Support System: Product Roadmap

Retrospective and forward-looking, but not a commitment. Items are ordered by what unblocks other work, not by estimated demand, because there is no demand data. Nothing here has a date.

The roadmap is built from the PRD, where every requirement carries an implementation status. The Now items are the "Not implemented" and "Partial" requirements that make the existing system more trustworthy. The Next and Later items add capability.

---

Now: make what exists trustworthy

1. Continuous integration

Run the backend tests, the frontend unit tests, the lint and token check, the production build, and the browser tests on every change. At present none of them run automatically.

• Closes: OPS-07.
• Depends on: nothing.
• Done when: a failing test blocks a change.

2. Direct tests for tenant isolation and request security

Row-level security, CSRF protection, refresh-token reuse detection, and role checks are exercised only by an end-to-end smoke script. Add isolated tests that fail loudly if a table loses its policy or a route loses its permission.

• Closes: SEC-08.
• Depends on: CI, so the tests run.
• Done when: removing a policy or permission fails a test.

3. Complete the OpenAPI document

Cover every public route so an integrator does not need to read source code.

• Closes: OPS-06.
• Depends on: nothing.
• Done when: a check compares the registered routes to the document.

4. Instrument the missing metrics

Reopen rate, time spent paused, and SLA performance by queue. The first two matter most: they are how the pending-pauses-the-clock rule gets watched.

• Closes: SLA-08, and the gaps in the KPI framework.
• Depends on: nothing.
• Done when: a supervisor can see how long tickets sit in pending, by agent.

5. Decide the open lifecycle questions

Should a customer reply on an on-hold ticket notify the assignee? Should warning lead time be configurable? Should pending have a maximum duration?

• Closes: the open questions in the PRD.
• Depends on: nothing, though real users would answer them better.

---

Next: make it usable by someone else

6. Hosted staging environment

Deploy to a reachable environment with object storage for attachments, and a real mail setup. Today the system only runs locally on seeded data.

• Closes: CHN-08.
• Depends on: CI.

7. Customer portal

Let a customer sign in, see their tickets and status, and reply without email.

• Depends on: a hosted environment, and a decision about customer identity.

8. Single sign-on

A requirement for most larger workspaces.

• Closes: SEC-07.
• Depends on: a hosted environment.

9. First conversations with real support teams

Not an engineering item, and probably the most valuable one. Test the personas, the lifecycle rules, and the competitive hypotheses against people who run support.

• Depends on: a hosted environment, so there is something to show.

---

Later: capability that needs the above

10. AI-assisted triage and suggested replies

Classification, priority suggestion, and drafted replies that an agent reviews before sending.

• Depends on: real ticket data and a human-review rule. A model that sends replies unsupervised is a different and riskier product.
• Not started.

11. Routing recommendations for supervisors

Suggest rule changes from historic load, such as a skill that is a bottleneck.

• Closes: RTE-10.
• Depends on: instrumented metrics and real volume.

12. Billing, plans, and usage metering

Only meaningful with a deployment and a customer.

---

What was deliberately left out

Assumptions

• Real users would change this order. The roadmap assumes trust comes before capability, and that assumption has not been tested.
• Effort is not estimated. Order reflects dependency, not size.