IT helpdesk intake
Screen helpdesk tickets for secrets and personal data, route them, check access requests, and triage reported phishing.
Scenario paths
Pick a predefined path to highlight it through the chain. No model is called; outcomes are illustrative.
- Steps
- Endpoints
- Verified self-service reset
- Service desk queue
- Ask for missing details
- Block sender, thank reporter
- Resource owner approval
- Analyst confirms
- Security desk
- Service desk agent
What each endpoint means
Verified self-service reset · Automated
A link to the identity-verified reset flow; no reset happens in chat.
Service desk queue · Automated
Placed in the hardware, software or network queue.
Ask for missing details · Automated
Asks for exactly the missing details in one message.
Block sender, thank reporter · Automated
Adds the sender to the spam filter and thanks the reporter.
Resource owner approval · Human review
The owner of the resource approves or denies; access is granted by the identity system.
Analyst confirms · Human review
An analyst confirms the message is genuine before the reporter is told it's safe.
Security desk · Human review
A security analyst handles exposed secrets and phishing.
Service desk agent · Human review
A person reads and routes the ticket.
Steps
Catch secrets and personal data before the ticket is stored.
- No sensitive dataRedact card dataRedact IDRestrict and redactIT request type
- Rotate and redactSecurity desk
- Privacy reviewService desk agent
Question, inputs and notes
Asks:What kind of sensitive data does this text contain?
Receives
- Inputrequest
- Inputdetector_hits
Branches
- No sensitive data, Redact card data, Redact ID, Restrict and redact: Code redacts any detected data before routing.
- Rotate and redact: Exposed secrets must be rotated.
- IT request typeChoice
Route the request.
- Access requestAccess request completeness
- Password or MFAVerified self-service reset
- HardwareSoftwareNetwork or VPNService desk queue
- Security concernPhishing report triage
- Service desk agentService desk agent
Question, inputs and notes
Asks:What kind of IT help is the employee asking for?
Receives
- Inputrequest (redacted)
- Inputrequester
Make sure the approver gets a complete request.
- Send to resource ownerResource owner approval
- Ask for resource and roleAsk for reasonAsk for durationAsk for missing details
- Service desk agentService desk agent
Question, inputs and notes
Asks:Does this access request include the resource and role, a business reason, and how long access is needed?
Receives
- Inputrequest
- From step 2Type: access request
- Phishing report triageChoice
Triage the reported email.
- Credential phishingMalicious attachmentSecurity analystSecurity desk
- Spam or marketingBlock sender, thank reporter
- Likely legitimateAnalyst confirms
Question, inputs and notes
Asks:What kind of email has the employee reported?
Receives
- Inputreported_email
- From step 2Type: security concern
About this chain
Goal and input
Every ticket ends in self-service, a queue, the resource owner's approval, a question to the requester, the security desk, or a service desk agent.
The employee's ticket, directory facts, and any forwarded email.
- request
- What the employee wrote.
- requester
- Department and device from the directory.
- detector_hits
- Pattern-based sensitive data matches.
- reported_email
- The forwarded email and gateway results, for security concerns.
How it runs
- The sensitive data step runs first so that secrets never reach queues, exports or other models.
- No step grants access or resets credentials. Those happen in your identity system after a person or a verified flow approves them.
Sources (3)
Primary documentation this chain builds on.
- SourcedTypeSafe docs · DocumentationSystem One
- SourcedTypeSafe docs · CookbookGuardrails for LLMs
- SourcedTypeSafe docs · PatternRouting by confidence