Don't start with AI. Start with your processes.

Why do most AI projects start wrong?

The pattern is always the same:

1. CEO reads about AI
2. Feels that "we have to do something"
3. Calls an AI company
4. "Can you implement AI for us?"

Our question: "AI for what?"

Silence.

This isn't a criticism. It's understandable. AI is everywhere. Everyone's talking about it. You don't want to fall behind.

But AI without a clear problem is a hammer without a nail. Powerful tool. Nothing to use it on.

What happens when you automate a bad process?

Here's the uncomfortable truth:

AI automates. It accelerates. It makes things consistent.

But it doesn't make them smart. It doesn't organize. It doesn't clean up.

Example: A company automates their quoting process with AI. Problem: the input for quotes is spread across 4 systems. Nobody knows the right prices. There are 3 versions of the product list.

Result: AI now generates quotes in 30 seconds that are wrong 50% of the time. It used to take 2 hours, but at least then they were right.

They didn't solve their problem. They sped it up.

The right order for AI implementation

Three steps. In this order:

1. Understand your process
2. Fix what's broken
3. Automate what works

Not the other way around.

Step 1: Understand your process

Can you explain exactly what happens from start to finish? Not vaguely, concretely?

• What triggers the process? "Customer sends a request by email or form"
• Who does what? "Sarah checks inbox, creates ticket, assigns to sales"
• Where does information come from? "Pricing in Excel, customer info in CRM, specs in Notion"
• Where does the output go? "PDF to customer, copy in Drive, note in CRM"
• What are the exceptions? "Orders over €10K need manager approval"

If you can't answer those questions, you can't automate the process.

Step 2: Fix what's broken

Now that you understand your process, you can see where it goes wrong:

• Unnecessary steps: things you do "because we've always done it this way"
• Double work: entering the same info in multiple places
• Unclear handoffs: tasks that stall because nobody knows whose turn it is
• Information that gets lost: knowledge in heads, not in systems

Fix this first. Sometimes without AI. Sometimes the fix is simpler than you think: a shared spreadsheet, a checklist, a weekly standup.

Step 3: Automate what works

Only once you have a process that actually works do you start automating.

Now you know exactly what AI should do:

• What input comes in
• What needs to happen to it
• Where the output goes
• When a human needs to step in

No guessing. No "maybe we could…" Concrete tasks that can be taken over.

Example: notary office saves 15 hours a week

A notary office in Flanders wanted "AI for documents."

The situation:

• Documents came in by email, WhatsApp, post, and in-person drop-offs
• Nobody knew where anything was stored
• 30 minutes of searching per case, on average
• Errors from wrong versions or missing documents

Step 1: Map the process. We drew out what actually happened. That alone was an eye-opener. 4 different ways to do the same thing. No standard.

Step 2: Get the basics in order.

• One central place for documents (not 4)
• Clear folder structure and naming
• Standard intake: everything through one channel

That alone saved 8-10 hours a week. No AI involved.

Step 3: Add AI. Now AI could actually help:

• Read and classify documents automatically
• Pull out data (names, dates, amounts)
• Generate draft deeds
• Check whether all required documents are present

"But we don't have time to document our processes…"

We hear this a lot. "We're too busy with the day-to-day."

The irony: you're busy because your processes aren't in order.

The time you currently lose to:

• Searching for information
• Explaining how things work to new people
• Fixing mistakes
• Answering the same questions over and over

You can invest that time once into mapping your process. And then never again.

It takes 4-8 hours to map a process properly. It saves you dozens of hours a month.

The questions to ask before starting with AI

Before kicking off an AI project:

1. What costs us the most time right now? Where's the pain? What frustrates your team the most?
2. Could we explain this to someone new? Is it documented? Or is all the knowledge in people's heads?
3. What regularly goes wrong? Where are the gaps? Which mistakes do we keep making?
4. What would change if this were 10x faster? Is faster also better? Or does it create new problems?
5. What's the business impact? How many hours does this save? How much revenue does it unlock?

The answers tell you where AI makes sense. And where it doesn't.

The bottom line

AI is powerful. But it only works on a solid foundation.

You don't need an AI strategy. You need a process strategy that AI can be part of.

Don't start with "how do we use AI?"

Start with: "what are we actually trying to achieve?"

Once that answer is clear, look at your process. Is it ready to be automated? Or do you need to clean up first?

That honest question saves you months of frustration and thousands of euros in failed AI projects.