How to Deflect 40% of WISMO Tickets with an AI Agent

If you run an ecommerce support queue, you already know what WISMO is — you’re just tired of answering it. WISMO (“where is my order”) questions are the single largest category of support volume for most online stores, and they’re also the one category where an AI agent can actually remove them from your queue entirely.

This post is about how. What AI agents do with WISMO specifically, what to watch out for, and how to measure whether the deflection actually happens.

The WISMO problem in numbers

Depending on who you ask, WISMO and closely-related “where is my refund” questions make up:

• 35-50% of ecommerce support volume (Gorgias and Zendesk benchmark reports, 2023–2025)
• ~$2-4 per ticket fully loaded when you count agent time, tooling, and context-switching
• Very little revenue impact either way — nobody buys more because you answered a WISMO fast, but they definitely complain loud when you don’t

Every ticket you deflect is a ticket your human team doesn’t have to touch. At a mid-sized store doing 500 tickets a week, 40% deflection is 200 tickets a week reclaimed — $400-$800 a week in support cost and probably twice that in sanity.

How an AI agent handles WISMO

The happy path looks like this:

1. Shopper opens chat, types “where’s my order?”
2. Agent asks for order number and email.
3. Shopper provides both.
4. Agent verifies the email matches the order, then calls the order-lookup tool against Shopify/WooCommerce/whatever.
5. Agent returns current status: “Order #4419 shipped on April 18 via FedEx. Tracking says it’s in Memphis, scheduled delivery Friday. Here’s the tracking link.”
6. Shopper is satisfied or asks a follow-up. If they ask “but why is it taking so long?”, agent either answers from carrier data or escalates.

No human touches this. It takes the shopper about 20 seconds.

The identity verification trap

The most important detail in that flow is step 4: the agent verifies the email matches the order before sharing status.

This is non-negotiable. If your AI agent will happily tell anyone the tracking status of any order number typed in, you have created a data-leak vector:

• Scrapers can enumerate order numbers and harvest customer emails + addresses
• Competitors can probe your order volume and shipping times
• Hostile actors can use tracking details for phishing follow-ups

Order numbers are not secrets — they appear in order confirmation emails that get forwarded, they’re printed on packing slips that get thrown out, they show up in screenshots on social media. The email-match check is what makes the flow safe.

Every vendor you evaluate should describe their identity-verification behavior before they describe anything else about their order lookup. If they don’t mention it, walk away.

What happens when the answer isn’t helpful

“Your order shipped yesterday, scheduled delivery Monday” is the easy case. The hard cases are:

• Carrier hasn’t updated in three days. The agent should say so honestly — “I can see the carrier received the package on April 15 but hasn’t scanned it since. I’ll escalate to our support team to follow up” — not invent reassurance.
• Order is lost. The agent should recognize this pattern (shipped more than 10 days ago, no delivery, no recent tracking) and escalate proactively.
• Order is delayed due to weather/carrier strike/etc. If you have policy docs covering this, the agent cites them. If you don’t, it escalates.
• Customer is angry. The agent hands off to a human. WISMO escalations from upset customers are not where you want an AI to be conservative and procedural.

A good AI agent recognizes the emotional tone of a message. “My order still hasn’t arrived and I’m so frustrated” is not the same message as “where is my order.” The first should escalate; the second shouldn’t.

Integrating order tracking

On the technical side, most WISMO deflection comes from connecting the agent to:

• Your ecommerce platform (Shopify Admin API, WooCommerce REST, BigCommerce) — this gives the agent order status, fulfillment status, and tracking numbers.
• Your shipping tool (Shippo, AfterShip, EasyPost) or directly to carrier APIs — this gives the agent live tracking details.

If the agent only has platform data and no carrier data, it can say “your order shipped via FedEx” but can’t say “it’s currently in Memphis.” Some shoppers will accept the former; most want the latter.

Measuring deflection accurately

The number you want is: of WISMO conversations the agent handled, how many did not result in a human ticket?

To measure this honestly:

1. Tag conversations by intent. WISMO, returns, product question, complaint, other.
2. Tag outcomes. Resolved in chat, escalated to human, shopper abandoned.
3. Compare WISMO volume in your support tool (Gorgias, Zendesk, Freshdesk) before and after launch, controlling for overall order volume.

Expect 30-50% deflection after a month of tuning. If you see 90% deflection in the first week, something’s wrong — the agent is probably “resolving” by refusing or redirecting, and those customers are re-emailing you or leaving.

Avoid these WISMO deflection anti-patterns

• Don’t hide the “talk to a human” option. Making escalation artificially hard boosts deflection metrics and destroys customer trust.
• Don’t auto-resolve after one response. Some conversations take three back-and-forths to actually resolve. Counting the first reply as a resolution is cheating.
• Don’t skip identity verification to improve UX. “Remove the email check, it’s a hassle” = “open data leak to boost CSAT.”
• Don’t skip tracking data. An agent that can only say “shipped” and not “currently in Memphis” converts fewer WISMO conversations to resolutions.

Getting started

WISMO deflection is the fastest-ROI feature of an AI agent widget. If you want to see it running against your actual order data, book a demo with us. We’ll wire up a test environment with read-only access to a sample of your recent orders and show you the flow live.

Or read the features page for how the Agentify Ecommerce agent handles order tracking specifically.