16 October 2026 · 5 min read · by Bart Broothaerts
How to start with AI in your company, without a big project
Short answer: Start with one concrete use case where time or money is visibly lost today, build it on your real data and existing systems, measure it against the old way of working, and only expand once it pays off. Most companies get further this way than with a large AI strategy or transformation programme.
Why big AI projects stall
Large AI programmes often start with a strategy, a vendor selection and a data clean-up before anything useful reaches the team. By the time something is live, priorities have changed. Starting small avoids that: value arrives early, and what you learn shapes what comes next.
Finding the first use case
A good first use case usually has these traits:
- It happens often: daily or weekly, not once a quarter.
- It follows recognisable rules, even if people apply judgement.
- It costs visible time or money today, or loses customers.
- The data it needs already exists in your systems.
Typical first steps
In established companies, the first use case is often one of these:
- Customer questions and orders on WhatsApp or email, answered and recorded automatically.
- Management questions about revenue, margins or operations answered from live data.
- A recurring follow-up, such as overdue invoices or expiring documents, handled by an agent.
What you need internally
One sponsor who can make decisions, someone who can give access to your systems, real examples of the work, and a team willing to test early versions. You do not need a data team or a new ERP to start.
Measure, then grow
Agree up front how you will judge the first use case: hours saved, faster responses, fewer errors, more orders. If it works, the same connections to your systems make the second use case faster to build.