed for this role
Helpful at the front desk. Harmless at the till.
Service agents read your knowledge and answer freely — but anything that moves money or changes an account is a governed action with your thresholds on it.
- Refunds and credits pend for approval above your threshold
- Account changes are typed actions with an audit trail
- DLP screens every conversation in and out
- Every resolution sealed in the ledger — QA can replay any run
audit/verify
ok: true
verifyChain ▸ chained SHA-256 · signed receipts
FAQ
What service leaders ask.
What happens when the agent doesn't know the answer?
It says so and routes. Answers are grounded in your approved knowledge with citations; low-confidence asks escalate to a human with the full conversation context instead of guessing.
Can the service agent issue refunds?
Only inside the threshold you configure. Refunds are governed actions — instant under the limit, pending on an approval board above it, recorded in the ledger either way.
How does QA review agent conversations?
Every run seals into the Trust Ledger with its sources, actions, and outcome. QA replays any conversation end-to-end and sees exactly why the agent answered the way it did.
Keep exploring
Where to go next.
Agent Library
Every Service agent and skill, filterable across the 21,500+ item catalog.
All roles
Ten more enterprise role packs — from Service to Field Operations.
Agent Studio
Extend the pack or build your own role — no code, full governance.
Pricing
Individual, Team, and Business — governance included on every plan.
Put a governed Service to work this week.
Start a free trial, install the pack, and connect your systems — or bring us one workflow and we'll demo it end to end.