How AI Agents Handle Order Status Queries on WhatsApp

"Where is my order?" is the most common message a support team receives. On WhatsApp it arrives at 11 p.m., with no order number, sometimes just a photo of a shipping email and the word "bhai??"

These are the WISMO tickets, short for "where is my order," and for most e-commerce brands they make up somewhere between 40% and 60% of all inbound support. They're low-value, high-volume, and almost entirely mechanical. Which is exactly why they're the first thing worth automating on WhatsApp, and a useful lens for understanding how an AI agent actually works under the hood.

Why WhatsApp changes the order-status problem

On email, a WISMO query is annoying but forgiving. The customer expects a reply in a few hours. On WhatsApp, they expect it in seconds, because that's how the channel feels to them. It sits in the same app as messages from their family.

There's a cost angle too. When a customer messages you first, they open a 24-hour customer service window, and free-form replies inside that window carry no Meta fee. So an agent that answers a WISMO query the moment it lands isn't just faster, it's answering for free. A proactive shipping update you push later, outside the window, gets billed as a utility template. The economics reward speed and reward letting the customer start the conversation.

That's the backdrop. Now the mechanics.

What actually happens between "where's my order" and the answer

To a customer it looks like magic: they type a half-sentence, and ten seconds later they have a tracking link. Under the hood it's four distinct steps, and each one is a place where a rule-based bot tends to fall over.

Step 1: Work out what they're actually asking

The first job is intent. "Where's my stuff," "order kab aayega," "did it ship," and a forwarded courier SMS are all the same question wearing different clothes. A keyword bot matches on "order" and "track." An AI agent on WhatsApp reads meaning, so it catches the forwarded SMS and the Hinglish version and the one that never uses the word "order" at all.

This is the single biggest reason keyword flows frustrate people. Customers don't phrase things the way your decision tree expects.

Step 2: Find the order

Now the agent needs an order to look up. Sometimes the customer gives a number. Usually they don't. So the agent resolves identity from what it has: the WhatsApp phone number is often enough to find the most recent order in Shopify or your OMS. If there are several, it asks a single clarifying question ("Is this about order #10432 from Tuesday?") rather than demanding an order ID up front.

That one design choice, resolve first and ask only if needed, is the difference between a flow that feels human and one that feels like an IVR phone tree.

Step 3: Pull the live status

With the order identified, the agent calls the real systems. This is where write and read integrations matter. It hits Shopify or your OMS for order state, then the courier API (Shiprocket, Delhivery, Blue Dart, FedEx) for the current tracking event. Not a canned "your order is on its way," but the actual scan: "Out for delivery, last scanned in Bhiwandi at 6:04 a.m."

The specificity is what kills the follow-up message. Vague reassurance generates another WISMO ticket tomorrow. A real tracking event ends the thread.

Step 4: Reply in the customer's language, in context

Finally the agent writes back. In the customer's language, with the tracking link, and with a next step if one's needed. If the parcel is genuinely stuck, it doesn't pretend everything's fine. It says the shipment hasn't moved in three days and offers to raise it with the courier or start a replacement.

A real thread, start to finish

Here's how it reads on the customer's screen. A customer messages a D2C skincare brand at 10:48 p.m.: "hi my order still not come??"

• The agent recognizes a WISMO intent with no order number and matches the phone number to a single open order placed four days ago.
• It queries the courier API and sees the parcel was marked "delivery attempted, customer unavailable" that afternoon.
• It replies: "Hi! Your order #10432 was out for delivery today but the courier couldn't reach you. Want me to reschedule it for tomorrow?"
• The customer says yes. The agent triggers the reschedule through the courier integration and confirms the new date.

No agent touched it. No ticket was created. The whole exchange took under a minute, inside the free service window, at nearly 11 p.m. That's the shape of a well-built order-status flow, and it's why WISMO is where most teams see the fastest return.

Where the agent should stop and escalate

Not every order query should be resolved autonomously, and a good agent knows the difference. The honest version of this feature has clear stop conditions.

• Lost or stuck shipments: when tracking shows no movement for several days, or the courier marks a parcel delivered that the customer says never arrived, that's a judgment call and a trust moment. Route it to a human with the full context attached.
• Refund or compensation requests: "it's late, I want my money back" is no longer a status query. It's a policy decision, and it should follow your escalation and handoff rules.
• Angry or repeat contacts: if the customer has messaged three times about the same parcel, the answer isn't a fourth tracking link. Tone-shift detection should pull a person in.

The goal isn't to automate every order message. It's to resolve the 70% that are genuinely mechanical and route the 30% that need a human, with everything the human needs already gathered.

What "resolved" has to mean here

A deflection metric will tell you the bot "handled" the query. That's not the number that matters. The number that matters is whether the customer comes back within 72 hours about the same order. If they do, it wasn't resolved, it was deferred.

This is worth being strict about because order-status automation is easy to fake. Send a generic "your order is on its way" to everyone and your deflection rate looks great while your true resolution rate quietly sinks. Pulling the real courier scan and answering the actual question is what makes the difference show up in the re-contact rate. If you're building a case for automation, tie it to autonomous resolution, not deflection.

Get the WISMO flow right and you've removed the largest, most repetitive slice of your WhatsApp queue, freed your team for the messages that actually need judgment, and done it on the one channel where customers expect an answer before they've put their phone down.

Ready to clear the WISMO queue on WhatsApp? Robylon AI resolves 60–80% of customer queries autonomously with agents that pull live status from Shopify, your OMS, and couriers like Shiprocket and Delhivery, plus 60+ other integrations. Start free at robylon.ai