The Crew

It answers “where’s my order?”

It never opens your wallet.

Support Triage reads every ticket and shipping exception the moment it lands. A routine delivery question gets a grounded draft, built from a live shipment scan. Anything touching money (damaged, refund, replacement) gets flagged for you instead. The agent can suggest; only you can spend.

LiveOperated & SovereignOff until you switch it on
Support Triage, this morningSample data
#RS-9214WISMO · Order #RS-9214

Drafted — waiting on your review

#RS-9188Damaged · High priority

Flagged for you — no draft attempted

#RS-9177NDR · reattempt needed

Drafted — waiting on your review

One ticket. Two very different paths.

Most helpdesk AI answers everything the same confident way — including the tickets where the honest answer costs money. Support Triage reads the ticket first, then decides whether it's allowed to write a word.

Ticket lands
Routine — “where’s my order?”Touches money — damaged or refund
Ticket #RS-9214
WISMO · Order #RS-9214Sample data
The draft

“Hi Ankit — your Signature Trio is with the courier, scanned at the Pune hub 6:42am today. Typical transit from there runs 1–2 days; I’ll flag it myself if it hasn’t moved by tomorrow.”

Grounded onlive courier scan · 6:42am · Pune hub

No refund, replacement or delivery date promised. That isn’t the agent’s to give.

Waiting in your queue — your “Send this reply” click is the only send
Ticket #RS-9188
Damaged · High prioritySample data
No draft attempted

“The seal was broken and the tin arrived dented — Order #RS-8890, ₹2,140.”

Auto-priority: a damaged report and an order over ₹500 both raise a ticket to high.

Sent straight to your Action Center. Flagged, never softened by AI copy
Either lane, the ticket’s status (open, in review, resolved, escalated) only changes when you change it. The agent drafts and flags; it never touches the record.

A draft that shows its working.

The mechanics behind the routine lane: grounded on a live read, not a language model's best guess at what a support agent might say.

Reads the real shipment scan, not a guess
Before it writes a word, it pulls the live courier tracking status for that order. If the tracking feed has nothing yet, it says so. It never invents an ETA to sound helpful.
Ties to the actual order, not a template
Order number, product and the promised delivery window come from the order record itself, the same one your ops team sees, rather than a canned WISMO reply with the blanks filled in.
Off the moment there’s no key
Without a live model connected, it still reads and triages every ticket correctly and flags what needs you. It just never fakes a draft to fill the gap: no key, no invented reply.
Ticket desk, at a glanceSample data
Auto-priorityDamaged, or order > ₹500 → High
StatusesOpen · In review · Resolved · Closed · Escalated
Reopen window7 days after resolve
First-response target4 hours

Support Triage reads this same desk. It doesn’t bolt on a second inbox.

Not a dead-end chatbot — your Action Center.

Every draft and every flag lands in the same inbox as your reviews and shipping alerts. Notice the WISMO ticket carries an “Escalate with AI” button — the damaged one doesn't, because there's nothing for the agent to draft.

Action Center

2 items need your call. Support Triage handled the rest.

Representative
  • AI-drafted reply · you review before it sends

    Hi Ankit — your Signature Trio is with the courier, scanned at the Pune hub 6:42am today. Typical transit from there runs 1–2 days; I’ll flag it myself if it hasn’t moved by tomorrow.

    Send this replyOnly your click sends it. Support Triage never does

Every rule here is code, not a setting.

This is the whole differentiator vs a generic helpdesk AI: it can acknowledge, explain and cite tracking — it can never spend your money on your behalf. None of that is configurable.

Words this draft is not allowed to reach for
RefundReplacementDiscountWallet creditPickupA specific date
A reply never promises money: no refund, replacement, discount, wallet credit, pickup or a specific date.
Damaged or refund-flavoured tickets skip the draft entirely — flagged for you, never softened by AI copy.
It never changes a ticket’s or an order’s status — open, in review, resolved, escalated: your call, always.
No live model connected? It still triages and flags correctly, and it never fakes a draft to fill the gap.
Support Triage runs at all — off by default; this switch is yours

The in-ticket Razorpay refund launcher stays exactly where it is today — a human action, on a human’s click. The agent can flag that a ticket needs one; it can never fire it.

Operated & Sovereign

Deflect the routine. Escalate the rest.

Support Triage ships with Operated and Sovereign, already wired in. Nothing to bolt on, nothing to configure. Off until you turn it on; grounded and human-approved from the first ticket.