Short answer
Business process automation means an event (a form, an order, an invoice, an email) starts a set of steps on its own. In small and mid-sized companies the most common ones are form-to-CRM alerts, invoice matching, order and stock sync, shipping notifications, a weekly report and sorting support requests.
Monday morning. The sales lead is copying the weekend’s form enquiries from the inbox into the CRM one by one. Someone in accounts is checking bank transactions against the invoice list line by line, and someone in the warehouse is matching the stock count on the marketplace to the one on the website. All three are doing the same job: moving information from one system to another by hand. In this article we have gathered business process automation examples for small and mid-sized companies, department by department: what starts the workflow, which steps it goes through, which tool builds it, what it saves and where it breaks.
n8n, Make and Zapier can build most of these workflows; we covered the differences in n8n vs Make vs Zapier. Even so, the first place to look is often the built-in integration of the software you already use; a connection your CRM or e-commerce platform already offers needs less upkeep.
Sales and CRM
From form to CRM to a salesperson alert
- Trigger: a quote or contact form on the website is submitted.
- Steps: the form fields become a contact and a deal in the CRM; if a record with the same email exists, no new one is opened. The enquiry is assigned to a salesperson by product group, who gets an email or WhatsApp alert.
- Tools: the CRM’s own form integration, or Zapier, Make or n8n. According to Zapier’s help centre, triggers are either instant or polling at intervals; an instant trigger means the enquiry lands without waiting for the polling interval.
- What it saves: no form sits in the inbox waiting for someone to open the email and copy it.
- What can go wrong: a field on the form is renamed and the workflow quietly starts creating empty records. On WhatsApp, messages the business starts need an approved template, and according to Meta’s documentation the template review can take up to 24 hours; prepare the template before you go live.
Quote follow-up
- Trigger: a deal in the CRM moves to the “quote sent” stage.
- Steps: if there is no reply within a set number of days, a task is opened for the salesperson, and at a second threshold the manager is told. When the quote is accepted, the deal closes and accounts gets a note to invoice.
- Tools: the CRM’s workflow or reminder sequence feature; otherwise a check that runs every morning in Make or n8n.
- What it saves: quotes no longer depend on a salesperson’s memory.
- What can go wrong: if stages are not kept up to date in the CRM, the workflow sends reminders to the wrong person at the wrong time. For the first few weeks, leave automatic reminders to customers for the salesperson to approve.
Finance and accounts
Matching incoming e-invoices to orders and payments
- Trigger: a new e-invoice arrives from a supplier.
- Steps: the invoice number, tax ID (VKN), amount and date are read and checked against the purchase order or delivery note. If they match, the invoice is marked “ready for approval”, if not, the difference goes on someone’s list; once paid, it is matched to the bank transaction.
- Tools: in Türkiye, GİB (the Revenue Administration) lists three ways to use the e-Fatura system: the GİB e-Fatura Portal, direct integration and a private integrator. If your invoices come through a private integrator or your accounting software, look at their API and ready-made matching features first; the missing step can be built with n8n or Make.
- What it saves: the accountant looks only at the invoices that don’t match, instead of opening every one.
- What can go wrong: two invoices for the same amount arrive from the same supplier and the workflow takes the second for a duplicate. Match on invoice number and tax ID, and let a person take the step that finalises the accounting entry.
Payment reminders
- Trigger: a sales invoice’s due date approaches or passes.
- Steps: a polite email before the due date, a second message with payment details on the day, a task for the account manager once it is overdue. The sequence stops when the payment lands.
- Tools: the reminder feature in your accounting software; otherwise a scheduled n8n or Make workflow and an email or WhatsApp template.
- What it saves: reminders stop depending on one person remembering them; the account manager talks only to customers who are really late.
- What can go wrong: a payment is recorded late and a customer who has paid gets a “your payment is overdue” message. Add a step that checks the payment status again just before sending.
E-commerce
Order and stock sync
- Trigger: an order is placed on the website or a marketplace, or stock is updated in the warehouse.
- Steps: the quantity sold comes off stock on every channel, the order goes to the accounting software or ERP, and purchasing is told when stock drops below a critical level.
- Tools: the platform’s and marketplaces’ own integrations, or marketplace integration software; n8n or Make for custom rules.
- What it saves: fewer out-of-stock items sold and then cancelled, which protects both the customer and your store’s performance.
- What can go wrong: two channels sell the last item at the same moment; however often the sync runs, the risk never drops to zero, so keep a small safety stock on best-sellers. Write down at the start which system is the reference for stock. If you advertise on a marketplace, the same sync protects the budget too; we cover spend that carries on as stock runs out in our guide to advertising on Trendyol.
Shipping notifications
- Trigger: the carrier assigns a tracking number or the shipment’s status changes.
- Steps: the tracking number is written to the order, the customer gets a tracking link, and a task is opened for support when a parcel is “undeliverable”.
- Tools: the platform’s carrier integration often does this itself; for WhatsApp you need an approved template and a provider connected to the WhatsApp Business Platform.
- What it saves: fewer “where is my parcel?” questions.
- What can go wrong: the carrier’s status updates lag and a delivered parcel gets an “on its way” message. Always include a link to the carrier’s own tracking page.
Returns
- Trigger: the customer fills in the returns form.
- Steps: the order number and return window are checked, and if eligible a return code is sent to the customer. When the item reaches the warehouse its condition is marked, a refund instruction is prepared if it is approved, and a resaleable item goes back into stock.
- Tools: the platform’s returns module, the carrier’s return code service, and Make or n8n for the steps in between.
- What it saves: the warehouse, accounts and support see the same record, and the customer doesn’t have to ask for an update.
- What can go wrong: the refund is triggered automatically, but the item arrived damaged. Where money goes out, let the automation prepare the instruction and a person approve the payment.
Marketing
We cover the email flows that come to mind first in marketing (welcome, abandoned basket, post-purchase, win-back) separately in how to set up email automation; here are two other examples.
Lead scoring
- Trigger: someone fills in a form, clicks a link in an email or visits a particular page.
- Steps: each action gets a score (a quote form high, reading a blog post low); someone who passes the threshold is handed to sales; the score of someone who stays inactive drops.
- Tools: the built-in scoring in some CRMs or segments in your email tool; we compared which suits whom in Klaviyo or HubSpot.
- What it saves: the sales team starts with the most interested people.
- What can go wrong: the scores rest on untested assumptions and most of the “hot” leads never buy. After a few months, compare them with the people who became customers and adjust the weights. A high score never makes it acceptable to send marketing messages to someone who hasn’t given consent.
Report automation
- Trigger: a set time every Monday morning.
- Steps: last week’s figures are pulled from the ad platforms, GA4 and the CRM, written to a spreadsheet or dashboard, and a short summary goes to the team.
- Tools: the data connectors of a dashboard such as Data Studio (formerly Looker Studio); a scheduled n8n or Make workflow for sources without one. We covered what belongs in it in the Monday report.
- What it saves: time spent gathering figures goes into interpreting them.
- What can go wrong: a connection’s access expires and the report arrives with missing data, looking error-free. Have the report show the date up to which each source was pulled.
HR and operations
Hiring and approval workflows
- Trigger: a manager opens a request for a new role, or an employee submits a leave or expense request.
- Steps: the request goes through a set approval order; each approver gets an alert and a one-click approve or reject; a reminder goes out if a reply is late. An approved role goes into a job ad draft, approved leave into the team calendar.
- Tools: the approval features of your HR software or office suite; otherwise a form plus Zapier, Make or n8n.
- What it saves: the “who is it waiting with?” question goes away, and every approval is on record.
- What can go wrong: the approver is on leave and the request waits for weeks. Name a deputy for every step.
Collecting documents
- Trigger: a new employee’s start date is confirmed, or documents are requested from a customer.
- Steps: the person gets a list of missing documents and an upload link; documents are named and saved to the right folder; reminders go out for anything missing, and the owner is told when the list is complete.
- Tools: a form tool, cloud storage and Zapier, Make or n8n in between.
- What it saves: no more documents lost in email attachments, and no more “which one was missing?” threads.
- What can go wrong: an ID copy or a medical report lands in a widely shared folder. Under KVKK (Türkiye’s Personal Data Protection Law, Law No. 6698), health data is a special category of personal data; restrict folder access, set a retention period and update your privacy notice.
AI in customer support
Sorting emails and tickets
- Trigger: an email or a new ticket arrives at the support address.
- Steps: a language model sorts the message by topic (order, return, invoice, technical) and urgency and pulls out the order number; the system checks the number exists; the ticket lands in the right queue.
- Tools: the built-in AI features of your support software, or the language model steps in n8n, Make and Zapier.
- What it saves: nobody has to read and route every message, and urgent ones don’t wait at the back of the queue.
- What can go wrong: the model puts messages it is unsure about into a category anyway. Define an “unclear” category and have a person route whatever lands there. Before sending customer data to a model, check the provider’s terms on how it uses data.
An FAQ assistant with a human in the loop
- Trigger: a customer asks a question on the website or on WhatsApp.
- Steps: the assistant drafts an answer only from your own FAQ, returns and shipping texts; if the answer isn’t there, it hands the conversation to an agent; at first, every draft is approved by an agent before it goes out.
- Tools: n8n’s AI agent and vector store nodes, or the built-in assistant in your support software.
- What it saves: answers to recurring questions arrive ready for the agent. The share of drafts approved without changes shows when the assistant can be given more freedom.
- What can go wrong: the assistant invents a return condition that isn’t in the source. Track the correction rate for weeks before removing the approval step, and always send questions about payments, complaints and personal data to a person.
Which one to start with
You don’t need to build all thirteen at once. To choose the first, put three things side by side: how long the task takes each time, how many times a week it is repeated and what an error costs. Time multiplied by frequency gives the weekly load; the cost of an error tells you how carefully the workflow has to be built. We explained the scoring, the one-week inventory and the two-week pilot step by step in where to start with automation.
An example, for illustration only: moving a form entry into the CRM takes a few minutes but happens many times a day, and a mistake reaches the customer; a monthly report takes long but is rare. Here the first candidate is usually the form workflow. Support workflows that need AI are safer to take on once the first rule-based workflow has settled.
Three safeguards common to every example
Put the “what can go wrong” lines side by side and the same three problems come up:
- Error alerts against silent stops. In n8n, a separate error workflow can be linked to a workflow; starting with the Error Trigger node, it runs when the linked workflow fails. In Make, error handlers are added to modules; the Retry handler stores the failed record as an incomplete execution to be retried. In Zapier, every run appears in the Zap history. In every case, the alert should go to a named person.
- A unique key against duplicates. A check on the order, invoice or ticket number asks “has this record been processed already?”.
- Human approval for expensive steps. Money going out, the first messages to customers and deleting records go through a person’s approval until trust is built.
These three layers are part of every workflow we deliver; the details are on our automation and AI service page. If you work with another team, look for the same three layers in the proposal; how to choose an automation agency shows how to ask.
Common questions
What is business process automation?
Business process automation means an event starts a set of steps on its own. For example, a record lands in the CRM when a form is submitted, stock updates on every channel when an order comes in, or a reminder goes out for an invoice that is nearly due. The aim is to remove the job of moving information from one system to another by hand, while leaving the decision to a person at the steps where a mistake is expensive.
Who maintains an automation once it is built?
Every workflow needs a named owner, and the error alert goes to that person. The need for maintenance usually comes not from the workflow itself but from the systems it connects: a field on the form is renamed, or an app’s connection authorisation expires. A half-page document describing what starts the workflow, which steps it goes through and who is responsible means maintenance no longer depends on the person who built it.
What tools are used for business process automation?
The first place to look is the built-in integrations of the CRM, e-commerce platform or accounting software you already use; they need less maintenance. Where the built-in connection falls short, Zapier, Make or n8n come in; we explain when each one fits best in our comparison of n8n, Make and Zapier.
What happens when an automation fails?
Without an error alert, the workflow stops quietly and the problem is often noticed only when a customer complains. That is why every workflow gets an error alert: in n8n, an error workflow that starts with the Error Trigger node; in Make, error handlers attached to modules; in Zapier, the Zap history. In every case, the alert should go to a named person.
Is it safe to use AI in customer support?
With human approval, the risk drops a great deal. The assistant should draft answers only from your own FAQ, returns and shipping texts, and hand the conversation to an agent when the sources hold no answer. A model can put a message in the wrong category or invent a condition that isn’t in the sources, so in the early period every draft is approved by a person. Check the provider’s data use terms before you send it customer data.
The most concrete thing to do after reading this list is to mark which of the thirteen workflows are still done by hand in your company today. Then write next to each one who does it and how many times a week. That half-page list usually shows on its own which automation should come first.
Sources
This article is general information, not legal, financial or technical advice. Platforms, their rules and the law change over time, so some details may have changed since 27 October 2026 or contain small errors. Before you make a decision for your business, check the current official sources or write to us.