Skilled people are scarce, administration grows with every new customer, and at the same time AI promises huge relief. That’s why many mid-sized companies want to automate processes – and face the same question: where do we start? This article shows a pragmatic path from a first overview to a measurable pilot, without a big-bang project and without replacing your existing software.
Typical time sinks in mid-sized companies
The biggest potential rarely sits in one large process, but in many small, repetitive tasks. Typical examples:
- Duplicate data entry: the same customer data is copied from the email into the CRM, then into a spreadsheet and later into a quote.
- Invoices and receipts: collecting, checking and matching receipts, and chasing approvals.
- Sorting requests: a shared inbox where every request first has to be read, classified and forwarded.
- Reports: copying numbers from several systems together to present them once a month.
- Onboarding: accounts, documents and appointments for new employees are triggered by hand every time.
- Knowledge in people’s heads: information only one person can find – and that’s missing when that person is on holiday.
Each of these tasks seems small. Together they tie up many hours of skilled working time week after week.
The prioritization matrix: which process first?
Not every workflow is an equally good starting point. Rate your candidates on five questions, each as “low”, “medium” or “high”:
- Frequency: how often does the task occur – daily, weekly, monthly?
- Effort per case: how much time does it take each time?
- Rule-based: does it follow clear rules or need many case-by-case decisions?
- Data: is the necessary information digital and accessible?
- Impact of errors: what happens if something goes wrong – and how easily can a human review be built in?
The best pilot candidates are frequent, time-consuming, rule-based and digitally available – with manageable consequences of errors. High-risk processes aren’t excluded, but they’re better started with a clear human approval step.
An example from practice
A company compares three candidates: invoice approval, sorting customer requests and creating individual quotes. The quotes take a lot of effort but are highly individual. Invoice approval is rule-based, happens often and can be safeguarded with a human approval. It becomes the pilot – the quotes follow later, once experience and trust have grown.
Planning the pilot
A good pilot is small enough to deliver results quickly and big enough to show real value. Before you start, clarify:
- Scope: which sub-process exactly? For example, only invoices from one inbox or only requests about one product.
- People involved: who works in the process today, who decides, who tests?
- Systems: which existing tools need to be connected? In many cases, existing applications can be connected via APIs instead of being replaced.
- Control: at which points does a human review and decide?
- Data privacy: which data is processed, which tools are approved?
- Acceptance criteria: how will you know the pilot is a success?
Also plan time for testing with real exceptions – faulty inputs, incomplete data, special cases. That’s exactly where you find out whether an automation holds up in everyday work.
Measuring the benefit
Without measurement, automation remains a gut feeling. So decide before you start what you’ll compare:
- Processing time per case,
- Lead time from receipt to completion,
- Rework and error rate,
- Running costs for licenses, usage and support,
- Satisfaction of the employees who work with the new workflow.
Collect the baseline values over a few weeks and compare after the pilot. That gives you a solid basis for deciding whether and how to automate further processes.
Frequently asked questions
Do we need an AI strategy first?
You don’t need a big strategy to get started. A clear first use case with a measurable goal matters more. The strategy often emerges naturally from what you learn in the pilot.
Do we have to replace our software?
In most cases, no. Automation connects existing systems and takes over the work between them. We check upfront which connections are possible.
Who in the company should own the pilot?
Ideally someone from the business side who knows the process and is allowed to make decisions – supported by IT and, if needed, an external delivery partner.
Conclusion
In mid-sized companies, getting started with automation works best small and concrete: collect typical time sinks, prioritize with a simple matrix, deliver a clearly scoped pilot and measure the benefit. Our four-step approach shows how we do it. If you want to find out which workflow pays off first for you, AI and process consulting helps – or discuss your process with us directly.
