The costs of AI rarely arise where you expect them. According to Bitkom, the biggest cost items for companies using AI are infrastructure such as cloud and computing power (51 percent), data preparation (50 percent) and integration with existing systems (41 percent). Only 8 percent name usage of the AI model itself as a major cost. No wonder 54 percent of users and 63 percent of non-users see hard-to-calculate costs as an obstacle. A simple calculation clarifies in advance whether an automation pays off.
What are the costs made of?
- Setup: analysing the process, building the automation, connecting it to your software, testing.
- Data: preparing data so the AI can work with it, for example consistent formats or cleaned master data.
- Operation: server or cloud, licences, AI usage fees.
- Maintenance: updates, adjustments when software or processes change, and monitoring.
- Rollout: your team needs to know how to work with the automation and what to check.
And what does it bring?
The benefit is usually time, plus fewer errors, faster responses to inquiries and work that no longer piles up when someone is on holiday. Time is the easiest to put into numbers.
A sample calculation
The following figures are assumptions for illustration. Use your own values.
A receipt is currently checked and booked by hand, which takes 15 minutes. This happens 40 times a week.
- Time per week: 40 × 15 minutes = 10 hours.
- Time per year: 10 hours × 45 working weeks = 450 hours.
- Value of that time: at assumed internal costs of 50 euros per hour, that is 22,500 euros a year.
- Realistic saving: the automation does not take over everything. Exceptions remain, and someone checks approvals. Assume, for example, 70 percent, so 15,750 euros.
- Offset: subtract the running costs per year and divide the one-off costs by what remains. The result is the time after which the automation has paid for itself.
If little remains after this calculation, that is a result too. Then automation does not pay off for this process, and it is better to know that beforehand.
Where are the hidden costs?
- Forgotten exceptions: the normal case is quickly automated; the special cases take the time.
- Messy data: inconsistent master data makes every automation more expensive.
- Starting too big: one huge project at once instead of one process after another.
- Overlooked running costs: operation and maintenance belong in the calculation, not just the build.
How we do it
We start with a free AI check and calculate your process together. You then receive a quote with a fixed price before you commit to anything. Our starting prices are in the price list. Local models can lower running costs further, more in Local AI instead of the cloud.
Sources
- Bitkom: For the first time, most companies use AI (14 September 2026, German)
