Network access request chatbot with auto-approval and policy enforcement
Build a Conversational Access Request & Approval Agent — a chat-based interface where employees and contractors can request network access (VLAN, VPN, guest Wi-Fi, or segmented network access) in natural language. The agent gathers required context conversationally, evaluates the request against policy, and either auto-approves low-risk requests or routes sensitive requests to a human approver. ## Interface Requirements Build a chat interface (Tines chatbot / conversational Page) as the primary entry point. It should: - Greet the requester and ask what kind of network access they need - Support natural language input (e.g., "I need VPN access to the finance network for the next two weeks" or "Can I get on the guest Wi-Fi for a vendor visiting tomorrow?") - Progressively gather any missing required fields rather than presenting a rigid form up front — this should feel like a conversation, not an interrogation - Confirm the full request back to the user in plain language before submitting ("Just to confirm: you're requesting VPN access to the Finance VLAN for yourself, starting today, for 14 days. Is that correct?") ## Required Context to Gather (via conversation, not a static form) - Requester identity (auto-pull from SSO session if available; otherwise ask and verify against Okta/Entra ID) - Requester type: employee or contractor - Access type requested: VLAN, VPN, guest Wi-Fi, or specific segment - Target network/segment name or description - Business justification (why they need it) - Device details: device type, OS, whether it's company-managed or BYOD, and MAC address/hostname if available - Role/department (pull from HRIS/Okta profile where possible instead of asking) - Location (office, remote, specific site) - Requested duration (start date/time and end date/time, or "ongoing" if standard for their role) ## Core Workflow Logic 1. **Session & Identity Handling** - Authenticate the requester via SSO where possible - If the requester is a contractor, verify they have an active contractor record and a designated employee sponsor on file (query HRIS/IAM); if no sponsor exists, prompt the agent to ask "who is your Tines employee sponsor?" and validate that person is active 2. **Policy Lookup & Risk Classification** - Query a policy reference (Tines Record, database, or config file) that maps network segments/access types to a risk tier: Low, Medium, High - Low: e.g., guest Wi-Fi, general VPN with no internal segment access - Medium: e.g., department-specific VLAN, standard VPN to internal but non-sensitive systems - High: e.g., finance/HR/engineering production segments, anything touching regulated data, or contractor requests for any internal segment - Cross-check requester's role/department against the requested segment for a mismatch flag (e.g., a marketing employee requesting access to the engineering production VLAN should be flagged even if that segment is normally "Medium" risk) - Check requested duration against policy max for the given risk tier; flag if it exceeds the max (e.g., Medium max = 30 days, High max = 7 days without additional sign-off) 3. **Decision Routing** - **Low risk + no flags** → Auto-approve immediately - Provision access directly via NAC/IPAM platform API (e.g., Cisco ISE, Aruba ClearPass, Infoblox) - Confirm to the requester in chat with access details and expiration - **Medium risk, no flags** → Auto-approve but notify the requester's manager asynchronously (visibility, not blocking) - **Medium risk with flags, or any High risk** → Pause and route to a human approver: - Determine the correct approver (requester's manager for role mismatches; network security team for High-risk segments; sponsor for contractor requests) - Send an approval request via Slack or email with full context: requester, justification, segment, duration, device info, and the specific risk flag(s) raised - Support approve/deny actions directly from Slack/email without needing to open another tool - If denied, notify the requester in the chat interface with the reason (if provided) and next steps (e.g., "contact your manager for an exception") - If not actioned within a defined SLA (e.g., 4 business hours), escalate to a backup approver or the network security team lead 4. **Provisioning** - On approval (auto or human), call the appropriate NAC/IPAM API to: - Assign the device/user to the correct VLAN/segment/VPN group - Generate any required credentials, certificates, or registration codes - Set an explicit expiration timestamp based on requested/approved duration - Log the full record: requester, approver (or "auto-approved" + policy rule matched), segment granted, device, duration, risk tier, and timestamp 5. **Expiration & Cleanup** - Run a scheduled check (e.g., hourly) for any access grants past their expiration timestamp - Automatically revoke expired access via the NAC/IPAM API - Notify the requester in advance (24 hours before expiration) with the option to request an extension through the same chat interface — this re-triggers policy/risk evaluation rather than blindly renewing 6. **Conversational Edge Cases** - If the requester asks a status question ("what's the status of my VPN request?"), the agent should look up the record and respond conversationally - If the requester's input is ambiguous or missing required detail, the agent should ask a clarifying follow-up rather than failing or guessing - If the requester asks for something outside policy entirely (e.g., permanent unrestricted access), the agent should explain policy limits and offer the closest compliant alternative ## Data & Audit Requirements - Every request, decision (auto or human), and provisioning/revocation action must be logged to a queryable record store with full audit trail: who requested, what was requested, risk tier assigned, who approved (or which policy rule auto-approved), what was granted, and when it was revoked - This log must support compliance reporting and be searchable by requester, approver, segment, or time range ## Integrations Needed (adjust based on actual stack) - Identity: Okta or Entra ID (SSO, role/department lookup, contractor sponsor validation) - NAC/Network: Cisco ISE, Aruba ClearPass, Forescout, or Infoblox IPAM - Notification/Approval: Slack and/or email with interactive approve/deny actions - Policy reference: Tines Record or connected database mapping segments to risk tiers and duration limits ## Success Criteria - Requesters can get access through natural conversation without needing to know exact segment names or fill out rigid forms - Low-risk requests are resolved in seconds with no human bottleneck - Sensitive/high-risk requests always have a human-in-the-loop checkpoint with full context provided to the approver - No access persists past its approved expiration - Full audit trail exists for every decision, whether made by policy logic or a human approver
What this prompt builds
A conversational AI agent that replaces rigid network access request forms with natural language chat, auto-approves low-risk requests via NAC/IPAM integration, routes sensitive requests to human approvers with full context, and automatically revokes expired access while maintaining a complete audit trail. Designed for organizations where IT and security teams manage network segmentation, VPN access, and guest Wi-Fi provisioning. Reduces manual triage burden, accelerates routine access grants, enforces policy rigor on high-risk requests, and eliminates lingering unused access.
The problem
Network access requests typically involve rigid ticketing forms, manual approval chains, and inconsistent policy enforcement — leading to slow turnaround for legitimate low-risk requests, forgotten access that never gets revoked, and thin audit trails for compliance. IT and security teams want strict policy enforcement and auditability, but end users need fast, frictionless access without learning VLAN names or filling out multi-field tickets. This conversational chatbot absorbs that complexity by gathering context naturally through chat, performing instant risk classification against policy, auto-approving routine requests while provisioning access directly via NAC/IPAM APIs, and routing only high-risk or flagged requests (contractor access, role mismatches, sensitive segments) to human approvers with full context in Slack or email. It automatically expires and revokes access at the end of the approved duration and logs every decision for audit — delivering faster access for employees and contractors, reduced manual burden on IT and security teams, tighter enforcement where it matters, and a complete compliance-ready trail.
Solution and impact
This workflow replaces manual access-request triage with a conversational agent that auto-provisions low-risk network access instantly, routes sensitive requests to the right approver with full context, and automatically revokes expired access — all while maintaining a queryable audit log of every decision. Employees and contractors get frictionless access in seconds for routine requests, IT and security teams eliminate manual processing of every ticket and gain policy-driven rigor on high-risk grants, managers approve only when judgment is genuinely needed, and compliance teams have a complete trail without added overhead. The result is faster provisioning, tighter security posture, elimination of stale access, and zero compliance gaps.