AI automation examples by department
Updated:
The AI automation examples that work in companies share one shape. Something messy arrives, such as an email, a PDF, a form or a call recording. AI reads it and turns it into clean data or a draft. Ordinary automation then moves the result into the system where it belongs, and a person approves wherever a mistake would cost money. Typical examples are reading supplier invoices into accounting, sorting customer emails and drafting the replies, putting call notes into the CRM, and comparing order confirmations with purchase orders. Below are examples for each department, what the AI does in each and what stays with a person.
AI automation or plain workflow automation?
Many workflow automation examples need no AI at all. If the input always arrives in the same format, a fixed rule is cheaper and more reliable. AI earns its place where the input is free text or every case looks a little different.
| Workflow automation (rules) | AI automation (an AI step in the flow) | |
|---|---|---|
| Input | Structured: a webshop order, a form field, a row in a spreadsheet | Unstructured: an email, a scanned invoice, a call, a photo of a receipt |
| Example | A paid order creates an invoice and a delivery note | An order sent as a PDF attachment is read and entered into the system |
| Mistakes | Predictable: it fails the same way every time until someone fixes the rule | Occasional and varied, so it needs checks and a person for unclear cases |
| Running cost | The automation tool or hosting | The same, plus AI usage fees per document or message |
Most useful systems combine both: rules do everything they can, and AI handles the step a rule can’t. Every example below works that way. When the AI has to choose the steps itself, not just do one of them, you need an agent instead; these AI agent examples show the difference. The platform that runs the workflow matters less than it seems; n8n vs Make vs Zapier compares three common ones on cost and where your data goes.
AI in finance and accounting: examples
- Supplier invoices from the mailbox to the accounting system. AI reads the PDF and pulls out the supplier, invoice number, amounts, VAT and due date, then creates the entry. A person approves invoices over a set amount or from a new supplier.
- Matching payments to open invoices. When a bank transfer has a missing or wrong reference, AI suggests which invoice it pays, based on the amount, the payer’s name and the text. Someone confirms the suggestion.
- Expense receipts. Staff photograph a receipt, and AI reads the amount, date and merchant and proposes a category.
- Month-end commentary. AI drafts the explanation of the largest changes against last month from the figures. The finance lead edits it before it goes anywhere.
Numbers need a second check that isn’t AI. A simple rule, such as “the line items must add up to the total”, catches most reading errors before they reach the books.
AI in HR: examples
- Answering staff questions. Leave rules, expenses, benefits: AI answers from your handbook and policies, and links to the paragraph it used. Questions it can’t answer go to HR.
- Onboarding. When a contract is signed, accounts, equipment requests and first-week tasks are created automatically. AI drafts the welcome email and the first-week plan from the role.
- Job ads and interview scheduling. AI drafts a job ad from a short brief, and reads candidates’ replies to propose interview times that fit the calendars.
- Survey and exit interview answers. AI groups free-text answers by theme so HR reads a summary instead of two hundred comments.
Be careful with anything that judges people. The EU
AI Act lists AI used to filter job applications or evaluate candidates as high-risk (Annex III), which brings extra obligations. Under GDPR, a person shouldn’t be rejected by an automated decision alone (Article 22; see GDPR and AI). In practice, let AI organise and summarise, and let a person decide. This is general information, not legal advice.
AI in sales: examples
- Qualifying inbound leads. A form or email comes in. AI reads it, checks the company’s website, compares it with the criteria you set, creates the CRM record and sends it to the right salesperson.
- Call notes into the CRM. AI turns the call transcript into a summary, next steps and updated fields. The salesperson reads it and corrects it in a minute instead of typing it from memory.
- Quote drafts from an email request. AI works out what the customer asked for, matches it to your price list and prepares a quote. A person checks it and sends it.
- Follow-ups with context. When a deal goes quiet, AI drafts a follow-up that refers to the last conversation, for the salesperson to send or bin.
AI in marketing: examples
- Product descriptions and translations. AI writes first drafts from the product data, in every language you sell in. Someone who knows the product checks them before they go live.
- Reusing what you already have. A webinar, case study or long article becomes a newsletter draft and a set of social posts.
- Sorting reviews and feedback. AI tags every review and survey answer by topic and tone, so complaints about delivery and praise for a product show up as counts, not anecdotes.
- The weekly campaign report. Figures are collected from the ad platforms and your analytics, and AI writes three paragraphs on what changed.
Anything published under your name needs a person to approve it. AI writes confidently even when it is wrong about a product detail.
AI in customer service: examples
- Sorting incoming messages. AI reads each email or chat, tags its topic, language and urgency, and sends it to the right queue.
- Drafting replies. AI looks up the order and your policies and drafts the reply. The agent edits and sends it.
- Resolving routine requests end to end. An AI agent tracks the parcel, changes the delivery address or starts the return itself, and hands anything unclear to a person. LeoAI, which we built for webshops, has handled 8,400+ conversations this way, 94% of them with no person needed (source:
leoai.hu). - Handover summaries. When a case reaches a person, AI summarises what the customer wants and what has been tried, so nobody asks twice.
More on what this takes is on the AI customer service page.
AI in operations and procurement: examples
- Orders that arrive by email or PDF. AI reads the customer’s order and enters it into the ERP or webshop. Anything it can’t match, such as an unknown product code, goes to a person.
- Checking order confirmations. AI compares the supplier’s confirmation with your purchase order and flags differences in price, quantity or delivery date.
- Comparing supplier quotes. Quotes in different formats are read into one table, so the comparison takes minutes.
- Reports across systems. Stock, orders and deliveries are collected every week, and AI writes a short summary of what needs attention.
How do you pick the first one?
Choose the example that scores well on all four:
- It happens often. Daily or weekly, not twice a year.
- The input arrives digitally. Email, PDF, form or chat, not paper in a drawer.
- A person can check the result quickly. Approving takes seconds, not a second round of the same work.
- A mistake is caught before it costs. A wrong draft is cheap; a wrong payment is not.
Start with the AI preparing and a person approving, and give it more freedom only where the numbers show it is right. The AI readiness checklist helps you score your own processes, and how to implement AI in a business covers the steps from pilot to rollout.
What vSites does
We build AI automation like the examples above: we connect the systems you already use and add AI only to the step a rule can’t handle. The price is fixed in writing before we start, and the developer you talk to is the one who builds it. If you aren’t sure which process to start with, AI consulting answers that first, and if a plain rule or no automation is the better answer, we say so.