Shopify

Your assistant reads the orders themselves rather than a dashboard tile, so you can ask in plain language and get the specifics back. It can pull orders inside a date range, open any single order to see line items, financial status and fulfillment status, and read shop-level settings so the currency and time zone in the answer match your store. Nothing is written back, so you can ask the same question five different ways while you work out what you actually wanted to know.

What you end up with. A written recap of the week's orders with the individual orders behind each number, so you can open the ones that look odd.

Start with it bringing you the information and you deciding.

Never without. Tell it which time zone and currency to report in, because Shopify stores orders in the shop's own settings and a recap that silently mixes presentment and shop currency will not reconcile against your payouts.

Built on: Shopify GraphQL orders and order search / Get order / Query Shop (GraphQL)

Customer search accepts criteria like order count, total spend, location and dates, so the assistant can build the list of people who used to buy and have not lately, then open individual records to see what they bought. It reads only. The list comes back as names and order histories you can read, not as a segment saved into Shopify.

What you end up with. A named list of lapsed customers with what each one used to buy and when they stopped, ready for you to decide who is worth a note.

Start with it bringing you the information and you deciding.

Never without. Say out loud what counts as gone quiet before you ask, because six months of silence means something different for a mattress store than for a coffee subscription, and the assistant will use whatever cutoff it guesses if you do not give it one.

Built on: Search Customers / Retrieves a list of customers / Retrieve a single customer / Shopify GraphQL customers and customer search

The assistant can read inventory items and stock levels per location, cross-check them against recent order volume, and tell you which SKUs are thin relative to how fast they are moving. For a catalog large enough that this would otherwise mean paging through screens, it can run a bulk query and read the whole set at once. It changes no stock counts.

What you end up with. A short list of SKUs to reorder, each with its current level, its location, and the recent sales rate that makes it urgent.

Start with it bringing you the information and you deciding.

Never without. Check that it read every location, because a multi-location store can look well stocked in total while the location that actually ships your online orders is empty.

Built on: Shopify GraphQL inventory items and levels / Shopify GraphQL orders and order search / Run Shopify bulk query operation

Discounts and discount codes are readable through the connection, including the redemption count on each code and the price rule behind it. Put that next to the orders in the same window and you can see which codes are carrying real volume, which ones nobody has touched since the campaign that created them, and which ones are still live months later. Nothing is changed by asking.

What you end up with. A list of your discount codes with how many times each was redeemed and which are still active, and the orders behind the ones that are moving.

Start with it bringing you the information and you deciding.

Never without. A redemption count tells you a code was used, not what it cost you. Read the orders behind a code before you call it a success, because a discount that only moved buyers who were going to buy anyway returns the same number as one that brought somebody new.

Built on: Shopify GraphQL discounts and discount search / Get Discount Code / Get Discount Codes for Price Rule / Get discount codes count / Shopify GraphQL orders and order search

Shopify has a bulk operation API built for exactly this: the assistant submits a query, Shopify runs it in the background, and the results come back as a file that stays available for seven days. That is how you get thousands of orders or customers into one analysis instead of a screen at a time. Shopify allows only one bulk query at a time per store, so the assistant may also need to cancel a job that is already running.

What you end up with. A complete export of the records you asked for, in one file, without a paginated crawl or a third-party export app.

Start with it doing the work and checking with you.

Never without. Confirm no other bulk job is mid-run before letting it cancel one. Canceling discards the result file that job was building, cancelation is not instant, and Shopify permits only a single bulk query per store, so an app or a colleague may be waiting on the job you just killed.

Built on: Run Shopify bulk operation query / Run Shopify bulk query operation / Cancel Shopify bulk operation

Bulk variant updates go through Shopify's asynchronous job pipeline, so the assistant can hand it a full set of price, SKU or inventory-setting changes in one operation and report back row by row on what succeeded and what errored. You describe the rule, it builds the rows, you look at them before it submits. Prices that go through are live on the storefront as soon as the job finishes.

What you end up with. Every variant repriced in one pass, with a per-row result showing exactly which ones failed and why.

Start with it doing the work and checking with you.

Never without. Read the generated rows against a handful of products you know before submitting, and run the change on a small test set first. The two GraphQL write actions behind this cover the whole mutation surface, so the same call that edits a price can unpublish a live product or delete its media, changes hit the storefront immediately, and Shopify keeps no undo for any of it.

Built on: Bulk update product variants via GraphQL / Shopify GraphQL products, variants, media, and publishing

The assistant can create a custom collection with its title, description, image, sort order and publication settings, and it can create products in bulk if the season needs new ones. It works from whatever you give it: a spreadsheet, a supplier list, a description of the theme. Everything lands in Shopify as real records, so review the collection before you publish it to a sales channel.

What you end up with. A named collection in your admin with the products in it and the sort order set, waiting on your decision to publish.

Start with it doing the work and checking with you.

Never without. Keep the collection unpublished until you have looked at it. The same write surface that adds products also changes publications and product media, a collection published to the online store is visible to shoppers the moment it goes live, and there is no version to roll back to.

Built on: Create a custom collection / Bulk create products via GraphQL / Shopify GraphQL products, variants, media, and publishing

It can search for the duplicates and the broken records, open each one, and update names, emails, phones, tags and addresses. Customer creation and updates return their errors in a userErrors array rather than failing loudly, so the assistant can report which records it could not fix and why. Marketing consent is one of the fields it can write, which is the one to be careful about.

What you end up with. A tidied customer list plus a written note of the records that could not be fixed automatically and what is wrong with each.

Start with it doing the work and checking with you.

Never without. Exclude marketing consent fields from any bulk cleanup unless you are deliberately fixing consent, and keep an export of the customer list from before the run. Consent state is a legal record in most jurisdictions, not a formatting detail.

Built on: Search Customers / Retrieve a single customer / Update Customer / Create Customer