This guide is a general technical overview. It is not legal advice and not a legal review within the meaning of the German Legal Services Act (RDG). For an authoritative legal assessment, we recommend having a lawyer review your individual case.
A process analysis for automation answers one simple question: which recurring work in your business takes the most time and follows rules clear enough for software to take over? To find out, you spend one week writing down your time sinks, score them by frequency, time, error rate and how rule-based they are, start with the easiest win and test it in a small pilot. In a small business this takes a few weeks, not a consulting project with a marathon of workshops.
My take on it: most small businesses don't need an AI strategy, they need an honest tally sheet. Anyone who spends a week writing down where the office hours go usually knows afterwards where to start.
The analysis has five steps. You can do the first three entirely on your own; for steps four and five an outside view helps.
Start a simple list, on paper, in Excel or in a notes app. Every time you or your office handle a recurring task, add a line: what was it, how long did it take, where did the information come from (phone, email, WhatsApp, a scrap of paper)? Pick an ordinary week, not the one before the holidays.
Ask your team, too. The person running the office usually knows better than the boss how often they chase customers for missing photos or measurements.
For every task on your list, ask four questions:
Frequency times time gives you the weekly hours for that task. That is your lever. Error rate and rule-based tell you whether the lever can actually be pulled.
Don't start with the biggest chunk. Start with the one that eats a lot of time and is easy at the same time. A fast, visible success convinces your team more than any presentation and shows you whether working with software fits your day-to-day. The big overhaul comes later.
Now describe how the workflow should run in future, step by step: what comes in, what happens automatically, where does a person take a look? At the same time, check which programs you use today (trade software, calendar, email, accounting) and whether they have an interface. That determines how much work the implementation will be.
Typical building blocks we use at Softstorm for this kind of workflow:
Whether a workflow needs a fixed chain of rules or genuinely needs AI is a separate question. We cover it in the guide Workflow vs. AI automation.
Before you switch everything over, run the new workflow on a small scale for four to six weeks, for example only for enquiries coming in via the web form. Measure the same things as in step 1: minutes per task, number of follow-up questions, errors. That is the only way to see in black and white whether it pays off, rather than just whether it feels good.
This table is enough for the first three steps. Copy it into Excel or onto a sheet of paper and fill it in for one week:
| Task | How often per week? | Minutes per task | Errors or rework? (1–3) | Always runs the same way? (1–3) | Which programs? |
|---|---|---|---|---|---|
| Example: answering enquiries | 15 | 10 | 2 | 3 | Email, phone |
| … |
How to evaluate it:
This template doesn't replace advice, but it gets you into any conversation with real numbers, including one with us.
In trade businesses, the same workflows almost always end up at the top of the list. What they have in common: they happen often, follow fixed rules and keep you away from the actual work on site.
You'll find an overview of these use cases on our page for trade businesses and specifically under automation for trade businesses.
That adds up to around 10 hours a week. If an automation takes over most of the work for enquiries and appointments, and quotes only need checking instead of writing from scratch, the arithmetic suggests savings of roughly 5 to 6 hours a week. That is an assumption, not a measured result. Which is exactly why step 5 exists: the pilot shows you what actually remains in your business.
One note in case enquiries will run through a chatbot on your website: since 2 August 2026, Art. 50 of the EU AI Act sets out a transparency obligation, so users must be able to tell that they are chatting with an AI. More on this in the guide on labelling requirements for chatbots. Whether a chatbot or an AI agent fits your case better is explained in Chatbot or AI agent?.
The most important limit first: you don't automate a bad process. If every enquiry arrives through a different channel today, prices only exist in the master craftsman's head and nobody knows who is responsible for callbacks, software just makes that chaos faster. Simplify first: settle on one inbound channel, maintain one price list, clarify who does what. That alone often saves noticeable time, without any AI.
I'd also advise against it in these cases:
What an automation costs depends less on the AI than on your setup. The main factors:
You have the biggest influence yourself: start with a single, clearly defined workflow, tidy up your data first and leave special cases with a person for now. How costs break down for individual solutions is covered in the guides on AI agent costs and AI chatbot costs.
A process analysis is not a major project. With one week of tally sheets and the template above, you'll know which two or three workflows really eat up time in your business, and you can test one of them in a targeted way. If you'd rather do the scoring together, take a look at our AI consulting and the implementation side under automation. The simplest next step: bring your completed list to a free AI potential check (30 min), or get in touch for an initial consultation.
Note: The information on the GDPR and the EU AI Act is a technical overview, not legal advice.
Free AI potential check (30 min)
Check for free nowIn five steps: first you spend one week noting down which recurring tasks eat up time. Then you score each task by frequency, time spent, error rate and how rule-based it is. Next you pick a quick win as your starting point, plan the target process with your existing systems and test it in a small pilot with clear measurement.
Good candidates are tasks that happen often and follow fixed rules: sorting and answering enquiries, preparing quotes, scheduling appointments, capturing measurements and delivery notes, preparing invoices and pre-screening job applications. Decisions on price, orders or hiring stay with a person; the AI only does the groundwork.
A good candidate happens at least several times a week, takes noticeable time each time, follows understandable rules and has clear inputs such as an email, a form or a photo. If every case runs differently or the knowledge only lives in one person's head, the process is not ready yet.
Yes, you can handle the as-is review and the scoring yourself with a simple spreadsheet. For one week, write down every recurring office task with how often it happens and how long it takes. Outside help pays off mainly for designing the target process, checking which systems can be connected and setting up the pilot.
When a process happens rarely, runs differently every time or is already chaotic today. Automating a bad process does not make it better, it just makes it go wrong faster. Simplify first, then automate.
Mainly the number of systems that need to be connected, whether they have an interface, how clean the data is and how many special cases there are. Starting with a single, clearly defined workflow keeps the effort small and quickly shows whether it pays off.
This guide is a general technical overview. It is not legal advice and not a legal review within the meaning of the German Legal Services Act (RDG). For an authoritative legal assessment, we recommend having a lawyer review your individual case.
What is inside
Step 1 of 3
We analyze your processes, identify automation potential and show you which digital systems really make sense.
