Process optimisation: map the process first, then the software that fits

Short answer. Process optimisation is improving the way work flows through your business, with fewer steps and fewer errors. The order decides whether it works. First you map the process and remove what adds no value. Only then do you adapt software, connect systems or build something new.

I spent years improving processes before I started building software myself. At organisations such as UWV, Centric and G4S, with a master's in Business Process Management. After that I ran my own e-commerce operation across several countries and channels. In all those years the software was rarely the first problem. The first problem was almost always a process that once made sense and then kept growing.

Process optimisation and software belong together, but in a fixed order. Start with the tool and you automate your mess. Only draw the process and never touch the software, and you are left with a nice report and the same manual work.

What is process optimisation?

Process optimisation is the systematic improvement of a business process so that it costs less time and money and produces fewer errors. You look at the actual flow of work, from request to delivery, and remove, simplify or automate the steps that add nothing.

An example I see in almost every growing operation. An order arrives by email. Someone copies it into an Excel list. A colleague then enters it into Exact Online. At the end of the week someone else checks whether the list and Exact still match. Four actions for one piece of data. Do the maths: if each action takes three minutes and sixty orders come in per day, twelve hours a day go into one piece of data that was already in the email. Automation makes one of those actions faster. Optimisation asks why there are four. Work out what it costs you with the calculator: what does manual re-typing cost you per year?.

Why the process comes before the software

Automation does not make a fragmented process whole. It makes the separate pieces faster. In Deloitte's global survey on intelligent automation, process fragmentation and a lack of IT readiness topped the list of reasons companies could not scale their automation, just as in the years before (Deloitte, 2020). Automate one step of such a process and you get a solution that works for that step and nowhere else.

Meanwhile adoption is moving fast. According to Statistics Netherlands (CBS), 45 percent of Dutch companies with 50 to 250 employees used AI technology in 2025, up from 20 percent in 2023 (CBS, 2025). Those are exactly the companies outgrowing their own processes. More figures, broken down by company size, are in how many Dutch companies use AI and automation. The faster you build, the more expensive it is to do so on top of a process nobody has looked at in years.

How I map a process

This is the order I follow on every project. It works for an order process in a web shop and just as well for invoicing at an installation company.

1. Start with the goal. A dashboard is a means. The goal is usually overview, numbers you can trust or less manual work, and that decides what you measure.
2. Map the actual flow. Who does what, in which system, and where does the work wait for someone? Draw how it really runs. The handbook is usually two reorganisations old.
3. Mark the waste: re-typing between systems, double checks, waiting for approval, searching for the latest version of a file. Lean calls these steps that add no value for the customer (Lean Enterprise Institute).
4. Simplify or remove. Some steps can be done smarter, some can go. That is cheaper than any software.
5. Only then choose the solution: configure the package differently, connect systems or build something new. How I make that choice is in off-the-shelf, integrate or custom.
6. Measure again. Is the manual work really gone, or has it moved to someone else?

Which method fits your process?

There are established methods for process optimisation. They are useful as long as you treat them as tools.

Method	What it is about	Good fit for
Lean	Removing waste, letting work flow	Lead time, waiting time, unnecessary steps
Six Sigma (DMAIC)	Reducing variation and errors with data	Recurring errors, quality problems
BPMN	Recording a process unambiguously in a diagram	Agreements between departments and with a developer
Value stream mapping	Making the whole flow from request to delivery visible	Seeing where work actually gets stuck
PDCA	Testing small improvements in short cycles	Continuing to improve after the first round

How to use these methods in a mid-sized company without turning it into a year-long project is covered in process optimisation models: Lean, Six Sigma, BPMN and when to use which.

From mapping to software

After mapping, there is almost always work left that a system has to do. There are four routes, from cheap to expensive: adapt your process to the package, extend or reconfigure the package, connect systems, or build custom software. The honest answer is usually a combination. Bookkeeping stays in a standard package, the web shop gets connected, and the exceptions no package knows get a small system of their own.

AI is a tool for one kind of step: the one where someone now reads messy information and transfers it. For anything that has to be exactly right, it is the wrong choice. Where that line sits is in automating business processes with AI.

Where it goes wrong in practice

The same process problems return in every industry. Most often I see data moving by hand from one system to another (re-typing between systems) and a patchwork of tools where nobody knows which source is leading. Right behind those: the logic of a core process living in a spreadsheet only one colleague understands (dependent on Excel and one person), and steering on yesterday's numbers because margin and reporting don't add up.

They are four faces of the same thing. The process grew, the systems were added, and nobody looked at the whole again.