How to use AI in your business: a guide in 5 steps
How to get your business started with AI: find the right tasks, say no to the wrong ones, test small and measure the effect. A practical guide without jargon.
Short answer
AI works best when you start with a concrete task, not the technology. Find a task that takes a lot of time, looks the same every time and requires manual work today. Test a small solution with the people who do the task, and measure whether it saves time before you expand.
Step 1: Find out where the time goes
Go to where the work happens and look for:
- Tasks where information is copied from one system to another.
- Questions that get asked over and over.
- Emails that have to be read, sorted and turned into something – a quote, an order, a case.
- Reports that are put together by hand every week or month.
Ask your employees. They know exactly what takes time.
Step 2: Choose the right tasks – and say no to the rest
Not every AI idea is a good one. Prioritise using two questions:
- How big is the gain? How much time or how many errors does it save?
- How hard is it? Is the data available, and is the task well defined?
Start with tasks that deliver a lot and are easy. Say no to the ones that are hard and deliver little – even if they sound exciting.
Step 3: Set the guardrails
Before the AI goes live, decide:
- Which systems and data can it see?
- What can it do by itself, and what must a person approve?
- How will you log what it does?
- Are personal data and GDPR in order?
Step 4: Test small
Build a small version for one workflow and a few employees. Spend 2–6 weeks, gather feedback and adjust. This is where you learn the most.
Step 5: Measure and decide
Measure before and after: time per task, number of errors, response time to customers. If it adds value, expand. If not, stop – that’s also a good result, because you saved the money on a big solution.
Five questions you should be able to answer before you start
The AI projects that stall rarely stall because of the technology. They stall because nobody could quite answer the basic questions before starting.
1. Where does the time actually go?
Most people have a feeling for where the work is heavy. But the feeling and the reality are often two different things. It is rarely the big, visible task that steals the time. It is the small things that repeat all day: looking for information, copying from one system to another and answering the same questions again and again.
Can you point to the 2–3 workflows where the most hours are lost every week?
2. Who does the work today – and how?
Management sees one picture of a workflow. The employees who do it every day see another. They know the shortcuts, the spreadsheets next to the system and the exceptions that never appear in any process description. If an AI solution is built from management’s picture, it rarely fits everyday work – and then it is not used.
Do you know how the work is actually done, not just how it should be done?
3. Which data and systems will the AI use?
AI is only as good as the information it has access to. Are customers, products and prices in the ERP? Are the agreements in emails? Is the data in one place, or does it need to be pulled from several systems? The answer decides how big the project becomes – and what it costs.
Can you say where the data the solution needs is stored?
4. What may the AI do – and what may it not?
Should the AI only read and suggest, or may it also create and send? Where must a person approve? Which tasks should not be left to AI at all – e.g. because of personal data or the EU AI Act? These questions are easier to answer before the solution is built than afterwards.
Have you decided where the line goes, and who is responsible?
5. How will you see that it works?
“We need to use AI” is not a goal. “We need to spend half as long on quotes” is. Without a concrete goal, nobody can say whether the project was worth the money – and it is hard to get approval for the next step.
Do you know what should be better in 90 days, and how you will measure it?
Most companies can answer a couple of the questions, but not all of them. That is completely normal. It just means the first step should not be to build something, but to understand your own work.
Common mistakes
- Starting with a tool instead of a problem.
- Buying AI for everyone without knowing what it’s for.
- Forgetting the employees – they need to be involved from the start.
- Going straight to a big solution without a test.
How BCT helps
Our Shadow Walk is exactly steps 1 and 2: we spend ½–2 days with you, follow the work, find the tasks where AI adds the most, and give you a prioritised plan at a fixed price. Then we can build the solution – and you own the code.
Frequently asked questions
Where should a small business start with AI?
With one task that takes a lot of time and looks the same every time – such as email, quotes or common questions.
Do we need an AI strategy first?
A short plan is good, but it should be based on concrete tasks – not on the technology.
How long does it take to get started?
A first test can often be running within a few weeks when the task is well defined.
What if employees are sceptical?
Involve them from the start. When they see the AI taking over the dull groundwork, they often become its biggest fans.
In short
AI delivers the most value when you start with a concrete task rather than the technology. Find out where the time goes, pick the tasks that repeat, and say no to the rest.
Set the guardrails, test small with the people who do the task today, and measure whether it works. Only once you can see the gain does it make sense to expand.
Want to see what it could look like for you?
We are happy to have a no-obligation talk about your day-to-day work and where AI and software can make a difference.



