Case study — one hour with AI saved me €1000 a year

This is not theory. It’s a continuation of a series of real cases. In the first article I described how AI helped me close a case with Polish debt collectors. Here — how I reduced my monthly car insurance from €140 to €60. That’s about €1000 a year on one car.

How it started

I bought a car in the Netherlands. A standard passenger car under 2 liters. The insurer quoted €140 per month.

Why so much? “That’s the standard rate for foreigners. Everyone gets this.” You’re new in the country, you have no local history, we have no data — pay like a beginner.

I knew what Dutch drivers with similar experience pay. Around €40 per month. So I was overpaying €100 every month. Over a year that’s €1200. Over five years — €6000. Serious money for “because you’re a foreigner”.

It bothered me.

First attempt — on my own

I googled. Asked friends. Wrote to the insurer asking them to review it. I also threw a few questions into AI — without context, just “my insurance is expensive, how do I reduce it”.

Everywhere the answer was the same: “This is for all foreigners, you don’t have history in the country, you can’t do anything.”

And again I thought — stop. That can’t be right. Twenty years of driving. No serious incidents. And they tell me “you can’t do anything”?

So I’m asking the wrong question. I need a different approach.

Second attempt — a one-hour conversation

I sat down and spent an hour telling AI my whole story. Not “help me lower insurance”, but:

who I am and where I moved from
when I got my license, when I replaced it, which license versions I currently have
how many cars I had and who they were registered to
which policies I bought before, whether there were claims, whether there were accidents
how exactly I bought this specific car in the Netherlands, which documents I submitted and which I didn’t
what the insurer already refused and on what arguments

This is not a regular AI prompt. It’s an interview. I spoke, it asked follow-up questions, clarified details, and nudged me with the right prompts.

After an hour it knew my situation better than I did.

What AI found immediately

First issue — the license date.

I have a Ukrainian license. I got it in 2004. In 2017 I exchanged the paper card for a plastic one. The plastic shows the date 2017.

The insurer looked at that date and assumed I’ve been driving for 9 years, not 22.

This isn’t “evil” — it’s a formal process: what’s in the document is what counts. I didn’t even notice it.

Second issue — insurance history on paper.

AI asked: “Do you have official proof that you’ve been insured as a driver all these years?”

And I realized the real problem.

Over 22 years I had many cars. But they were registered to different people and entities — partly to family, partly to friends, some were work cars, some rentals, some leases. I drove, I managed, I maintained. Legally, the policyholder was someone else.

My “history” was spread across a dozen names. Policies were in my name only when I personally paid for them. As a result, my official insurance history is about 1.5 years — on paper, for my whole life.

That’s a painful discovery. But AI said: that’s not a failure. That’s simply the baseline we work from.

How AI gathered evidence

First — to figure out what can actually be confirmed with documents.

AI itself found the official site where I could request a certificate about my insurance history from Ukraine. It told me where to write, what data is needed, and what format to use. I submitted the request and received the certificate. Yes, 1.5 years. At least it’s official.

Then — workarounds to demonstrate the real driving experience:

**Photos of my original 2004 license.** I still had it at home. AI told me to photograph it and attach it to the email. Proof that I’ve been driving since 2004, not 2017.
**Proof that 2017 was an exchange, not a first issue.** On the back of the plastic license, in small print, there is the original category issue date. I didn’t pay attention. AI did.
**No-claims / no payouts.** I never received insurance payouts. AI suggested emphasizing this explicitly — it matters in the premium formula.
**Careful framing about family use of cars.** AI formulated it delicately — without any lies, but focusing on actual driving experience, not only the formal “policyholder” label.

What went to the insurer

AI drafted a structured letter with four blocks of evidence and references to specific attached documents.

I verified it. I don’t speak Dutch, so I asked AI to explain each paragraph in Ukrainian. I checked the meaning. Agreed. Sent it.

A week later — the reply: €140 → €100.

That’s already €40/month saved, or €480/year. Most people would stop there.

I didn’t.

The second letter

AI suggested adding one more document — a clear photo of my original paper license from 2004 with the issue date visible. Separately emphasize that the plastic license was an exchange, not a new license. And separately add the document about no payouts across the years.

Second letter. Another week — another reply.

€100 → €60.

That’s the final number. €60/month instead of €140. Savings of €80/month, or about €1000 per year.

How much time it took

That first one-hour conversation with AI was the key. It’s not “Q&A” — it’s transferring full context.

After that — 15–20 minutes per letter: verify, add documents, send.

Total — about 3–4 hours of active work over a week. For €1000/year saved.

If you calculate five years of owning the car — that’s €5000 for 4 hours of work.

The main lesson — context decides everything

In the first attempt, AI didn’t help me. Why?

Because I gave it fragments: “insurance is expensive, how do I reduce it”. It gave generic advice — google it, write a letter, check official tables. Like a search engine.

In the second attempt, I spent an hour on full context: biography, documents, car history, license history, previous refusals. Everything that could matter.

And AI became useful. Because it saw what I didn’t see.

The 2017 date. Policies under family names. No-claims history. A detail in small print on the license back. That’s not magic — it’s careful analysis. And careful analysis is impossible without data.

If you give AI half the story — it will produce half a solution.

If you hide things (“it doesn’t matter”, “it’s unrelated”) it won’t see the connections. It can even lead you to an outcome where your payments go up. I’ve seen that happen.

Don’t hide details from AI. It’s not a social network. It’s for your result.

You can’t change the past — but you can describe it correctly

When AI found that my official history is 1.5 years, my first reaction was: “Damn.” For twenty years I thought I had strong experience — but on paper there’s almost nothing.

What do you do with that? Be ashamed? Make excuses? Lie to the insurer about policies that never existed?

No. You can’t change it.

Twenty years already happened. Cars were registered the way they were — under family, friends, business, rentals, leases. That’s history.

What’s left is to describe the situation correctly today. From what exists. Without hiding, without shame, without trying to “polish” the story.

And it works the same way in business.

If you’ve been running a business for 20 years, you have many things you’d do differently now. Contracts registered to the wrong entity. Tax shortcuts that look weird today. Documents that don’t exist because “we didn’t keep them”. Clients where it ended badly.

It already happened. As a consultant, I can’t rewind time. AI can’t. Nobody can.

What we can do is start from the real current situation — not an ideal one, not the one you wish you had, but the one you actually have.

That’s why when a client comes to me — or to AI — and starts withholding, decorating, hiding uncomfortable details, I ask the same thing: tell it as it is. Not how it looks on the website, not how you explain it to partners, not how you justify it to yourself.

As it is.

Then you can work. Then you can find solutions from a real point A — not a fictional one. That’s the difference between real consulting and just “a conversation”.

In my insurance case, I told AI everything. That I have little formal history. That documents aren’t perfectly “clean”. That I already tried and got refused. And only after that did AI find a real path forward.

Don’t hide from a consultant what you can’t change anyway. It’s not a diagnosis. It’s the baseline.

Human in the loop — as always

I wrote about this in detail in the previous article. Here’s the short version:

AI drafts. I send.
AI suggests. I verify and decide.
I don’t write in Dutch — I ask AI to explain in Ukrainian what it wrote.
I double-check numbers and dates.

If someone says “AI did it for me” — that’s not true. AI helped me do it. Those are different things.

What if you have a business?

Everything I described here is personal. My car, my data, my risk, my savings.

In business the approach is fundamentally different — there’s client data, contracts, confidential information. I wrote about that in the previous article. The short version: for business you need isolated environments, DPAs with vendors, and access control. Not just “connect AI to work email”.

In the next articles I’ll describe which processes I consider safe for AI use in business.

What to take away

If you have a situation where:

you pay more than locals (insurance, bank, rent, contract)
you’re told “it’s like this for all foreigners” or “it’s a standard procedure”
you already received a generic refusal
you have real facts that support your case — but they are spread across documents and countries

Don’t stop after the first refusal. Give AI an hour of your time. Tell it everything. Show the documents. Let it ask questions.

There’s a good chance the solution exists — it’s just not visible at first glance.

This is a consultant’s work

I help people and small businesses navigate situations where AI can genuinely save time and money. I show where it helps, where you need more caution, and how to work safely.

For personal situations — fast results.

For business — a separate approach with data isolation and normal security hygiene.

Review the situation →