Use cases / Customer and revenue
AI Customer Service
Resolve customer requests with agents that know your policies and stay inside them.
01
Problem
Support teams answer the same questions across channels while context lives in tickets, wikis and CRM notes. Ungoverned chatbots promise refunds they cannot give or expose data they should not see.
02
AI capability
Understand the request, retrieve the right policy and account context, draft or send a reply, and take bounded actions in the help desk and billing systems.
03
Data
The agent works from the context you connect, not from the open internet by default.
- Help center and policy documents
- Ticket history
- CRM account records
- Order and billing status (read-only by default)
04
Agent
Customer Service Agent
Triages incoming requests, answers from approved knowledge, and escalates what it should not decide.
05
Workflow
- 1. Request arrives
- 2. Classify intent and risk
- 3. Retrieve policy and account context
- 4. Draft reply or action
- 5. Policy check
- 6. Send, or route to a human
- 7. Log outcome
06
Governance
Governance runs inside the agent at runtime: policy changes what it can actually do.
Can
- Read approved knowledge and the requesting customer account
- Draft and send replies within templates
- Create and update tickets
- Prepare refund recommendations
Cannot
- Access other customers records
- Issue refunds above its limit
- Change account ownership
- Share internal-only notes
Requires approval
- Refunds and credits above threshold
- Replies to flagged accounts
- Goodwill exceptions to policy
Records
Agent identity, Data accessed, Model used, Output, Tools called, Policy applied, Decision, Approval, Action, Outcome.
Suggested starting autonomy: L2 Approve
Start at L2 so a person approves replies and credits; move to L3 for low-risk intents once approval rates and corrections are reviewed.
07
Outcome
What you measure, so the agent earns more autonomy on evidence:
- First-contact resolution rate
- Time to first and final response
- Escalation and re-open rate
- Share of replies edited by humans
- Policy exceptions flagged
Questions
Can the agent issue refunds on its own?
Only within limits the organization sets. Refunds above the threshold require human approval, and every refund is recorded with the policy that applied.
Which knowledge does it answer from?
Only sources you connect and approve. Answers can be traced back to the documents used.
Build it on the platform
Related use cases
- AI Knowledge ManagementGive AI the context to understand the organization, without flattening access rules.
- AI SalesMore selling time, with outbound messages under control.
- Autonomous WorkflowsEmbed AI in how the organization operates, step by governed step.
- AI MarketingContent at pace, inside brand and claims rules.