
How RedTape runs the order lifecycle from checkout to doorstep, and back
One of India's best-known footwear brands runs customer conversations on an AI employee with write access into its couriers, warehouse, and storefront: it tracks orders against the source of truth, reschedules failed deliveries directly with the courier, executes returns and exchanges front to back, and sells from the live catalogue, on WhatsApp and email, in ten languages.
Challenge
A shoe order generates work at every stage: sizing doubt before checkout, delivery anxiety in transit, failed courier attempts, returns inside a 7-day window, and 45-day defect claims. Most of it lands on WhatsApp, in whichever of ten languages the customer thinks in, and most tools can only look things up and fetch a human.
Solution
An AI employee wired into the storefront, the warehouse system, the returns platform, and four courier partners, with write access where it matters: it books redeliveries, cancels shipments, updates addresses, builds carts, and generates invoices instead of just answering about them.
Results
- Failed Ekart deliveries rescheduled or returned to origin directly with the courier, zero human touch
- Returns and exchanges executed front to back, portal walkthrough included
- Every handover arrives pre-briefed, so the human starts at minute five, not minute zero
Shoes, bought sight unseen
RedTape is one of India's best-known footwear and fashion brands: shoes, apparel, and accessories sold online at redtape.com and through stores across the country.
Footwear is a demanding thing to sell online. Every order carries sizing doubt before checkout, delivery anxiety while the box is in transit, and a 7-day return window the moment it arrives. The customer asking about all of it writes on WhatsApp, in English, Hindi, Hinglish, or one of seven other languages, and expects the answer in the same one.
Between dispatch and doorstep
The hardest conversations in footwear e-commerce happen after the order is placed. A delivery attempt fails and the courier's log says one thing while the customer says another. A parcel turns back as return-to-origin. A return is eligible, or isn't, by a margin of one day. An exchange hinges on whether the next size is actually in stock.
Most AI agents are read-only: they look the answer up, apologise, and fetch a human. The lookup was never the hard part. Acting on it was.
Wired into the whole stack, with write access
RedTape's AI employee runs on Sagepilot with real system access on both sides of the conversation. It reads from Shopify for orders and the live catalogue, from EasyEcom, the warehouse system that serves as the source of truth for what actually happened to a shipment, from Return Prime for live return and exchange status, and from Ekart, Shadowfax, Delhivery, and Shiprocket directly for courier tracking and failed-attempt records.
And it writes back. It books redeliveries with Ekart on a date the customer confirms, marks shipments back to origin when they choose to cancel, raises formal escalations with Shadowfax, updates addresses on undispatched orders, generates invoices, builds ready-to-checkout carts, and emails named logistics contacts directly when a delivery ETA is breached, with the full shipment dossier attached.
Use cases across the lifecycle
The same employee covers the journey end to end, identifying the customer silently from the channel before they have typed anything:
- Discovery and selling: searches the live catalogue by article, size, and colour, presents five to six products as rich cards with images and prices, reads buying intent, and hands back a ready checkout cart link
- Order tracking: pulls the order from both the storefront and the warehouse system, reconciles the two, and answers with carrier, tracking number, and a clean link, or a full journey summary when the order is already delivered, returned, or refunded
- Failed deliveries: reads the actual courier record, tells the customer exactly what was logged, then books the redelivery or sends the shipment back, directly with Ekart
- Return-to-origin orders: confirms against the warehouse, explains the refund path by payment method, and guides the reorder in one consolidated message
- Returns and exchanges: verifies delivery, calculates the 7-day window, collects the right product photos, walks the customer through the portal, and checks size availability before promising an exchange
- Self-serve: stock and restocks, invoices, COD limits, delivery timelines, store hours and locator, gift cards, and bulk order routing
It also holds the line on policy without escalating: expired windows, hygiene items, missing tags, and claims without an unboxing video are declined firmly and finally, because a human could not overturn them either, and pretending otherwise costs a headcount and still disappoints the customer.
Handovers that start at minute five
Escalation happens only where a human has a lever the AI does not: cancellations and address changes past dispatch checks, delivered-but-not-received disputes, custom GST invoices, comfort and fit complaints inside the window, and manufacturing defect claims inside the 45-day window, which go to the product team with photos, invoice, and contact details already collected.
No transfer arrives as a bare ticket. Every one carries the order, the verified status, what was already checked, the evidence collected, and the specific reason a human is needed.
The discipline that makes write access safe
Write access into couriers and commerce is only safe with hard verification rules, and RedTape's agent runs on them: no delivery status is ever stated without checking the warehouse system first, and the warehouse wins any conflict with the storefront or the courier. If verification fails twice, the agent hands over rather than guess.
- Never invents a delivery date, an ETA, or a discount
- Quotes the courier's exact wording, never paraphrased or upgraded
- Confirms and re-confirms dates with the customer before anything is booked
- Never announces an action as done unless it completed in that same moment
- Never tells a customer they were escalated unless the escalation actually happened
- Resolves on the channel the customer arrived on, never bounces them elsewhere
The tone carries the same care: a genuine apology first when an order missed an occasion, one consolidated reply instead of holding messages, twenty to thirty words when that is all it takes, and a clean close once the case is resolved.
From a size and a budget to a ready cart
The agent searches the live catalogue, presents five to six products as rich cards with images and prices, reads buying intent without waiting for a magic phrase, and hands back a ready checkout cart link.
- Live catalogue searched by size, colour, and budget
- Five to six picks shared as rich product cards
Cart Link Sent · Ready To Checkout
Answered from the source of truth
The agent identifies the customer from the channel, pulls the order from both the storefront and the warehouse system, and reconciles the two before saying a word. Delivered, returned, or refunded orders get a full journey summary instead of a stale tracking link.
- Order pulled from the storefront
- Cross-checked against the warehouse, the source of truth
- Live status fetched from the right carrier
- Carrier, tracking number, and a clean link sent back
The failed attempt, fixed with the courier
When a delivery attempt fails, the agent reads the actual courier record, tells the customer exactly what was logged, and then acts: a redelivery booked on a date they confirm, or the shipment sent back, directly with Ekart. Even complaints about courier behaviour close without a human.
- Failed-attempt record read, exact wording quoted
- Redelivery booked directly with the courier
Returns run front to back
Delivery verified, the 7-day window calculated, condition checked, the right photos collected, and the portal walked step by step, with exchange sizes confirmed against live stock before they are promised. A human enters only if the portal itself fails the customer.
- Delivery verified, 7-day window calculated
- The right product photos collected, condition checked
- Next size checked in live stock before promising
- Walked through the returns portal, step by step
Handovers that arrive pre-briefed
Escalation happens only where a human has a lever the AI does not. And no transfer arrives as a bare ticket: every one carries the order, the verified status, what was already checked, the evidence collected, and the specific reason a human is needed.
- Warehouse and courier records pulled and compared
- Timeline and evidence collected from the customer
- Routed to Support with the full dossier attached
More stories



Join the brands that run their customer experience on Sagepilot
Live in about 48 hours, across every channel your customers use.


