Business process automation is one of those phrases that sounds obvious until somebody asks you to point at an example. Most definitions describe it as using software to run repeatable work without human intervention, which is true and almost entirely useless when you are trying to decide what to build on Monday morning. What helps far more is a list of real processes, in the order teams usually automate them, with enough detail to recognise your own operation in the description.
That is what follows. Twelve examples, grouped by the part of the business they sit in, each written as a trigger, a sequence of steps, and an honest note on what it is worth. None of them require a data science team or a six-figure platform. Almost all of them are built with an orchestration tool sitting between systems you already pay for, which is how the vast majority of useful automation gets done in companies under a few hundred people.
As you read, keep one filter in mind: frequency multiplied by friction. A process that happens forty times a week and irritates somebody every time is worth automating even if each instance only takes four minutes. A process that happens twice a year is usually worth documenting rather than automating, no matter how painful those two occasions are.
What counts as business process automation
A business process becomes a candidate for automation when three things are true. It starts with an identifiable event, such as a form submission, a payment, a date passing, or a status changing in a system. It follows rules that a reasonable person could write down, even if those rules include exceptions. And it produces an outcome somebody else depends on, whether that is a record, a message, a document, or a decision.
Notice that judgement is not disqualifying. Plenty of processes involve a human decision in the middle, and the automation still earns its keep by handling everything either side of that decision. A useful mental model is that you are not removing the person, you are removing the twelve small acts of copying, chasing, checking and remembering that surround the one moment where their judgement matters.
This is also the line between automation and artificial intelligence, which gets blurred constantly in marketing material. Classic automation moves structured data along a defined path. AI handles the unstructured parts: reading a messy email, summarising a call, classifying a document, drafting a reply. Most modern workflows use both, with AI as a step inside an otherwise deterministic process rather than as the process itself.
Sales and marketing examples
1. Inbound lead capture, enrichment and routing. The trigger is a website form or an email to your shared inbox. The workflow creates or updates a CRM record, looks up the company domain to fill in size and industry, scores the enquiry against your qualification criteria, assigns it to the right owner, posts a notification into the team chat, and sends an acknowledgement to the prospect within a minute. This is the single most common first automation for a reason. Response time is one of the strongest predictors of whether an enquiry converts, and manual handling almost always introduces hours of delay overnight and at weekends.
2. Meeting follow-up and CRM hygiene. After a discovery call ends, the recording is transcribed, an AI step summarises it into agreed actions and objections, the summary is written back to the deal record, a follow-up email draft is placed in the salesperson's inbox, and a task is created for the next step. Sales teams rarely fail to follow up because they do not care; they fail because writing notes after five calls in a row is the least appealing part of the day. Automating the admin protects the pipeline data everything else depends on.
3. Proposal and contract generation. When a deal moves to the proposal stage, the workflow pulls the agreed scope and pricing from the CRM, generates a document from a template, routes it for e-signature, and updates the deal when it is signed, triggering the handover to delivery. This removes a class of error that costs real money: proposals sent with last month's pricing, or with another client's name still in the header.
4. Re-engagement of stalled deals. A scheduled job scans for opportunities with no activity in twenty-one days, checks whether the contact has opened anything recently, and either drafts a personalised nudge for the owner to approve or moves the deal to a nurture sequence. The value here is not the email itself but the discipline: no pipeline review meeting can match a job that runs every morning without needing to be remembered.
Finance and admin examples
5. Invoice creation and payment chasing. When a project milestone is marked complete, the workflow generates the invoice in your accounting package, sends it to the billing contact, and then chases on a schedule: a polite reminder three days before the due date, a firmer one on the day, and an escalation to the account owner at fourteen days overdue. Cash collection is one of the few automations whose return you can measure in days of working capital rather than hours of labour, which makes it unusually easy to justify.
6. Supplier invoice and receipt processing. Incoming bills arrive by email as PDFs. A document parsing step extracts the supplier, amount, tax and due date, matches them against the purchase order, routes anything above a threshold for approval, and posts the rest straight into the ledger with the file attached. This is the classic example people mean when they say intelligent document processing, and it is now accurate enough on standard invoice formats to run with light human review rather than full manual entry.
7. Expense and reconciliation checks. A weekly job compares card transactions against submitted receipts, flags anything unmatched, and messages the individual directly rather than routing everything through finance. Small, unglamorous, and it removes one of the most reliably resented recurring tasks in any growing company.
Customer support examples
8. Ticket triage and routing. Every inbound ticket is classified by topic and urgency, tagged, assigned to the right queue, and given an initial acknowledgement that sets a realistic expectation. Where the customer is a known high-value account, the workflow escalates immediately. Triage is a good automation candidate because it is genuinely rule based once you have written the rules, and because the cost of getting it wrong manually, an urgent issue sitting in a general queue overnight, is high.
9. Deflection of repeat questions. An assistant trained on your help documentation, past resolved tickets and product information answers the questions that make up the bulk of volume: order status, password resets, opening hours, delivery timelines, refund policy. Anything it cannot answer confidently is handed to a person with the conversation history attached. Done well this typically absorbs a meaningful share of first-line volume without the experience degrading, because the customer still reaches a human on anything genuinely unusual.
10. After-hours voice handling. Calls outside office hours are answered by a voice agent that identifies the caller, captures the reason for the call, answers common questions, and books a callback in the right person's calendar. For service businesses where a missed call is a lost job, this is often the highest-value automation in the entire list, and it is the one clients are most sceptical about until they hear it working.
HR and people examples
11. Employee onboarding. A signed contract triggers the whole sequence: accounts created in email, chat and the tools relevant to the role, equipment requested, the payroll record opened, documents sent for signature, a first-week schedule built with the right introductions booked, and a checklist shared with the manager. Onboarding is a strong example because it touches many systems, happens often enough to matter, and fails visibly when done manually. Offboarding is the same workflow in reverse, and it carries a security benefit that is easy to undersell: access revoked the same day, every time, with an audit trail.
Operations and fulfilment examples
12. Order, inventory and fulfilment sync. An order is placed, stock is decremented across every sales channel, the fulfilment provider receives the pick request, the customer gets a confirmation and later a tracking link, and low-stock thresholds trigger a purchase suggestion for the buyer. Retail and e-commerce operations tend to accumulate manual reconciliation between channels, and that reconciliation is precisely the kind of work that scales badly and errs often.
Alongside it sits the reporting workflow that almost every company eventually builds: a scheduled job that pulls figures from sales, finance and support into one place and distributes a short management summary on Monday morning. It is not glamorous, but it replaces the recurring half day somebody spends copying numbers between spreadsheets, and it removes the version-control arguments that follow.
What the successful examples have in common
Look across the twelve and a pattern emerges. Every one of them starts from a clear event rather than from a person remembering. Every one of them writes its result back into a system of record rather than ending in somebody's inbox. And every one of them has a defined behaviour for failure, whether that is a notification, a retry, or a handoff to a person, because the automations that erode trust are the ones that fail silently.
They also share a scope discipline. None of these examples tries to automate a department. Each takes one path through one process and does it reliably. Teams that attempt the grand version first, a single workflow covering every variant of enquiry across three business lines, generally spend two months building something nobody trusts. Teams that ship the narrow version in a fortnight usually broaden it themselves within a quarter, because success creates appetite in a way that a roadmap never does.
The last commonality is measurement. Before building, the team wrote down the current numbers: how many enquiries a week, how long the average response takes, how many invoices go past thirty days, how many tickets are repeat questions. Without those baselines the automation still works, but nobody can prove it, and unproven automation is the first thing cut when budgets tighten.
How to pick your first one
Score your candidate processes on four things and the answer is usually obvious. How often does it run, since frequency is where the return comes from. How rule based is it, since a process nobody can describe cannot be built. How many systems does it touch, because two or three is a healthy first project and seven is not. And who feels the pain, because an automation with an internal champion gets used and one without a champion quietly gets bypassed.
In practice the winner for most small and mid-sized companies is the first example on this list. Lead capture and follow-up is frequent, rule based, touches few systems, and its owner, whoever runs sales, is highly motivated to see it work. Finance chasing is a close second when cash collection is the visible constraint. The specific choice matters less than the principle: start where the volume is, and make the first build small enough that it is running within a fortnight.
Frequently asked questions
Where to start
Take the example above that most resembles something you do manually today, and spend an hour writing it out as a trigger, the steps, the systems involved, and what should happen when it goes wrong. That page is a build brief. It is also the fastest way to discover whether the process is genuinely ready to automate or whether it needs tidying up first, which is a useful thing to learn before anyone opens a tool.