NHU Group-Based Ticket Routing — Staff Training

The signed_ntv chain — full 10-step sequence (in order)

1. Internal note: "✅ NTV Signed - Notice Accepted" with auto-filled property/tenant/lease/surrender details
2. Tags added: lifecycle_ntv_signed + ntv_workflow_active + ntv_workflow_fired
3. Group set to Support & Requests
4. Assignee cleared (parks ticket in Open Pool for staff pickup)
5. Status flipped to open
6. 🔌 Utility Coordination internal note with SOP checklist — Ryan's task
7. Tag added: pending_utility_research
8. 📬 Owner Decision email draft auto-resolved with all owner/property details — Ashley copy-paste-sends
9. Tag added: pending_owner_decision
10. Roadmap internal note listing what's still queued (RentCheck auto-schedule, tenant reminders, watchdogs, magic links, PDF referral generator, downstream decision chains)

What's NOT firing automatically yet (queued for next refinements)

• RentCheck Move-Out Inspection auto-schedule (pending RentCheck automation audit)
• Tenant pre-vacate reminders (14d / 7d / 2d / 0d) — drafts only until Steven approves auto-send
• pg_cron watchdogs for utility (2 business days) and owner decision (3 business days) SLAs
• Magic-link buttons + hosted owner-decision forms (one-click UX instead of structured reply)
• PDF Referral Agreement generator for the Sell decision path
• Downstream chains by decision: owner_decision_rerent / owner_decision_sell / owner_decision_discuss

What staff should do differently starting today

• Trust the auto-routing. Within ~5 seconds of a ticket landing, the router may change its group / form / status / tags. That's expected. Don't undo it unless it's clearly wrong (flag it to Steven if so).
• NTV signature flow: AppFolio doc-signed notifications get folded into the parent NTV request ticket automatically. Look for the original NTV request ticket (with the form name Notice to Vacate) — the View Document link from AppFolio will be in the merged comment.
• One macro for NTV countersignature. The old NTV :: 1-7 stage gate macros are gone. Use ✅ NTV Signed - Notice Accepted and Submit-as-Open. The ticket flips into the Open Pool and the workflow chain fires in the background.
• Internal notes are drafts, not sends. Owner emails, tenant comms, vendor messages drafted by AI all post as internal notes for you to copy-paste-send. Nothing leaves Zendesk automatically until Steven flips that switch.
• Steven's cockpit gets only exceptions. When a step is overdue, blocked, or escalated, it shows up there. Don't worry about updating him on routine progress.

What staff should know

• Every inbound channel becomes a ticket. If something needs our team's attention, it should exist as a Zendesk ticket. Side conversations, internal Slack chatter, and one-off emails don't count as "work" until they're a ticket.
• You don't decide where it goes. The routing happens in the next two steps automatically. Your job starts in Step 4.
• The ticket has a requester — the person who created or is asking about the issue. Most tickets have the tenant or owner as requester. For internal-source tickets (digests, automations), the requester is a system user.

What staff should know

• Groups define ownership. A ticket's group_id is the answer to "who owns this?" — collectively, by the group, not by one person.
• Triage is the safety net. If the routing trigger can't make a confident decision, it falls through to Triage. Whoever scans Triage classifies the ticket and re-routes it. Repeating patterns that hit Triage should prompt a routing-trigger improvement.
• One group, many views. Inside a group like Support & Requests, separate views filter by requester type (tenant vs owner vs vendor vs agent). One ownership group, many lenses.
• The 8th and 9th groups — Sales | Call-Ins and zPMWBatPhone — exist but are dormant. Don't route to them.

What staff should know

• You'll see fewer applicant emails in Triage. Applicants whose emails are on file with AppFolio now auto-land in Convert & Commit on the Applications form. Triage is for unknown senders.
• Look for the applicant_routed tag. If a ticket has that tag, the Smart Router placed it — not a human. If you see it routed to the wrong place, flag the ticket and we'll tune the rules. Don't just re-route silently or we miss the signal.
• Mailer-alias applicants get tagged appfolio_mailer_unmatched. These come from AppFolio's anonymized email forwarders. The Smart Router can't see the applicant's real email, but it can see the pattern. Tagging keeps them visible for audit.
• The Smart Router never assigns to a person. Per NHU's Open Pool rule, automation sets the group only. Cross-trained staff pulls from the pool.
• Audit table: smart_router_log in Supabase. Brandon can pull any time to see what the router did and why.

What staff should know

• Move-out emails, renewal confirmations, and lease-signing notifications now route themselves. When an AppFolio email arrives, within seconds Zendesk's AI Copilot tags it (e.g. lifecycle_ntv), the relevant trigger fires, and the ticket lands in the correct view. No manual triage needed.
• The audit tag tells you the trigger ran. If you see appfolio_routed_ntv on a ticket, you know the routing was automatic. If a ticket should have been routed but doesn't have the tag, flag it — that's a routing miss worth investigating.
• NTV and Renewals are simplified to the rollup view. As of 2026-05-12 the granular sub-views (NTV :: Surrender Gate, NTV :: RED Escalations, Renewals :: Owner Decision Gate, Renewals :: RED Escalations) are removed. The trigger lands tickets in 📦 NTV :: All Active or 🔄 Renewals :: All Active. We'll re-introduce granular gates later when there's real volume to justify them.
• The catchall is one-time on backlog. When the catchall trigger shipped, every active AppFolio tenant ticket got Requester is filled in and tagged appfolio_routed_support. That's a one-time fix, not a recurring storm.

What staff should know

• The Open Pool is a shared queue. No one "owns" a ticket in the pool until they decide to take action on it. Everyone in the group can see everything in their pool.
• Specific assignee = signal. If your name appears on a ticket, the system is telling you: YOU are the only person who can do this work. Treat it as a personal flag, not a routine assignment. Examples: signatures land on Steven, water bills land on Ryan, HOA Elbow Creek work lands on Ryan.
• If you see your name on a ticket and you genuinely can't do the work (out of office, wrong skill set), reassign back to group-only — don't leave it stuck on you.
• If you find a ticket without a group at all, it's a triage gap — push it to Triage and tell Brandon so the routing trigger (or Smart Router) can be improved.

What staff should know

• Open Pool, not your inbox. You're not waiting for tickets to be assigned to you — you're choosing what to pull next. Inbox is push; the pool is pull. Pulling is faster.
• Cross-trained team. Whoever gets to a ticket first works it. Don't ask "is this mine?" — if it's Open in a group you belong to and not assigned to a specific person, the answer is "yes if you have time."
• ★ Calls to Merge (NEW). Voicemails routed to Maintenance with the call_overflow tag show up here. Open the voicemail, find the parent WO ticket it's about, merge them. The voicemail disappears from this view automatically once merged.
• Avoid double-work. If a teammate is already on a ticket (you see them in the activity log or they've added a comment), move on to another one in the pool.
• If two people accidentally work the same ticket, the second one to comment is the one who keeps going. Cheap to coordinate.

What staff should know

• Keep the conversation on the ticket. If a tenant texts you directly or a vendor calls your cell, copy the key info into the ticket as an internal note. Otherwise the next person who picks it up has no idea what happened.
• Use side conversations for vendors. Replies from vendors land on the parent ticket and auto-reopen it — verified working in production. Don't email them outside Zendesk.
• Voicemails about a WO go in Calls to Merge. Open the voicemail, find the parent WO ticket (search the WO number or property address), and merge. The merged voicemail disappears from the Calls to Merge view automatically.
• Internal notes are free. Cheap to write, high value for the next person. When you put a ticket on Pending or Hold, leave a quick internal note explaining what you're waiting on.

What staff should know

• The distinction matters for the next person scanning the queue. "Pending" tells them the tenant ghosted us. "Hold" tells them the vendor stalled. Different follow-up actions.
• Both states auto-reopen on reply. You don't have to babysit either one. Once a reply lands, the ticket flips back to Open and reappears in the Open Pool. Verified at 100% pass rate in production audit.
• Don't leave a ticket Open if you're actually waiting. Open means "needs our attention." If we can't do anything until someone else responds, it shouldn't be in the Open Pool — that wastes everyone's scan time.
• If you're not sure whether it's Pending or Hold, ask: "Is the person I'm waiting on the requester on this ticket?" Yes → Pending. No → Hold.

What staff should know

• Solved is the success state. Don't be afraid to use it — clean queues beat full queues. But also don't use it to make the queue look smaller when work isn't actually done.
• Solved auto-closes after 4 days in Zendesk's default. Closed is permanent — the ticket can never be re-opened. If a customer follows up after that, a new ticket is created and linked.
• If a Solved ticket reopens (via reply, side conversation, or trigger), it moves back to Open and reappears in the pool. Pick up where it left off.

What staff should know

• Trust the safety net. Once you've put a ticket on Pending or Hold with a clean internal note about what you're waiting on, you're done. No reminders to set, no calendar entry, no follow-up file.
• The trigger watches every channel. Email reply, side conversation reply, portal message — all of them flip the status back to Open and surface the ticket in your pool.
• If you notice a reply came in but the ticket didn't reopen, tell Brandon — that's a routing bug worth fixing.

The closed loop — what to do when Unassigned grows

• Smart Router never assigns a ticket to a person (per the Open Pool model). So unassigned alone doesn't mean unrouted — only unclassified means unrouted. The view tightening (2026-05-12) excludes the 4 routed classifications so the view actually reflects "needs triage."
• When a pattern shows up repeatedly in Unassigned, that's the signal to build automation. The workflow is:
  • Identify the pattern (sender domain, subject keyword, email channel, lifecycle tag…)
  • Add a trigger / extend Smart Router to classify it
  • Trigger sets WO Ticket Category → ticket exits Unassigned automatically going forward
• Goal: Unassigned Tickets count trends toward zero over time. Anything still in the view after we've watched it a few days = work the system can't yet do without a human. That's the priority for the next routing automation.
• Goal: Open Pool count is the real workload signal. Tells you how much active work is waiting on the team. Triage-by-classification stops being a daily task; pulling from Open Pool becomes the only daily task.
• Validated on 2026-05-12: tickets #86376 (Final Move Out info from tenant) and #86378 (Tenant Request Form — General Question) were both showing in Unassigned despite Smart Router having already routed them to Support::Tenant. After the view tightening, both dropped out automatically — exactly the desired behavior.
• Audit finding 2026-05-12 (late) — the "No assignee" bug: Supabase mirror audit found 163 active tickets, 26 truly unrouted — but the prior filter's "No assignee" condition hid all 26 because each had been auto-assigned by default routing. Dropping "No assignee" from the filter (v10) surfaces these as the real daily triage queue.
• Next 3 routing automations — patterns identified from the 26 audit blind spots. Each is a concrete trigger to build:
  • AR auto-tag — subject contains "invoice", "payment", "past due", or "payment agreement" → set WO Ticket Category = receivables___payables. ~7 tickets/cycle today (Athletic Cleaning invoices, Past Due notices, Payment Agreement tenant requests).
  • Tenant maintenance auto-convert — keyword match on appliance / repair / quote / paint terms with a tenant requester → tag work_order + set group to Maintenance. ~3 tickets/cycle today (Dishwasher, home warranty service, Elderberry Run Paint).
  • Reply tag inheritance — when a reply creates a child ticket OR a parent reply arrives without tags, copy the parent ticket's category + lifecycle tags forward. Compounds as reply threads multiply.
• Open follow-up for Brandon — verify AppFolio Bridge triggers set the WO Ticket Category field. The 4 AppFolio bridge triggers (NTV, Renewals, Move-In, Support::Tenant catchall) set form + tags. If they don't also set WO Ticket Category, lifecycle-routed tickets can still leak into Unassigned.

How to scan this view in 10 seconds

• First pass — find your name. Any ticket where you appear as the Assignee gets worked first. (Above: WO #5701-2 is Steven's signature work, only Steven can do it.)
• Second pass — scan the Open rows. Open + unassigned = the pool you pull from. Pick by age (older first) or priority (urgency).
• Skip Pending and On Hold. Those are waiting on someone else. They'll auto-reopen when ready. Don't waste scan-time on them unless you're investigating why something's stuck.
• Watch the Updated column. Anything that hasn't moved in 5+ days is a flag — either it's genuinely stuck or someone forgot to update the status.