What is AI automation?
Automation connects programs so that data flows from A to B without retyping. AI comes in when a step has to understand text: reading an email, recognising a request, writing a reply. Tools such as n8n or Make run the workflow, and a language model such as GPT, Claude or Apertus handles the text work.
Which seven processes are worth it?
| Process | What happens automatically | Effort |
|---|---|---|
| 1. Sort enquiries | AI recognises the request, creates the contact in the CRM and suggests a reply | 1 to 2 weeks |
| 2. Prepare quotes | Form entries become a draft quote with prices from your price list | 2 to 3 weeks |
| 3. Capture invoices | Incoming PDF invoices are read and passed to accounting | 1 to 2 weeks |
| 4. Transfer data | New customers from the store land automatically in accounting and the newsletter | 1 week |
| 5. Coordinate appointments | Appointment requests in emails are recognised and proposed in the calendar | 1 to 2 weeks |
| 6. Maintain product data | Product range, prices and stock flow from the supplier into the store | 2 to 4 weeks |
| 7. Answer standard emails | Frequent questions get a draft reply, and a person approves it | 1 week |
How do I find the right process to start with?
Pick a process that happens at least once a week, has clear rules and where someone copies data today. Do a rough calculation: minutes per run times runs per month. From about five hours per month, an automation usually pays off within the first year.
How much time does an automation save?
Do the maths yourself: minutes per run times runs per month. If someone transfers 20 enquiries a day into the CRM by hand and needs 3 minutes for each, that adds up to around 20 hours a month. With an automation, this time disappears almost completely. Running costs are the automation tool and the use of the AI model, usually a small amount per month at SME volumes.
What about data protection?
Personal data may only go to services you name in your privacy policy, and providers outside Switzerland need an adequate level of protection. More on this in the article on AI and the revised Federal Act on Data Protection (revFADP).