Ecommerce Automation in 2026: Flow's Internal Email Is for Staff

Ecommerce automation is software that carries out store tasks without a person doing each one: holding risky orders, flagging low stock, hiding sold-out products and messaging shoppers. Shopify’s help center splits that list in two. Its Send internal email action takes no variable recipient, so Shopify says that action suits staff and points customer email to marketing automations.

That split is written into Shopify’s help center, not inferred. The page for Shopify Flow’s Send internal email action says the address field does not accept variables, so the action “is best used to send emails to staff. To automate emails to customers, create a marketing automation.” Shopify, which calls Flow “an ecommerce automation platform”, draws a useful line in that one note about one action: automation that changes the store or tells the team, and automation that markets to a shopper. This guide maps both halves for a Shopify operator, reading Shopify’s and Zapier’s own pages as served on October 15, 2026. It is published by Relvino, which decides per shopper across email, SMS and pop-ups, so it has a stake in the second half; read it as informed but interested.

What ecommerce automation covers

The page-one definitions agree on the broad shape. Zapier calls it “the process of using tools to streamline online store operations and reduce manual work”, and its examples run from routing a new order to fulfillment to sending a follow-up when a checkout is left unfinished. Shopify’s own description of Flow is the same idea in platform terms: it is “an ecommerce automation platform that you can use to automate tasks and processes within your store and across your apps.”

In practice the work falls into five areas for a Shopify store:

• Orders and fraud. Tag, hold, cancel or capture payment depending on risk, discount codes or customer history.
• Inventory. Alert someone when stock runs low, hide a product when it sells out, ask a vendor to reorder.
• Fulfillment. Combine same-day orders from one customer, flag late shipments, route orders to a location.
• Catalog and admin. Catch duplicate SKUs, publish or unpublish products, keep tags and metafields tidy.
• Shopper messages. Abandoned checkout and cart reminders, welcome emails, post-purchase and win-back messages.

The first four change the store or inform the team. The fifth lands in a customer’s inbox or phone, and that is where Shopify’s own documentation changes products.

Where Shopify draws the line

Flow’s vocabulary is three words: a trigger starts a workflow, conditions test the data, actions do the work. Its action reference groups the actions into store actions, communication actions, app actions and advanced actions. The communication group is described as covering “emailing yourself or your team, posting a message in Slack, or sending a marketing email to your customers.” Two of those three go to staff. Flow’s own action list also includes order paperwork that goes to a customer, such as Send order invoice, whose recipient address is its own required field, and Send payment reminder, which goes to a B2B company’s main contact, but its general-purpose email action is Send internal email, and its recipients are typed in as fixed addresses.

Marketing email to customers belongs to Shopify Messaging, whose marketing automations page states that those emails “are sent only to customers who are subscribed to marketing.” The two products point at each other. Messaging’s page says that a custom marketing automation, or one involving a third-party app, can instead be built as a workflow in Flow, and its requirements page says “For full functionality, your store needs to have the Shopify Flow app installed.” Flow is the engine under both halves. What differs is who may be on the receiving end and which rules come with that.

The trigger pages show the same seam. Flow’s “Customer left online store without making a purchase” trigger lists five actions it can use on its own: add customer tags, remove customer tags, update or remove a customer metafield, and update the customer note. Every one of them edits the customer record. The templates listed on that same page are the ones that send email, and they carry a consent and frequency rule with them: each “excludes customers who received an abandoned checkout, cart, or product browse email in the last 14 days.” Recording what a shopper did is a store task. Writing to the shopper is a marketing task with its own guardrails.

The staff side: tasks with one right answer

The tasks Flow handles well share a property. Each has an outcome that a rule can state in advance and a check can confirm afterward. An order flagged as high risk should not ship until someone looks at it. A variant at zero stock should not be for sale. Two orders from the same customer on the same day should go out in one box. Shopify’s condition model fits this exactly: its help center says “Each condition step has 2 possibilities”, true when the criteria are met and false when they are not.

The template library reads like an operations manual. One template holds fulfillment for customers with a chargeback on another order from the past 60 days and emails the operations team. Another, in Shopify’s words, will “send an email to your vendor initiating a reorder” when a product runs low. A third cancels orders flagged as high risk, refunds and restocks the items, tags the order and the customer, and sends a notification, and Shopify’s own description of it ends: “We recommend manually checking orders after cancellation.” Even on the side of automation with one right answer, the vendor tells merchants to look twice at the step that cannot be undone.

For these jobs, automation is simply better than a person doing the task by hand. It does not forget the Sunday-night order, it does not tag inconsistently, and it leaves a run log. Zapier’s page makes the same case: “Automation also reduces human error, since tasks are handled the same way every time.”

What Flow’s own limits say about its design

Flow’s documented limits are worth knowing before building on it, and read together they describe a tool built for predictable store work.

Runs are deterministic by design

The Run code action executes JavaScript, but Shopify notes that it does not support importing modules or making external fetch calls, and “Random and clock-based functions can't be used.” Triggers are, in the help center’s phrase, “very literal, with rigid definitions”: Order created and Order transaction created are different events, and choosing the wrong one quietly breaks a workflow. Each workflow can have only one trigger step. The same inputs produce the same run, which is exactly what an operations team wants from an order hold.

Time and volume have ceilings

The Wait action pauses a workflow “for a maximum of 90 days per workflow”, and Shopify suggests a scheduled workflow for anything longer. Data is fetched again when a wait ends. Shopify’s own example is a customer whose lifetime spend crosses $750 with a new order, who then cancels during a seven-day wait: when the workflow resumes, it reads the lower total and the VIP tag is never applied. The Scheduled time trigger runs at most every 10 minutes, and “Get data actions can only run on lists up to 100 items long.” The optimization guide warns that a Get data action with no query filter means the API “will either return all resources or none”, which can make a workflow act on the wrong records.

High-volume triggers multiply mistakes

The same guide singles out the inventory quantity trigger, which runs on every stock change, and notes that a workflow fetching 100 products and running 1,000 times a minute fetches 100,000 products a minute. Its advice is to combine workflows that share a trigger so they do not conflict, and to put heavy logic behind a condition and a short wait.

• Cancel a high-risk order · Shopify tool its help center points to: Flow template (Order risk analyzed) · Documented limit or rule: “We recommend manually checking orders after cancellation.” · Who receives the result: The store record and the team
• Reorder when stock is low · Shopify tool its help center points to: Flow template, Send internal email · Documented limit or rule: Recipient addresses are fixed, not variables · Who receives the result: The vendor or buyer named in the workflow
• Daily list of late fulfillments · Shopify tool its help center points to: Flow, Scheduled time trigger with Get data · Documented limit or rule: Every 10 minutes at most; 100 items per Get data · Who receives the result: Staff
• Pause before a follow-up step · Shopify tool its help center points to: Flow, Wait action · Documented limit or rule: 90 days per workflow; data refetched after the wait · Who receives the result: Whoever the next step addresses
• Recover an abandoned checkout · Shopify tool its help center points to: Shopify Messaging marketing automation · Documented limit or rule: Subscribed customers by default · Who receives the result: The shopper
• Welcome a new subscriber · Shopify tool its help center points to: Shopify Messaging marketing automation · Documented limit or rule: Sent only to customers subscribed to marketing · Who receives the result: The shopper

The shopper side is a different kind of task

A message to a shopper fails the test the staff side passes. There is no outcome that can be written down in advance as correct. Whether a given shopper should get a reminder an hour after leaving a cart, a text the next morning, a discount, or nothing at all depends on that shopper, and the result can only be judged afterward, against what would have happened without the message.

Shopify’s own considerations page warns about overlap between message automations: “if you activate two welcome email automations, then your customers receive twice as many emails.” Consent adds another layer the operations side never touches, since a marketing message may only go to someone who agreed to receive it, and every send spends a little of that agreement.

That is why this map treats shopper messaging as its own slice rather than one more workflow type. The detailed playbooks live elsewhere: the abandoned cart recovery on Shopify guide covers cart and checkout reminders, and ecommerce email automation covers the email programs themselves. Where machine learning sits in those programs, inside the rules or in place of them, is the subject of AI marketing automation.

Outside Shopify: connector tools

Not every store runs everything inside Shopify. Connector platforms join the store to a help desk, a spreadsheet, a shipping tool or an ad account. Zapier’s page describes the model plainly: “You set the rules once, and the system runs them every time.” Its own product line on the same page offers “reliable, deterministic automation across 9,000+ apps.” The examples it gives are the staff-side kind: post new paid Shopify orders to Slack, add new customers to a Google Ads list, grant access to a digital product after purchase.

Flow covers part of this ground natively. Its action reference names connector apps such as Google Sheets, Klaviyo and ShipStation, and Flow can call outside services through its request actions. A store that already lives in Flow usually adds a connector platform only for tools Flow does not reach.

Where agents fit on the map

The newest entries in ecommerce automation are agents, and they sort onto the same two halves. Some act on the store: Shopify’s help pages now mention Sidekick, which can draft a Flow condition from a plain-language request, and the result is still a workflow with fixed rules. Others act toward shoppers, deciding what to say and to whom. A third group buys on the shopper’s behalf, which changes what reaches the store in the first place; agentic commerce on Shopify covers that side. The AI agents for ecommerce guide sorts the full field.

Where Relvino fits

Relvino works only on the shopper half. It decides per shopper, in under 80ms, whether to send an email, a text or a pop-up, or nothing, inside guardrails the brand sets, such as margin floors, channels and brand voice. Its Large Retail Model is trained on 7M+ data points across 10K+ retailers, so a store gets category knowledge before its own history is deep. It does not tag orders, hold fulfillment or reorder stock; those workflows stay in Flow, where Shopify built them to live. The technical cutover takes about 30 minutes, and revenue is measured in a 14-day pilot beside the existing setup.

Reading the map

For an operator planning what to automate next, the useful question about any task is who receives the result. If the answer is the store record, the team or a vendor, the task is usually one with a single right answer, and Flow’s documented limits describe a tool built to execute it the same way every time. If the answer is a shopper receiving marketing, the task has consent attached and a result that can only be read after the fact. Shopify did not leave that distinction to interpretation. It wrote it into the help page for Send internal email, Flow’s general email action, whose address field takes only fixed addresses and whose note points customer email to marketing automations.

Frequently asked questions

What is ecommerce automation?

It is the use of software to carry out repeatable store tasks, from order handling and inventory to customer messages, without a person doing each one. On Shopify the native tool is Shopify Flow, a separate app installed from the Shopify App Store, which builds workflows from triggers, conditions and actions. Shopify’s help center now also mentions Sidekick, which can write a Flow condition from a plain-language prompt such as continuing only when an order is over $100 and the customer is tagged VIP.

What is the 80/20 rule in ecommerce?

The 80/20 rule, or Pareto principle, is the rule of thumb that a small share of causes produces most of the results: a minority of products, customers or channels accounting for most of the revenue. It is a heuristic, not a measured law, and the real split differs by store. Applied to automation, it argues for starting with the few high-volume, error-prone tasks, such as order holds and low-stock alerts, before the long tail of one-off workflows.

Is Shopify Flow free?

Yes. Shopify’s help center calls it a free app available on the Basic, Grow, Advanced and Plus plans. Some features are plan-gated: the Send HTTP Request action needs Grow, Advanced or Plus, and tasks created by custom partner apps need Plus. Usage limits also differ by plan, because Flow follows the API limits that each plan carries.

What happens if a Shopify Flow workflow is edited while runs are waiting?

Runs paused on a Wait step pick up the edit. Shopify’s Wait documentation says that if conditions or actions after the wait are changed, a run that resumes uses the updated version of the workflow rather than the original, and deactivating or deleting the workflow cancels the waiting runs. Shopify suggests checking the Activity page for runs in waiting status before changing an active workflow.

Can Shopify Flow connect to apps outside Shopify?

Yes, in two ways. Installed connector apps add their own actions to a workflow, and the Send HTTP Request action calls outside services on the plans that include it. A third action is easy to mistake for a route out: Send Admin API request, which Shopify says can use most mutations in the GraphQL Admin API, including ones not yet available as Flow actions, changes the store itself rather than reaching an outside tool.

How long does it take to migrate to Relvino?

The technical cutover takes about 30 minutes: connect the Shopify store, point the sending domain, and shopper data ingests automatically. Revenue is then measured in a 14-day pilot beside the existing setup. Operations workflows in Shopify Flow, such as order holds, fraud tags and low-stock alerts, keep running untouched, because Relvino replaces the shopper messaging layer and not the store’s operations automation.