In short
What a company should automate is decided by measuring, not by gut feeling. Zircon's AI Potential Audit takes two weeks: we talk to the team, map the processes, measure where time goes on manual work, and rank what is worth automating and what isn't. On the company's side, it takes about two working days.
Key points
- Automation pays off where there is a lot of repetitive, rule-based manual work and the data is already digital.
- The loudest complaint is not always the biggest loss of time, so measure before you decide.
- The AI Potential Audit takes two weeks and asks for about two working days from the company.
- At the end of the audit you get a fixed price for the next step, before anyone commits.
- Our view: measure first, build second, and one process at a time.
What is worth automating in a company?
Automate what repeats often, follows rules and already has its data in digital form: incoming invoices, orders, quotes, e-mails, regular reports. Don't automate what is rare, different every time, or needs human judgement to decide.
In practice the line is rarely this clean. A task can be annoying and still not worth automating, because it happens once a month. Another looks trivial but repeats a hundred times a day. These signals help you sort them.
| Signal | Worth automating | Probably not worth it |
|---|---|---|
| Frequency | many times a day or a week | once or twice a month |
| Rules | you can write down what to do when | every case is different |
| Data | in an e-mail, a PDF, a spreadsheet or a system | on paper or in someone's head |
| Type of work | copying data from one place to another | judgement, negotiation, relationships |
| The process | stable, everyone does it the same way | still changing |
- Frequency
- Worth automatingmany times a day or a week
- Probably not worth itonce or twice a month
- Rules
- Worth automatingyou can write down what to do when
- Probably not worth itevery case is different
- Data
- Worth automatingin an e-mail, a PDF, a spreadsheet or a system
- Probably not worth iton paper or in someone's head
- Type of work
- Worth automatingcopying data from one place to another
- Probably not worth itjudgement, negotiation, relationships
- The process
- Worth automatingstable, everyone does it the same way
- Probably not worth itstill changing
It can also turn out that a task needs no artificial intelligence at all, just a simple rule or a setting in software you already have. That is a good result too.
Why measure before you automate?
Because gut feeling picks the loudest problem, not the most expensive one. Measuring turns opinions into working hours, and working hours can be compared.
According to Eurostat's 2025 data, 19.95% of EU enterprises with at least 10 employees used some form of AI technology. Among small companies with 10 to 49 people the share was 17%, among medium-sized companies with 50 to 249 people 30.36%. EU companies that considered AI but did not go ahead most often named lack of expertise as the reason: 70.89% of them. In our view, lack of expertise often simply means that nobody can say where to start. That is what measuring is for.
How does the AI Potential Audit work?
The AI Potential Audit is Zircon's two-week assessment that shows what is worth automating in a company, and what isn't. It has four steps.
- 01
We talk to the team. We care about the process as it actually runs, not as it is written down.
- 02
We map the processes. Where the data comes from, who does what with it, which system it lands in, and where it is copied by hand from one place to another.
- 03
We measure where time goes on manual work. How often a task comes up, and how long it takes each time.
- 04
We rank what is worth automating and what isn't. The ranking also says what is not worth your attention right now.
On the client's side this is about two working days of effort. The rest of the two weeks is our work.
What happens after the two weeks?
At the end of the audit we give a fixed price for the next step, before anyone commits. If the company goes ahead, the Sprint follows: AI automation of a single process from start to finish, connected to the tools the company already uses.
Where money or customers are involved, we build in validation rules, a log of every step and human approval. A person presses the button.
After the Sprint comes the handover, with up to 30 days of aftercare, then operation: we keep running the system after go-live. That way measuring, building and running stay in one pair of hands.
What does this mean for a company of 10 to 250 people?
A company of 10 to 250 people rarely has someone whose job is to measure processes. So the decision to automate is often made on instinct: someone saw a good demo, or got fed up with a spreadsheet. That is understandable; it just gets expensive if you start with the wrong process.
Our advice: don't buy a new tool or start a big project until you know where the time goes. Start with one process whose benefit you can state in working hours. If that works, move on to the next. That is how we work too: measure first, build second, one process at a time.