Business Process Automation: Where to Start and What to Prioritise

Business process automation is one of those topics where companies most often go wrong right at the start – they buy a tool, deploy it on a randomly chosen process, and wait for results. Real business process automation works the other way round: first you understand exactly what is happening in the company, then you decide where change makes the biggest difference, and only after that do you look for a technical solution. This article is a practical guide to doing it step by step.

The most common reason for failure isn't bad software, but a poorly defined brief. A company decides to automate invoicing or onboarding, but nobody has first described how the process actually runs – who triggers it, what exceptions come up, where waiting times appear, and where data gets copied manually between systems. Without this map, you end up automating chaos, which only speeds it up instead of simplifying it.

The second common mistake is mixing up terms. Business process automation and process digitisation are not synonyms, and companies that confuse them often end up buying the wrong solution – we'll come back to this below.

Before any decision about a tool is made, you need answers to three questions for every candidate process:

• Who performs the process, and how often – one person once a week, or dozens of people every day?
• Where does manual work happen – retyping data, copying between spreadsheets, email and internal systems, manual approvals.
• Where do errors and delays occur – this is usually where the biggest opportunity for improvement lies, not just in time saved but in output quality.

Mapping doesn't need to be complicated – a simple diagram or a table of steps with an owner assigned to each one is enough. It's important to involve the people who actually carry out the process, not just their managers. They are the ones who know the exceptions that never made it into any internal policy. When someone says "automate invoicing" or "automate onboarding" without their input, it's easy to end up automating only part of the process while the exceptions continue to be handled manually.

Once you have mapped several candidates, you need to rank them by priority. We recommend scoring each process against four criteria:

The ideal first candidate is a process with high frequency, clear rules, and data that already exists in digital form. Conversely, processes full of exceptions or dependent on paper documents are better suited to a later phase, once the company already has some experience with automation.

Process digitisation means converting a paper-based or informal process into electronic form – a form instead of paper, a shared document instead of email attachments. Business process automation goes a step further: the system itself carries out steps that a person used to do – routing an approval to the right person, filling in data, triggering the next step without manual intervention.

The order matters. A process that hasn't been digitised generally can't be properly automated – you first need structured data and a clear digital flow, and only then does it make sense to build automation logic on top of it. Companies that skip this order often end up with automation that simply copies the original inefficient process, just faster.

The diagram below illustrates only the general principle – how the share of manual work in a process changes as it moves step by step from a paper-based state through digitisation to full automation. It does not represent measured values or data from any specific project, only a simplified illustration of the relationship between phases.

Across industries, a handful of process types repeatedly meet the criteria from Step 2:

• Invoice processing and approval – clear rules, high frequency, data often already in electronic form.
• Onboarding new employees – a repeated series of steps (access rights, documents, training) that can be standardised well.
• Answers to recurring customer questions – well suited to AI agents, which can handle routine queries and escalate only the exceptions.
• Reporting and data transfer between systems – eliminating manual copying between CRM, accounting, and spreadsheets.

If you're tackling specific areas, take a look at our article on automating invoicing and accounting or on how to automate onboarding for new employees – both describe concrete steps for common business processes.

Once you've chosen the first process, the solution-design phase follows. Here it's worth distinguishing between off-the-shelf tools (SaaS automation platforms) and a custom-built solution – the choice depends on how many exceptions the process contains, how much it needs to integrate with existing internal systems, and what data security requirements the company has.

Implementation cost and time vary from project to project – they depend mainly on the number of systems that need to be connected, the volume of exceptions in the process, security requirements, and the scope of testing before it goes live. That's exactly why it's worth discussing these factors individually rather than estimating them on a blanket basis.

For processes that require their own logic or need to connect several internal systems, it's worth considering custom software development instead of piecing together multiple partial tools that become hard to maintain over time.

Business process automation doesn't start with choosing software, but with mapping how the company actually operates. Choose your first process based on frequency, clarity of rules, potential impact of failure, and data availability – not on what looks the most impressive. Process digitisation is usually a necessary precursor, not an alternative to automation.

If you're not sure which process in your company has the greatest potential and which approach suits it best, it's worth discussing individually before any decision is made about a tool. Get in touch via contact and we'll go through together where automation makes the most sense for your business.