Process automation for SMEs: what is worth it

Process automation is worth it for SMEs when a task happens often, follows fixed rules and goes wrong when done by hand: transferring orders, passing on invoice data, reconciling master data. Tasks with many exceptions or genuine judgement calls are better left to people. Looking at the workflow itself before choosing a tool means automating the right steps and avoiding expensive detours.

In many companies the trigger is very concrete: someone types web shop orders into the inventory system every morning, copies customer details from one program to another, or exports lists to merge them in Excel. These steps take time, are tiring and produce exactly the typing errors that later surface in accounting or shipping. The question is rarely whether automation is possible, but what to automate first and with which tool.

How do you tell whether a task is suitable for automation?

A task is suitable for automation when three criteria come together: frequency, rule-based logic and error-proneness. The more all three apply, the clearer the benefit. If only one applies, it is worth a closer look before investing effort.

• Frequency: the task comes up regularly, not just now and then.
• Rule-based: the task follows fixed rules that can be written down – if field A looks like this, B happens.
• Error-prone: done by hand, it produces typos, forgotten steps or duplicate entries that cause rework later.
• Clear data source: the data already exists digitally, ideally in a system with an API or at least in a fixed export format.
• Clear ownership: someone in the company can say what the correct process looks like.

Which tasks should a small business not automate?

Not every annoying task is a good automation candidate. Tasks that are rare, keep changing or require professional judgement often cost more upkeep as automations than they save in effort. A special offer for a long-standing customer, a complaint with a history, or the decision to bring an order forward belong with a person who knows the situation.

Another warning sign: nobody in the company can describe the process clearly. If several employees handle the same task in different ways, the process needs to be clarified first. Automating an unclear process only makes it unclear faster – and errors surface later because no person is looking any more.

• Rare tasks where setting up the automation takes longer than years of manual work
• Processes with so many special cases that they would constantly need new rules
• Decisions involving discretion, customer relationships or legal consequences
• Processes that are being redesigned anyway – redesign first, then automate

Which workflows do SMEs typically automate?

Most useful automations in small and mid-sized companies are unspectacular. They connect programs that so far only talk to each other through a person, and remove retyping from daily work. Typical examples from retail, manufacturing and services:

• Importing web shop orders automatically into the ERP or inventory system
• Detecting incoming invoices in a mailbox, filing them and passing them to accounting
• Creating new CRM customers in the accounting system without entering them twice
• Regularly reporting stock levels to the web shop or marketplaces
• Generating delivery notes and shipping labels from an order and notifying the customer
• Sending internal notifications when an order reaches a certain status

Workflow tool or custom integration: which fits when?

Workflow tools such as n8n, Make or Microsoft Power Automate connect applications using ready-made building blocks. For many workflows with moderate data volumes they are a good choice: quick to set up, easy to follow and no custom software development needed. Power Automate is a natural fit if a company already works with Microsoft 365. n8n can also be self-hosted, which can matter for sensitive data and GDPR questions. Make is a cloud service with many ready-made connectors.

A custom integration – a small, purpose-built link between two or more systems – makes sense when the tool reaches its limits: large data volumes, complex mapping rules, a system without a suitable API, strict error-handling requirements, or a business-critical process that should be versioned and traceable. The answer is often a mix: simple workflows in the tool, the critical core connection as a custom integration.

• A workflow tool fits with: few systems with ready-made connectors, moderate data volumes, simple rules
• A custom integration fits with: missing or unusual APIs, many special rules, business-critical processes, high traceability requirements
• Needed in both cases: error alerts, documentation and a clearly responsible party

Why does maintenance belong in automation from the start?

An automation is not a one-off project but a small system inside the company. Vendors change their APIs, credentials expire, an application gets an update and a field suddenly has a different name. If nobody is told, the workflow silently stops working: orders no longer arrive, invoices pile up, and it only gets noticed when a customer complains.

That is why every automation needs three things from day one: monitoring that detects when a run fails or does not happen; an alert to a person who takes care of it; and short documentation of what the workflow does, which accounts it uses and who to ask when something goes wrong. What this ongoing operation looks like in practice is covered in our article on operating and maintaining automations.

How do you approach process automation step by step?

1. 01Collect tasks: ask employees which recurring activities annoy them most and where errors happen.
2. 02Write down the process: for the most promising candidates, record step by step who does what with which data in which program.
3. 03Assess against the criteria: rate frequency, rule-based logic and error-proneness, and start with the clearest process.
4. 04Choose the tool: check which APIs the systems involved offer, then decide between a workflow tool and a custom integration.
5. 05Start small: implement one workflow first, test it with real data and check it in parallel for a while.
6. 06Define operations: settle monitoring, alerts, documentation and ownership before switching off the manual route.
7. 07Expand: only tackle the next workflow once the first one runs reliably.

When does process automation need AI?

Most automations in SMEs do not need artificial intelligence, but clear rules and clean interfaces. AI becomes interesting when unstructured content has to be processed: assigning emails to a case, reading data from PDF documents or pre-sorting enquiries. Here too, AI starts at the wrong end if the process behind it is not understood. Where several steps are meant to run on their own across different systems, an AI agent with approval points is an option. When personal data is processed, GDPR and, depending on the use, the EU AI Act need to be considered – a locally hosted model can be an option here.

How NDVDL takes on automation in your business

NDVDL does not start with the tool but with the workflow. We come to your business and have you show us who does which steps in which program, where data is entered twice and where errors creep in. You then receive a written proposal: which workflows are worth automating, which are not, and what to build them with – a workflow tool, a custom integration or a combination.

After implementation, we operate and maintain the automation. NDVDL monitors the workflows, receives the error alerts, adapts the interfaces when vendors change something and keeps the documentation up to date. You have one fixed contact person for network, servers, applications and the connections between them, and do not need to check yourself whether everything is still running.