03/ AI Motorbike Roadtrip: Budapest

AI Motorbike Tour: Day 3 🥁 Budapest 🇭🇺
An AI take on security, accuracy and cost.

Day 3 was a long one.
So long that I am writing this from my hotel lobby the next morning 😅

It started early from Belgrade, trying to avoid the heat after spending the previous day trying to avoid the rain.

I failed at both.

But the nice part was that I had enough time in the morning to take the smaller roads and cross through a small border point.
One of the last border controls in this part of Europe.

And just like that, I was back in Schengen.

It makes you appreciate how lucky we are to travel so easily nowadays.

But back to serious topics.

Budapest was a busy day.

First, I had the chance to meet Attila Andrejkovics, we talked about AI, GenAI, finance, our stories, and what led each of us into this world. A very insightful conversation.

Then I met Ricardo Bravo de Rueda, and we had a very pragmatic discussion about where AI really stands today:
How companies are setting things up, What is working, What is still unclear.
And where European businesses are right now.

Daniel Oghenederhie, ACCA joined us after that, and we had a great conversation about our stories and what brought us to Hungary. Although Ricardo clearly wins that one, coming from Chile and being in the Hungary for 20 years 😄

And finally, last but definitely not least, I met Nancy Kathini.
Again, one topic came back in every conversation:
AI, yes.
But how?

Because today, the technology is amazing.
But for businesses, there are still real challenges:
1️⃣ Prompt-based agents
2️⃣ Data flows
3️⃣ Token costs
4️⃣ Security
5️⃣ Accuracy

And this is where I believe the next step is very clear.
🧠 Agents should not only be “prompted.”
They should be given proper skills.
Some of those skills should be built with code.

💻 The agent should be close to the data, sometimes local, sometimes on a company server.

🧮 And companies should not always rely on the biggest model for every task.
Small, specialized models can often be faster, cheaper, more secure, and good enough for the right use case.

At Capto, this is exactly the kind of AI infrastructure we are building for finance teams. But more importantly, we are also happy to share openly about it. So if you want to understand what is really happening in production AI, feel free to reach out.
Or contact Peter if you are along my route while I am on the road.

A big thank you to Attila, Ricardo, Daniel and Nancy for the great conversations and the interest.

Now leaving the lobby.
Next direction: Vienna 🇦🇹

Have a great day,
Alexis

—

Originally posted on LinkedIn:

https://lnkd.in/p/e8e92Ngd