Which process should you automate first? A guide in four steps
5 min read
Short answer
Automate a process first that comes up often, follows clear rules, costs noticeable time and is important but not business-critical. Collect all candidates, rate frequency, time per case, rule-based structure and risk, estimate the effort, and start with the process that offers the best ratio of benefit to effort.
What does this mean for your company? Book an intro call
Almost every company has more ideas for automation than time and budget. The key question is therefore not whether, but where to start. A good first process pays off quickly, builds trust in the team and creates appetite for the next step.
Step 1: Collect candidates
Ask the people who do the work: “Which task would you most like never to do again?” The answers are often remarkably specific. Typical candidates are transferring data from one program to another, entering invoices, writing order confirmations, compiling reports or forwarding enquiries.
Write each candidate down in one sentence: what happens, how often, and who does it?
Step 2: Rate each candidate
Give one to three points for each of four criteria:
- Frequency: many cases a day earns three points, once a month earns one.
- Time per case: the longer a case takes, the more can be saved.
- Rule-based: does the process follow clear rules, or does every case need judgement?
- Low risk: are errors easy to spot and correct? Then it earns three points.
The candidates with the most points make the shortlist. For them, the detailed calculation from Is process automation worth it? is worthwhile.
Step 3: Estimate the effort
Benefit alone does not decide. Also ask:
- Is the data available digitally, or does it arrive on paper?
- Do the programs involved have interfaces? More in Typing data in twice?
- Is the process stable, or is it being rebuilt right now?
A process with somewhat less benefit but clearly less effort is often the better start.
Step 4: Start with one candidate
Choose one process and implement it completely instead of starting five at once. Measure how long a case takes before and after. This visible success is the best argument for the next project.
When should you not automate yet?
If responsibilities are unclear or everyone does the process differently, you only automate the confusion. Then it pays to simplify the process together first. Especially for processes with lots of text and documents, AI can help where classic rules reach their limits. How to find such cases is covered in Finding AI use cases.
For a first calculation with your own numbers, use the automation calculator.
Julius Launhardt helps small and mid-sized companies find the right first process, and then automates it. More on the process automation page.
Frequently asked questions
Should I start with the biggest process?
Usually not. A medium-sized process with a quick, visible success is the better start. Your team builds trust, and you gain experience before tackling the big topics.
Who should have a say in the selection?
Above all the people who do the work today. They know where time is lost and which exceptions exist. Add the team lead and, if you have one, IT.
What if no process clearly stands out?
Then take a closer look at the numbers. For two or three candidates, measure for a week how often they occur and how long they take. After that the decision is usually clear.
Do I need outside help for the selection?
Not necessarily. You can carry out the rating in this article yourself. An outside view helps mainly with estimating effort, where experience with systems and interfaces counts.