What business process automation means
Business process automation means letting software carry out the repeatable steps of how your company works: moving data, creating tasks, sending documents, checking conditions, reminding people and chasing things that are late.
It covers a wide range. At one end, a rule that sends an email when a form is submitted. At the other, a system where a finished job automatically becomes an invoice, the invoice is sent to accounting, the payment is matched from the bank, and a reminder goes out if it's late, with nobody touching it unless something is wrong.
The goal isn't automation for its own sake. It's fewer hours spent copying, checking and chasing, and fewer mistakes that cost money. If an automation doesn't do one of those, it's decoration.
Which processes are worth automating
The best candidates share a few traits. Score your processes against them before building anything.
- Frequent: it happens daily or weekly, not twice a year.
- Rule-based: the right action can be described, even if the description has a few branches.
- Painful when missed: forgotten follow-ups, unbilled work, expired documents, late payments.
- Involves re-typing: the same data entered into two or more places.
- Crosses a hand-off: work moves from one role to another and tends to get lost in between.
- Stable enough: the process won't be redesigned next month.
Common automations by area
Our article on business process automation examples goes through specific cases in detail. Here's the map of where automations usually pay off in a company of 20 to 300 people.
- Sales: enquiries captured from every channel, leads assigned by rule, follow-ups scheduled, quiet deals flagged, offers generated from templates.
- Operations: won deals turned into jobs, jobs scheduled against availability, checklists and reports required before a job can close, stock reserved and reordered.
- Finance: invoices created from completed work, sent to accounting, payments matched from the bank, reminders sent on a schedule, overdue lists for the owner.
- People: onboarding checklists, shift plans published, leave requests routed, expiring certificates flagged, hours prepared for payroll.
- Customers: confirmations, reminders, status updates, satisfaction checks and portal access without anyone sending emails by hand.
- Reporting: a daily or weekly summary of the numbers that matter, delivered without anyone building it.
Three ways to automate, compared
There are three main approaches. Most companies end up using more than one, and that's fine if you know which job each one does.
- Glue tools (Zapier, Make and similar). Speed: fast to start. Cost: low monthly fees that grow with volume. Strength: connecting SaaS tools quickly. Weak spot: dozens of small automations with no owner, silent failures, and logic scattered across accounts nobody reviews.
- Automation inside each SaaS tool. Speed: fast. Cost: usually included or an add-on. Strength: works well within one tool. Weak spot: stops at the tool's edge, so cross-department processes stay manual.
- Automation inside a company system. Speed: needs the system first. Cost: part of the build. Strength: rules run on one shared database, see the whole process and fail visibly. Weak spot: bigger upfront decision.
- Robotic process automation (screen bots). Speed: medium. Cost: licences plus maintenance. Strength: automating old software that has no API. Weak spot: breaks when screens change; best treated as a bridge, not a foundation.
Glue automation vs system automation
This distinction decides whether your automation holds up in two years.
Glue automation connects tools that were never designed to work together. It's great for a quick win or a temporary bridge. But each automation is a small piece of logic that lives outside your main systems, often built by one person, rarely documented. When a field changes in one tool, something breaks quietly, and you find out when a client complains.
System automation lives where the data lives. A rule like 'when a job is marked done and has signed paperwork, create the invoice' runs in the same system that holds the job, the paperwork and the invoice. It can't lose sync with itself, it logs what it did, and it's visible to anyone with access.
A sensible pattern: use glue tools to test whether an automation is worth having, then move the ones that matter into the core system.
How to automate a process, step by step
The order matters. Automating a broken process just breaks it faster.
- Map the process as it really runs, including the workarounds and the person who 'just knows'.
- Remove steps that exist only because the tools don't talk to each other.
- Decide the single source of truth for each piece of data.
- Write the rules in plain language: when this happens, and these conditions are true, do that.
- Define the exceptions: what the automation should not handle and who gets it instead.
- Build it, run it alongside the manual process for a short period, and compare results.
- Switch off the manual process and keep a log of every automated action.
- Review monthly for the first quarter: what failed, what was overridden, what should be added.
Where AI fits in automation
Classic automation needs clear inputs: a status change, a date, a number. Much of the work in a company arrives messy: an email from a client, a scanned delivery note, a voicemail, a photo of a damaged part.
AI closes that gap. It reads the messy input and turns it into structured data the rules can use. An email becomes a new job with the right client, address and urgency. A supplier invoice becomes lines ready for approval. A call becomes a lead with notes. AI can also draft the human-facing parts, like a polite second reminder or a status update to a client.
We treat these as AI staff: a receptionist that answers calls, a caller that works through lists, a clerk that files, invoices and chases payments. They work on the same data and permissions as your people, and anything uncertain goes to a human for approval.
Where automation goes wrong
Most failed automation projects fail for boring reasons.
- Silent failures: an automation stops working and nobody notices for weeks. Every automation needs logging and an alert when it fails.
- No owner: automations built by someone who left, in an account nobody can access.
- Over-automation: automating a decision that needs judgement, then spending more time fixing its mistakes than it saved.
- Automating around bad data: rules that depend on fields people don't fill in reliably.
- Too many small tools: every department builds its own, and the company ends up with a hidden system nobody designed.
Owning and documenting your automations
Every automation is a small promise: when this happens, that will happen. Over time a company accumulates dozens of these promises, and someone has to know they exist.
Keep a simple register, even if it's a single page: what each automation does, what triggers it, what systems it touches, who owns it and what happens when it fails. Automations built into a company system can show this directly, with a log of every action taken. Automations scattered across glue tools usually can't, which is one more reason to move the important ones into the core.
Make sure the accounts that run your automations belong to the company, not to one employee's personal login. It sounds obvious. It's also one of the most common reasons an automation suddenly stops when someone leaves.
Bringing your team along
People worry that automation means they're next. In practice, in companies of this size, it removes the parts of the job nobody likes: re-typing, checking whether something was sent, chasing the same late payer for the fourth time.
Say that clearly and early, and involve the people who do the work in designing the rules. They know the exceptions. They know which client always pays late but always pays, and which supplier's invoices are always wrong. An automation built without them will miss those details and they'll quietly work around it.
After launch, give them an easy way to flag when the automation got something wrong. Those reports are the fastest way to improve the rules, and they turn the team into owners of the system instead of its victims.
How to size the payback
Before building, estimate the value with numbers you already have. For each candidate process, note how often it happens, how long it takes each time, who does it and what that person's time costs. Add the cost of mistakes from the last year: unbilled work, late payments, lost leads, penalties.
Then estimate what share the automation removes. Be conservative. If the payback is clear on a conservative estimate, build it. If it only works on an optimistic one, wait or fix the process first.
Don't forget the second-order value. When invoices go out the same day as the work is done, cash arrives earlier. When follow-ups never slip, conversion improves. These are harder to predict, so don't count on them, but they often end up bigger than the hours saved.
Measuring automation after launch
An automation that nobody measures tends to drift. It keeps running, but the process around it changes, and after a year it's doing something slightly different from what anyone intended.
Track a few things for each important automation: how many times it ran, how many times it failed or was overridden by a person, and the business number it was meant to move. For invoicing, that might be days between job completion and invoice sent, and days until payment. For lead follow-up, time to first response and the share of leads with a next step scheduled.
Look at these monthly for the first quarter, then quarterly. When overrides pile up, the rule is wrong or the process has changed. When failures pile up, something upstream has changed, usually a field, a format or a connected tool. Both are cheap to fix early and expensive to discover late.
What it costs
Glue tools cost a monthly fee that grows with the number of tasks run. Individual automations built by freelancers are often priced per workflow. Both are fine for small, isolated needs.
When automation is part of a company system, it's priced with the system. At Company Maxxing, Core at €18,000 gives you one company on one system with up to five roles, data migration and three months of support, and the hand-offs are built into the system from day one. Maxxed at €48,000 adds automations that chase invoices, follow up leads and fill gaps, the AI layer, up to four integrations, a client portal and six months of support. Empire starts at €120k for groups of companies. Extra integrations are €4,500 each.
After the support period, Maxxing Care at €2,900 per month covers changes and new features, which is where automation lives long term. Processes change, and the rules need to change with them.
Getting started
Pick one process that crosses at least two roles, happens every week and costs money when it slips. Invoicing after completed work is the classic choice. Map it, fix it, then automate it.
If you'd rather do it with us, apply and show us how the process runs today. We map every hand-off, give you a fixed price, and you get a first working version on your own data within 24 hours of kickoff.