Process optimisation for SMEs: how to find where your time goes

Almost every company with twenty to a hundred people has a process everyone knows and nobody enjoys. An order arrives by email, someone retypes it into the ERP, a colleague checks it, and if anything is unclear, emails go back and forth. It works. But it costs hours every week, and those hours grow with your revenue.

Process optimisation is about that kind of work. Not a new system or an AI project, but a simple question: what do we do now, what does it cost, and what could be better? This article shows how to tackle that without a months-long consulting programme.

What is process optimisation?

Process optimisation means improving a recurring workflow so it takes less time, produces fewer errors or finishes sooner. You look at the steps as they are now, remove what adds no value, make the rest unambiguous, and only then automate what is left.

The difference with automation is the order. If you automate straight away, you lock a messy process into software. The bad part just runs faster from then on.

Why process optimisation works differently in an SME

A large company has a process team, process documentation and people with time to maintain it. In an SME, knowledge of the process sits in a few heads, and those heads are busy. That has three consequences.

• The process isn't written down anywhere. Everyone does it slightly differently, and you only notice when someone is off sick.
• The team is the link between systems. Data moves from email to Excel to the accounting package by hand.
• There is no time for a months-long programme. An improvement has to pay off within weeks, or it disappears under the day job.

So an approach for SMEs is small and concrete. One process, with the people who do the work, and a figure you can check.

A step-by-step plan for process optimisation

Use this plan for one process at a time. Allow a few half-days for the first three steps, spread over a week.

1. Pick one process, not a department

"Finance" is not a process. "A supplier invoice from arrival to payment" is. Choose a workflow with a clear start and end that happens often and that people complain about. Complaints are a good guide: they point to work that is done twice or gets stuck.

2. Map how it actually runs

Sit down with the people who do the work, not their manager. Draw every step: who does what, in which system, and what happens when something doesn't add up. That last part matters. Most of the time isn't spent on the normal route but on the exceptions: the invoice without an order number, the customer who exists twice, the order that came in by phone.

A whiteboard or a sheet of A3 is fine. You don't need modelling software to see where work gets stuck.

3. Work out what it costs now

Without a figure, every improvement is an opinion. Add up three things: the hours spent, times the hourly cost including employer costs; the time spent fixing errors; and what gets left undone, such as quotes that go out too late. The manual work calculator gives you a first estimate in a few minutes.

Write your assumptions down too. When someone later asks where the figure comes from, you want to be able to show every step.

4. Cut before you automate

Go through each step and keep asking the same question. What happens if we stop doing this? Often a check turns out to be done twice, a report is read by no one, or an approval is unnecessary below a certain amount. Cutting costs nothing and saves time straight away.

5. Make the exceptions unambiguous

Whatever is left, make predictable. Agree what happens for each known exception, and who decides when something doesn't fit. A process with fixed rules for its exceptions is easy to automate. A process that runs on gut feeling is not.

6. Automate what is left

Only now do you choose your tools. Often it is a setting in software you already have, or an integration between two systems. Where the input is free and messy, such as emails, PDFs or phone notes, AI can take over the reading and sorting. When to choose which, see AI agent or regular automation. Which process to tackle first is covered in which business processes suit AI. A step-by-step plan for the automation itself is in automating business processes.

7. Measure again and appoint an owner

After a few weeks, measure the same three things as in step 3. Appoint one person who looks after the process and decides what changes. Without an owner, the old work slowly creeps back in.

Lean, Six Sigma or BPMN: which model do you need?

Search for process optimisation and you soon run into models. The three best known:

• Lean: comes from manufacturing and is about removing waste, such as waiting, duplicate work and unnecessary steps. The thinking translates well to office work.
• Six Sigma: aims for less variation and fewer errors, with a fixed cycle of measuring and analysing. Strong for high-volume processes with measurable quality.
• BPMN: not an improvement method but a notation for recording processes unambiguously. Useful when a process crosses many departments or systems.

For an organisation of ten to a hundred and fifty people, the lesson of Lean is usually enough: see the waste, remove it, and measure whether things improved. You don't need a certification or a methodology handbook for that. The plan above is essentially Lean on a small scale.

Examples of process optimisation

Three patterns we often see in SMEs. The details differ per company, the pattern doesn't.

• Orders from the inbox: orders arrive as emails or PDFs and get retyped. First reduce the channels and agree fixed fields, then automate the input with a human check on anything unusual.
• Quotes that sit waiting: the quote itself is quick to make, but gathering prices and terms takes days. One source for prices and a template with fixed building blocks remove most of the work. After that, an assistant can prepare the draft.
• Supplier invoices: invoices are matched to orders and coded by hand. How to tackle that step by step is in automating invoice processing.

Where does AI fit into process optimisation?

AI is good at work with a lot of variation in the input: reading text, summarising, classifying, making a suggestion. It is less suited to work that must be exactly the same every time and where an error is not acceptable. There, fixed rules are faster, cheaper and more reliable.

So work in this order. Improve the process first, then automate the fixed rules, and use AI only where people now read and interpret. When AI is not a good idea, read this article.

Do it yourself or bring someone in?

With the plan above you can improve one process perfectly well yourself. Outside help mainly makes sense in three situations: when nobody inside has time to do it properly, when you need to compare several processes to decide where to start, or when the solution needs an integration or software you won't build yourself.

For that situation we have the AI Opportunity Audit: five days, one figure, three opportunities. We talk to the people who do the work, work out what the current work costs and rank the three opportunities with the highest value. The report also says what we would not do. At De Duurzame Jongens, the audit identified €500,000 a year in growth opportunities.