Business process automation: definition, examples and method for an SME

Process automation means having tools carry out the repetitive steps of a business process, such as data entry, passing information on, chasing or approving, so that people only step in where judgment is needed. In an SME, the gain comes mostly from the handoffs between people and between tools.

On the ground, most companies are starting from a long way back. In France, only 5% of very small and small-to-medium businesses say they use AI to automate tasks (France Num 2025 barometer, 11,021 companies surveyed in spring 2025). In Wallonia, the Digital Wallonia 2025 barometer gives companies 29/100 on its "processes" axis, a score that "reflects a lack of automation and process transformation, with digital remaining mainly a support for existing activities" (our translation).

Search for "business process automation" and you land on SAP, IBM, Blue Prism, Ricoh, Atlassian. Software vendors who define automation by what they sell, for companies with several thousand employees. Useful for understanding their catalogue. Far less useful for knowing what to do in a 40-person SME that runs on a CRM, an invoicing tool and a dozen or so Excel files.

I have no software to sell you, so let's talk about your SME: what a process is, six examples you find everywhere, and how to choose the first one to automate. The topics that deserve more room have their own article, linked along the way.

Process automation: the definition, without the vendor jargon

Automating a process means removing from a chain of work the steps a machine does as well as a human, and automatically connecting the ones that remain. The key word is "chain". A process has a start (a request, an order, a new hire) and an end (a customer served, an invoice paid, an employee up and running). In between, several people and several pieces of software hand the work over to each other.

Process, task, workflow: the difference that changes everything

• A task is a single action done by one person: entering an invoice, sending a reminder email, copying a row from Excel into the CRM.
• A process is the complete chain that produces a result for a customer, internal or external. The "quote → order → invoice → payment" cycle is a process. It contains dozens of tasks.
• A workflow is the way that process is tooled: who receives what, in what order, with what approval. It is the process as wired into the software.

You can automate ten tasks without the process moving a single day faster. If the invoice is generated in two seconds but then waits three days for approval from someone who is travelling, the customer still pays late.

BPA, RPA, iPaaS, AI: four ways to automate

You will come across these acronyms in every brochure. The main thing to remember is that each one names a technique, and a single process often mixes several.

• BPA (Business Process Automation): the generic term, the overall approach. It is the subject of this article.
• RPA (Robotic Process Automation): a software robot that clicks in an interface instead of a human. Useful when an old piece of software has no API. Fragile as soon as the screen changes. The details are in RPA for SMEs.
• iPaaS (integration platforms, such as Make, n8n or Zapier): connectors that move data between your tools according to rules. They are the basic tool of automation in an SME.
• AI: it handles what rules cannot, such as reading a free-form email, extracting data from a PDF or classifying a request. More on that below.

And before all of that, there is an option everyone forgets: the native features of your software. Your invoicing tool can probably already send reminders on its own. You still have to have opened the menu.

Why you automate a process, not a task

The time lost in an SME is rarely found inside the tasks. It is found between them. A signed quote arrives by email, someone forwards it, someone else re-keys it into the ERP, a third person waits for stock to be confirmed, and the invoice goes out a week after delivery. Each task on its own is quick. The process is slow.

Four losses come up almost every time at the handoffs:

• Re-keying: the same information typed two or three times into different tools, with a possible mistake each time.
• Waiting: the file is complete, but it sits in the inbox of someone who does not know they need to act.
• Manual approval: a sign-off that could follow a rule ("under €500, approved automatically") but goes through a human out of habit.
• Chasing: someone has to remember to follow up, and nobody remembers at the right time.

That is also why automating an isolated task is often disappointing. If the task sits in the middle of a broken process, you speed up one piece and the problem moves to the next. I go into this in detail in the tasks to automate first, in the section "the mistake to avoid". Look at the process end to end first, then choose the tasks.

Looking at the process end to end has a name: process mapping. In an SME, you do not need modelling software. A wall, sticky notes, the people who do the work, and two hours. For each step, write down who does it, in which tool, how long it takes, and above all how long the file waits before the next step. The waiting is almost always longer than the work itself, and it is usually the surprise of the workshop.

6 examples of business processes you can automate in an SME

These six come up in almost every B2B services and manufacturing SME. For each one, three questions: where the time is lost, what you can hand over to a tool, and what you should leave to a human.

1. From quote to invoice

The signed quote comes back by email or as a PDF. Someone re-keys the lines into the management tool, then the invoice waits for delivery to be confirmed. You invoice late, so you get paid late.

Most e-signature tools can trigger an action on signature: create the order, notify logistics. The delivery confirmation can in turn generate the invoice and send it to the right contact. Off-grid discounts and disputes stay with a human.

If you only tackled one, it would be this one. It hits your cash flow directly.

2. Chasing unpaid invoices

A reminder takes a few minutes per invoice. The problem is that nobody sends it at the right time, and the tone depends on the mood of the day. Payment delays grow without anyone having decided it.

A fixed schedule solves most of it: a reminder on the due date, then at day 8, then at day 15, with messages written once and for all and the payment link in every email. The day a strategic customer has a real problem, it is you or your accounting lead who calls them, not the automation.

3. Onboarding a new employee

Contract, IT accounts, equipment, access to tools, training schedule. Five people are involved and none of them has the full list. The new hire spends their first week waiting for access.

Here, a single entry can trigger everything else. Signing the contract creates the record in the HR tool, sends the account requests to IT and schedules the welcome meetings. It is the textbook case of HR process automation, which also covers leave and expense reports. The welcome itself, on the other hand, happens in person.

4. Entering and matching supplier invoices

Invoices arrive by email, as PDFs, sometimes still on paper. They have to be entered, matched against the purchase order, approved and passed on to the accountant. Four manual steps, four chances to get it wrong.

An invoice-reading tool extracts the supplier, the amount and the date. Matching against the order happens on its own, approval goes to the right person based on an amount threshold, and the document lands in accounting. That is most of what accounting process automation means for an SME. When the invoice does not match the order, the tool flags it and someone decides.

5. Qualifying and routing inbound requests

Website form, generic email address, phone, LinkedIn: requests come in through several channels and are sorted by hand. Some sit for three days before reaching the right salesperson, and a request that waits goes cold.

You centralise everything in the CRM, enrich the record (company, size, sector), assign it according to two or three simple rules and send an immediate acknowledgement. The first real conversation is still with a salesperson.

6. Monthly reporting

At the end of every month, someone exports figures from three tools, pastes them into an Excel file, repairs the broken formulas and sends the file. One to two days of work for a document half the recipients skim.

A dashboard connected directly to the source tools makes this exercise disappear. The time you get back can finally go into understanding why the numbers move.

Which process to automate first: the grid

The first process to automate is the one that comes up often, follows a stable rule, relies on reliable data and is expensive when it goes wrong. Often, it is not the one that annoys the owner the most.

To decide between several candidates, score each one on these five criteria. No tool is needed, just a week of observation:

• Frequency: how many times the process is triggered per month. A monthly process pays back twelve times less than a daily one.
• Total time per occurrence: actual work plus waiting, from start to finish. It is often the waiting that dominates.
• Cost of a mistake: a wrong invoice, a forgotten customer, an employee without access. Estimate what one mistake costs, and how many happen per month.
• Stability of the rule: does the process run the same way 9 times out of 10? If it changes every time, it is not ready.
• Quality of the input data: does the information you need exist, in the right place, in a usable format?

The gain calculation is the same everywhere: time per occurrence multiplied by frequency. Take a process that requires 20 minutes of work and comes up 60 times a month. That is 20 hours a month. If automation removes three quarters of it, you get back 15 hours, roughly two working days a month for a single process.

The last two criteria are knockouts. An unstable process or unreliable data, and automation amplifies the mess instead of reducing it. On our automation agency page, we frame them as three questions to ask before building anything, the first being "should this process even exist?". Sometimes the right answer is to remove a step, not to automate it.

The 3 mistakes that sink a process automation

When an automation fails in an SME, the cause is rarely technical. It almost always lies before the project, or after go-live.

Buying the tool before looking at the process

This is the most common one. An automation tool shines in a demo, the licences are bought, and then you look for something to automate with it. You end up automating what the tool does well, not what costs the company. The reverse order is detailed in where to start to automate an SME: the process first, the data next, the tool last.

Automating the exception instead of the usual case

When you map a process, the exceptions take over the whole discussion. "Yes, but when the customer has two delivery addresses…" If you try to automate every exception, the project triples in size and never ships. Automate the 80% of cases that look alike, and let the exceptions go to a human, with a clear alert.

Not appointing anyone to look after the automated process

An automated process breaks one day. A piece of software updates its API, a supplier changes the format of its invoices, a CRM field is renamed. If nobody is responsible for the process, the failure goes unnoticed for weeks, and you discover at closing time that 40 invoices were never sent.

Every automated process needs an owner on the business side: someone who knows what the process should produce, who receives the error alerts and who has the documentation to understand what was built. Vendors never talk about it, and I understand why, it does not sell. Yet it is what decides whether your automation lasts two months or five years.

What about AI?

AI widens the scope of what can be automated, but you always need to know where in the process you put it. Classic automation follows rules: if this field has this value, then take this action. It stalls as soon as information arrives in free form, like an email written by a customer, a scanned PDF or a request phrased in ten different ways.

That is where AI helps. It reads the email and extracts the intent, it pulls the amounts out of a badly scanned invoice, it classifies an inbound request. It becomes one step in the process, surrounded by rules and by human approval wherever the stakes justify it.

SMEs are getting there, at very different speeds depending on the country. In 2025, 28.8% of Belgian companies with 10 to 49 employees use at least one AI technology, compared with 15.0% in France and 17.0% on average in the EU (Eurostat, 2025). And among European companies that considered AI without taking the step, 70.9% cite a lack of expertise (Eurostat, December 2025). What holds companies back is know-how, far more than technology.

And if the process is fuzzy or the data poorly kept, AI gets things wrong with great confidence. The scoping logic we apply to AI projects is described in the C.A.R.E. method. It starts from the same principle as this article: structure before you automate.

Where to start, in practice

Choose a single process using the grid above. Map it with the people who do it, noting the working time and the waiting time of each step. Measure the starting situation for two weeks. Look first at what your current software can already do, and only then at what needs to be added. Automate the usual case, appoint an owner, measure again.

My take: one process handled properly teaches your team more than ten patched-together automations, and it gives you the numbers to choose the next one.

If you want us to look at your processes together and identify the one that will pay back the most, that is exactly what we do in a diagnostic. Book a 30-minute discovery call. We will tell you honestly what is worth automating, and what is not.