How to build a helpdesk with AI

These are the parts a helpdesk needs before it is useful. Each one is something the plan covers.

• Support teams and customer accounts: Separate teams, brands and customer organisations.
• Agent, requester and admin roles: Agents work tickets, requesters see their own, admins configure SLAs.
• Ticket lifecycle: Capture, triage, work and close requests.
• Inbound email to tickets: Create and update tickets from email threads safely.
• SLA clocks and escalation: Measure and enforce response and resolution targets.
• Assignment and routing rules: Send tickets to the right queue and person.
• Asset register: Track IT hardware and software assets, owners and lifecycle.
• Ticket and asset audit trail: Show who changed priorities, closed tickets, or reassigned assets.
• Support analytics: Measure workload, SLA attainment and satisfaction.

• Knowledge base and canned replies: Deflect tickets and speed up answers.
• Change and incident records: Plan risky changes with approval and keep incident history.
• Satisfaction surveys: Collect a rating after resolution.

Each table below becomes part of the Postgres database that belongs to your app.

• requester_orgs: tier drives the SLA policy.
• tickets: kind: question|incident|problem|change|task; status: new|open|pending|on_hold|resolved|closed; number is allocated from a per-org counter row.
• ticket_events: kind: reply|note|status|assignment|merge|system; visibility: public|internal; insert only.
• sla_policies: targets: first response and resolution minutes per priority.
• ticket_sla_timers: state: running|paused|met|breached; due_at is recomputed on pause and resume against the business calendar.
• routing_rules: first matching rule wins; every evaluation outcome is logged on the ticket as a system event.
• articles: full-text index over title and body for search.
• assets: status: in_stock|assigned|repair|retired|lost; assignment history in asset_assignments (asset_id, assignee_id, from, to).
• change_requests: CHECK (approved_by is distinct from requested_by); status: draft|pending_approval|approved|implemented|failed|rolled_back|closed.

• No live chat, voice or SMS in phase 1.
• No AI auto-resolution of tickets without human review.
• No network discovery or agent-based asset scanning.
• No remote desktop or endpoint management.
• No certification of ITIL or ISO compliance.
• Live chat, phone and SMS channels: plan for it, build it later.

Forge64 matches your description to one of 25 reference architectures, writes the plan and data model into the project, generates the app, and runs a quality gate over the code before you see it. Each app gets its own Postgres database and sign-in with no per-user fee. Hosting on Forge Cloud and Forge Payments are rolling out.

This is a reference design for the category, not a promise that every item is generated today. Built-ins such as background jobs, file uploads and search are rolling out; email sent from your app is coming soon, so anything here that needs it is something you add yourself for now.

What does a helpdesk need first?

Support teams and customer accounts, agent, requester and admin roles, ticket lifecycle, inbound email to tickets, sla clocks and escalation and assignment and routing rules. The plan for a helpdesk covers all of these before it generates code.

What should I leave out of the first version of a helpdesk?

No live chat, voice or SMS in phase 1. No AI auto-resolution of tickets without human review. No network discovery or agent-based asset scanning.

Which products define the helpdesk category?

Well-known examples include Zendesk, Freshdesk, Jira Service Management, Help Scout and Snipe-IT. Use them to decide which features your users expect, then describe your own version.

• Helpdesk architecture
• Example prompt for a helpdesk
• Helpdesk security checklist
• Helpdesk scaling checklist
• Often combined: SaaS dashboard
• Often combined: CRM
• All app types