Look Up Order Details Mid-Conversation, Without Leaving Your Helpdesk

The fix isn't another browser tab. It's letting the agent simply ask for the order, in plain language, right inside the conversation, and getting a clean, structured answer back. That's what this Macha use case does: an agent says "pull up order #3019" and Macha looks it up in Shopify and surfaces the customer, items, fulfillment status, and tracking — without anyone leaving Zendesk.

This is the agent-assist version of order lookup. It keeps a human in the loop (the agent reads the result and decides what to send), which makes it a good first step before you trust an agent to auto-resolve order-status tickets end-to-end. Macha is an AI agent layer that sits on top of Zendesk and connects to Shopify — it's not a helpdesk you have to migrate to.

The job to be done

The native Zendesk Shopify app already does part of this well: it matches the ticket requester's email or phone to their Shopify account and shows an order accordion with date, value, payment and fulfillment status, and tracking numbers inside the ticket. That's genuinely useful, and if all you need is the requester's own recent orders shown passively in a sidebar, it may be enough.

Where it stops is exactly where agents get stuck: it's identity-bound (it keys off the requester's contact details, so a B2B agent fielding "can you check order #3019 for a customer of mine" is back to manual lookup), it's read-only context rather than an actor, and it can't read an order number out of a pasted screenshot. The agent-assist pattern below closes those gaps.

How it works

The whole interaction is one trigger, two tools, and a human checkpoint.

Trigger: a Zendesk ticket where the agent needs order context. In practice the agent invokes Macha in the conversation — an internal note or the chat panel — and asks the question.

Tools the agent uses:

• Shopify → Get Order — "Look up an order by ID or order number with full line items, fulfillment, and payment status."
• Shopify → Get Customer — retrieve the customer profile, including order history and contact details.

You wire those up once when you build the agent. In Macha's tool picker you grant the agent only the Shopify (and Zendesk) actions it should be allowed to call — nothing more.

That's the security model in one screen: an agent can only do what its enabled tools allow. A lookup agent gets read actions (Get Order, Get Customer); it does not get refund or cancellation tools unless you deliberately add them.

Before any of that, both apps have to be connected to your Macha org — a one-time OAuth for Zendesk and Shopify.

A real run

Here's an actual Macha run, captured live in the chat — not a mockup. We asked the agent, in plain language, to pull up the latest order and return the details:

The prompt was "Look up the most recent successful payment and show the customer name, email, items, amount, and payment status." Macha made three tool calls and came back, in under six seconds, with a clean structured card:

Most recent order: - Customer: Dean Legg ([email protected]) - Items: Business Plan — AI Assistant for Zendesk - Amount: $15.00 USD - Payment status: Succeeded - Order ID: 23225355

Two things matter here. First, the agent asked in natural language and Macha resolved it — no manual search, no tab-switch. Second, the result is structured for a human to act on, not a dump of an admin page: every field the agent needs to answer the customer is right there, with the source tool and timing shown for trust.

One honesty note on this exact shot: it reads the order from Stripe, the billing tool connected in our demo org (Shopify isn't connected there). The pattern is identical for Shopify — point the same agent at Shopify → Get Order and the card carries line items, fulfillment status, and a tracking number instead of (or alongside) the payment detail. And because Get Order takes any order ID or order number, the agent can ask for order #3019 even when the person on the ticket isn't the buyer — the exact B2B case the email-matched native app can't handle.

The human checkpoint

Because this is agent-assist, Macha offers the next step rather than taking it. If the agent says "yes, draft it," Macha composes the customer-facing reply and hands it back for review — the agent approves, edits, or discards before anything is sent.

This approval gate is the whole point of the assist pattern. The agent stays accountable for what the customer sees, while Macha does the fetching, formatting, and first-draft writing. When you've watched a lookup agent be right a few hundred times, that's your signal to promote the workflow to a fully autonomous order-status agent.

Where it earns its keep

A few situations make this pattern punch above its weight:

• B2B and concierge support, where the person on the ticket is rarely the person who placed the order. Email matching fails here; "pull up order #X" doesn't.
• Order numbers buried in custom fields. Macha's Zendesk Get Custom Fields tool resolves field labels, so an agent can ask "look up the order in the Order # field" and Macha reads it off the ticket.
• Screenshots and receipts. Zendesk's Read Attachment tool gives the agent AI vision — if a customer pastes a photo of their order confirmation, Macha can read the number off it and then look the order up.
• Multi-order tickets. "Compare orders #3019 and #3022" is one prompt, not four tab switches.

Because Macha is the same agent layer across your stack, the lookup doesn't have to end at Shopify. The same conversation can pull a payment record from Stripe or a customer's recent ticket history from Zendesk — whichever tools you've granted.

Watch-outs: when not to reach for this

It would be dishonest to pretend this fits every team, so here's the contrarian read.

• If you only ever need the requester's own orders shown passively, the free native Zendesk Shopify sidebar may already cover you — there's no reason to build an agent for a job a built-in accordion does. Macha earns its place when the agent needs to ask, act on arbitrary orders, or read attachments.
• It's not a free action. Every Macha lookup is one or more AI actions billed in credits (0.5–9 by model; the default GPT-5.4 Mini is 1 credit), plus the Shopify API call. For ultra-high-volume "where is my order" tickets, the math often favors letting an autonomous agent resolve them outright rather than assisting an agent on each one — assist is the stepping stone, not always the destination.
• Garbage in, garbage out on identity. Macha returns what Shopify has. If an order number is wrong or the store data is stale, so is the answer — the agent's review step exists partly to catch that.
• Permissions are on you. Grant a lookup agent only read tools. If you also enable refund or cancellation actions, keep the human approval gate firmly in place.

How to set it up

1. Connect Zendesk and Shopify to your Macha org (one-time OAuth each).
2. Build an agent and, in the tool picker, enable Shopify → Get Order and Get Customer (add Zendesk → Get Custom Fields and Read Attachment if you want field/screenshot lookups).
3. Give it a short instruction: "When an agent asks about an order, look it up in Shopify and return the customer, items, fulfillment status, and tracking. Offer to draft a reply."
4. Test it on a real ticket, then let your team invoke it in the conversation.

The full walkthrough — agents, tools, and connectors — lives in the Macha docs.

FAQ

Does the order have to belong to the ticket's requester? No. The agent can ask for any order number. That's the main advantage over email-matched sidebar integrations, which only surface the requester's own orders.

Will Macha take actions, or just look things up? By default this pattern is read-only and assist-only: it fetches the order and offers to draft a reply, which the agent approves. You can grant action tools (like refunds) separately, but the human approval gate stays in place for anything customer-facing.

Can it read an order number from a screenshot the customer sent? Yes — with Zendesk's Read Attachment tool, Macha uses AI vision to read order confirmations, receipts, or product photos, then looks the order up.

Is this only for Zendesk and Shopify? Those are the connectors in this example, but Macha is a layer across your stack — the same agent can also pull from Stripe, search past tickets, or use other connected tools in the same conversation.

How is it priced? Each lookup is billed as Macha credits per AI action (0.5–9 by model; GPT-5.4 Mini, the default, is 1). There are no per-seat surcharges for using it — see the pricing page for plan details.

Try it

If your agents are still alt-tabbing to Shopify for one tracking number at a time, this is the smallest, highest-frequency win to automate first. Start a 7-day free trial, no credit card required, connect Zendesk and Shopify, and build a lookup agent — then watch how fast "pull up order #3019" replaces five minutes of tab-switching.

Written by Abbas (Customer Support & AI, Macha) · Reviewed by Ankeet Guha (Co-founder & CTO) · Published 2026-06-24 · Last updated 2026-06-24.