Article

Where to start with AI? The smallest first project

Not knowing where to begin with AI is normal. This step-by-step guide helps you choose the smallest meaningful first project.

Written by Loek Delahaye, founder, Delahaye Solutions · 10+ years software architect and CTOPublished:
Short answer
  • Do not start with the most impressive application. Start with the dullest, most repetitive task that currently eats hours.
  • Write down for a week which tasks you do that are always the same. That list is your roadmap.
  • Choose the task with the highest frequency and lowest variation. That is almost always the best starting point.
  • A first project does not have to be a big project. It can also be small, as long as it proves that automation works.

Most businesses start in the wrong place

When business owners think about AI, they usually think about the most visible application: a chatbot on the website, generated product photos or a fully automated sales process. Those applications exist, but they are rarely the best starting point. They are complex, expensive and require the foundations to already be solid. The best starting point is the task nobody enjoys, that is always the same and comes back every week. That task is not glamorous, but it is predictable. And predictable is exactly what AI handles well.

The step-by-step plan

Five steps to your first AI project

This is not a theoretical framework. This is what works in practice for businesses that actually start with AI, not just talk about it.

Step 1: Write down everything you repeat for a week

Track for a week which tasks you do that are always roughly the same. It does not need to be a formal system: a notepad or a simple notes app on your phone is enough. Note the task, how long it takes and how often it occurs. After a week you have a list you can work with.

Step 2: Choose the task with the highest frequency and lowest variation

From your list, choose the task that occurs most often and varies least from case to case. The more often something recurs and the more it is always the same, the more there is to save. Variation makes automation harder. Frequency makes the saving larger. Those two factors together determine the value of the project.

Step 3: Write down the step-by-step plan

Take the chosen task and write down exactly which steps you take, from start to finish. What information do you need? Where does it come from? What do you do with it? What is the outcome? If you cannot write this down, the task is not yet ready for automation. If you can, you have already written half the specification.

Step 4: Build the smallest working thing

The first version does not need to do everything. It only needs to automate the core of the task, so you can measure whether it works. An automation that handles 80 percent of cases and forwards the rest to you is already valuable. Perfection is for version two.

Step 5: Measure the result and decide whether to continue

After four weeks you know how much time the automation has saved you. Compare that with what you invested. If the saving outweighs the cost, you expand. If not, you learn something valuable: which assumption was wrong? That is also worth the effort.

The pitfall

Do not wait for the perfect project

Most businesses that do not start with AI are waiting for a project that feels big enough to justify the investment. That project does not exist, or at least not as a starting point. The businesses that do start choose something small, learn how it works, and build from there. After six months they have three or four working automations that together save them dozens of hours a month. That is the difference.

Frequently asked questions

Frequently asked questions about starting with AI

Do I need technical knowledge to start with AI?
No. As a client you only need to know what task you want to automate and how it currently works. The technical execution is up to the party you hire. What you do need is a clear description of the process. The better you can explain it, the faster and cheaper the build will be.
What is a realistic result from a first AI project?
A realistic first project saves you a few hours per week on a specific task. That is not a revolution, but it is tangible and measurable. If that works, you know that automation works in your context and you can build further. Do not expect company-wide transformation from the first project.
What if my processes are not standardised?
Then start there. Documenting processes is a valuable step in itself, even without AI. Once you know how a task always runs, you can automate it. If every execution is different, standardisation is the first task.
How long does it take before a first AI automation is live?
A single automation is typically live within one to three weeks with us. That depends on the complexity of the task and the systems involved. We make an estimate up front so you know what to expect.
Can I start with AI without replacing existing software?
Yes. Most first AI projects connect to existing systems rather than replacing them. They add a layer that reads, processes and forwards information, while your existing tools keep working as normal. Replacement only becomes relevant if the existing tool itself is the bottleneck.
Read more

Want to know which task you should tackle first?

Book a free call. We look at your daily processes together and advise honestly which task is the best starting point.

Book my free call →

Free and without obligation · Reply within one business day · Honest advice, even if automation turns out not to be the best choice