5 Manual Processes Still Running by Hand at SAP/Netsis Companies

By Can Kuşaklı · Published October 16, 2026

At manufacturers on SAP/Netsis, the same five processes keep running by hand.

All five share one trait: nobody complains. Everyone knows the work is done by hand, but nobody has named it a problem. It has become "that is how the work runs."

In this article we take the five one by one: why they stayed manual, what automating each looks like technically, and which to start with. None of them is exotic. You probably have at least two.

The common fix for all five is an automation layer. An automation layer is a piece of software that sits on top of your existing business system - whether that is SAP or a Microsoft stack - and runs the repetitive work a person does by hand, with the same permissions and without changing the system itself.

Which five processes run by hand at SAP/Netsis companies?

How do you automate a spare parts stock report?

How it runs today: Every Monday morning the same query is run, the output goes into Excel, two columns are deleted, one is added, headers are formatted and the file is emailed to five or six people. The only thing that changes is the date.

Why it stayed manual: Because it is nobody's job. The person preparing the report does it next to their real work, and because the task looks small enough, it never became a project.

What automating it looks like: The easiest in the group. No step needs a decision, the filters are fixed, the format does not change. A scheduled task pulls the report, the formatting rules are defined once, and delivery goes to the same mailing list. No AI needed - this is a scripting job.

Watch out for: When extracting the formatting rules, ask "why is it done this way" at every step. There is usually a reason a column gets deleted, and that reason is written down nowhere.

Can you automate order and delivery note matching?

How it runs today: Two screens open side by side, rows compared by eye, mismatches noted in a third file. Then phone calls start from that file.

Why it stayed manual: Because matching is not always one-to-one. Product codes may be written differently, partial shipments can correspond to one order, unit conversions get in the way. It looks too variable to solve with rules - so nobody tries.

What automating it looks like: This is the first job in the list that needs AI. Matching two sets of records is done by similarity-based matching, not full-text comparison. The system marks high-confidence matches automatically and leaves low-confidence ones to a person.

The key design decision: the goal is not one hundred percent automation. If the large majority of rows match automatically, the person only looks at what is truly uncertain. The workload does not vanish, but it shrinks and, more importantly, attention goes where it is needed.

Watch out for: This process has the highest error rate, because comparing rows by eye is tiring and tired eyes skip. The real gain from automation here is not time but accuracy.

How do you automate receivables ageing follow-up?

How it runs today: The ageing report comes out of the system fine. But the person deciding who to call is still a human, working from notes set aside on the side. "This customer was called last week." "A payment plan was discussed with that one." "Do not touch this one, the sales manager is handling it."

Why it stayed manual: Because the real information is not in the report. The report only gives numbers; the context that drives the decision has not been entered into anyone's system.

What automating it looks like: What gets automated here is not the report but the follow-up loop. The layer takes the ageing data, adds the last contact information on top, and produces a prioritized list: who, how much, when they were last called, what the next step is.

The precondition is that those notes are written somewhere. At most companies this step means moving to a CRM or a simple shared sheet. That is the part that takes more effort than the automation itself.

Watch out for: Before trying to automate this process, pin down what information the decision rests on. If it is recorded nowhere, record it first.

How do you automate production plan revision notices?

How it runs today: The plan changes. Then someone informs the people they think are affected, one by one. Usually over WhatsApp.

Why it stayed manual: Because the list of "who needs to know" is defined nowhere. The person sending the news knows it from experience.

What automating it looks like: An event-based trigger. When the plan record changes, the layer starts, works out what the change affects and notifies the affected parties.

The hard part is not technical: defining who is affected by what. That is a rule set to extract by sitting down with the person who knows the process. Once extracted, the automation works and that knowledge no longer lives in one person's memory.

Watch out for: Who was not told only becomes clear when work goes wrong. So the real cost of this process is invisible and delayed - which is why it keeps being postponed.

Can a month-end close checklist be automated?

How it runs today: The same twenty checks, every month, by one person who holds them in their head. Sometimes the list is in an Excel file, mostly it is only in that person's head.

Why it stayed manual: Because it runs once a month. Tasks that do not repeat often never get automation priority - yet the risk is highest here.

What automating it looks like: Two layers. The lower layer automatically runs the checks a machine can do and reports the result. The upper layer queues those that need human judgement and marks the completed ones.

So the output is not a fully automatic close - it is a close whose status is visible. Which check is done, which is waiting, which is stuck. Even that is a big gain, because today that information is gathered nowhere.

Watch out for: When that person goes on leave, the list goes on leave too. Writing the list down is work worth doing even if the automation is never built.

Which of the five should you start with?

Do not take on all five at once. Order them by this rule: not the biggest, the one that repeats most often.

A two-hour report that runs weekly is worth more than a one-day job that runs monthly. It saves more time over a year, and because the result shows sooner, it builds trust in the team.

By that rule the typical order is: first the spare parts report (easiest, most frequent), then order and delivery note matching (highest error risk), then the other three.

We also wrote about how the layer connects to SAP and where AI is truly needed: How to Automate SAP Reporting Without Replacing SAP. To work out what these jobs cost your company: The real cost of manual reporting.

Does your company have this gap?

Five questions:

• Do you pull the same report by hand more than once a week?
• Is there someone who compares two screens by eye?
• When the one person who knows a process goes on leave, does the work stop?
• Does a critical notification go out over WhatsApp?
• Does the month-end close live in someone's memory?

If you ticked two or more, the automation layer is missing.

None of this comes from a shortcoming in SAP/Netsis. The data is in the system, exactly where it should be. What is missing is a thin layer on top that nobody ever wrote.

Let's talk

Akyalabs builds this layer in three ways: implementation training that teaches your team to build it, one-to-one collaboration on specific processes, and custom product development partnerships where it makes sense.

Whichever of the five is running by hand at your company, we can start the conversation from there.

The automation layer is missing - not the system.