Another adventure with Frenchie, my AI agent. This time, instead of fixing servers, he was helping with our European holiday.
You know how sometimes you ask AI to do something simple, and it ends up doing something you never expected?
Well, this is one of those stories.
And it started with my wife asking me to set an alarm.
The joys of travelling through Europe
We were travelling around Europe, moving between countries, enjoying the sights, and generally having a fantastic time.
There was just one small logistical challenge.
Time zones.
When you're constantly travelling between countries, trying to remember when tickets go on sale in another country gets surprisingly annoying.
My wife wanted us to visit Tonnara di Scopello in Sicily, a beautiful coastal location with some rather spectacular views.
She knew that tickets were released 15 days in advance and asked me to set an alarm so we wouldn't miss out.
Simple enough, right?
Except I was already trying to remember which country we were in, which time zone we were using, and whether I'd need to set the alarm for today, tomorrow, or yesterday.
So I did what any reasonable person with a slightly overengineered AI setup would do.
I delegated it.
Enter Frenchie
If you've read my previous article about the PHP vulnerability, you've already met Frenchie.
For everyone else, Frenchie is one of my AI agents, named after the character from the TV series The Boys.
He's the resourceful technical genius who somehow manages to fix just about anything.
My version of Frenchie normally looks after Linux servers, updates, monitoring and technical troubleshooting.
But this time, I had a slightly different assignment.
I sent him the booking link and asked:
"They release the tickets 15 days in advance. Can you monitor this for me and tell me when tickets for 3rd of August are available?"
Frenchie investigated the website, worked out when the tickets should become available, and created an automated check every two hours.
Done.
No alarms. No time-zone calculations. No remembering to check a booking calendar while enjoying our holiday.
I went back to being a tourist.
Or so I thought.
"Mon ami, the tickets are available!"
A couple of days later, Frenchie messaged me.
Apparently, tickets for 3 August were available.
Excellent!
Except when I opened the website, I couldn't see them.
The calendar wasn't showing the date I needed.
So I replied:
"I cannot see 3rd of August."
Now, this is where I think the story gets interesting.
A basic reminder would have completed its job. It would have told me the tickets were available, and the rest would have been my problem.
But Frenchie wasn't just running a reminder.
He had actually been checking the booking system.
And when I challenged his result, he didn't simply repeat his previous answer.
He went back to investigate.
The website said no. The booking system said yes.
Frenchie inspected the booking system more closely.
He discovered that the website's backend was returning availability for 3 August, even though the public-facing calendar wasn't displaying it.
The booking system was effectively saying:
Yes, tickets are available.
While the website was saying:
Sorry, nothing to see here.
Frenchie confirmed that the booking system was returning ticket availability and pricing information.
But there was another question.
Could we actually book them?
Naturally, I asked:
"Can we book it through the backend?"
And this is another thing I appreciated about his response.
He didn't pretend he could do everything.
He explained that he could inspect the availability, identify the ticket products, and understand parts of the booking process.
However, the final payment and confirmation involved browser-based checkout and potentially additional payment authentication.
So a fully automated booking wasn't something he could confidently guarantee.
Fair enough.
But he wasn't finished.
He found another way
After investigating the booking process further, Frenchie discovered something useful.
The website accepted a direct booking URL containing the requested dates.
Instead of relying on the public calendar to expose 3 August, he constructed and verified a link that opened the booking page with those dates already selected.
In his technical breakdown, this was the third part of his findings:
"Direct URLs I found — Human-friendly date-prefill URL."
Not exactly the most exciting sentence in the world.
But that link was gold.
I opened it, and it worked.
I was able to book our tickets a full day before the date became available through the normal website calendar.
No special access. No bypassing payment. I simply followed a working booking link and completed the normal reservation process.
And all because I'd originally asked an AI agent to remind me when tickets went on sale.
The interesting part wasn't getting the tickets early
Don't get me wrong. Getting the tickets was fantastic.
And, more importantly, I could report back to my wife that I'd sorted them.
Always a win.
But what impressed me was everything that happened between my initial request and the final result.
Think about it.
I asked Frenchie to set a reminder.
He created a recurring monitoring task.
He detected availability and notified me.
When I couldn't see the tickets, he investigated the discrepancy.
He discovered that the booking system and public calendar weren't presenting the same information.
He explored the booking process, recognised the limitations around payment, and found a legitimate link that allowed me to complete the booking myself.
What began as a reminder became a small investigation, followed by a practical solution.
And I barely had to do anything beyond asking questions and following the final link.
This is why I'm so interested in AI agents
There's a lot of conversation about AI these days.
Which model is best? Which chatbot writes better emails? Which AI can generate the most impressive images?
All interesting questions.
But for me, the real opportunity lies somewhere else.
What happens when AI stops simply responding to questions and starts taking responsibility for getting things done?
Not independently making every decision.
Not being given unrestricted access to everything.
But having the right tools and permissions to monitor information, investigate problems, work through unexpected complications, and help people complete actual tasks.
In this case, the stakes were relatively low. We were booking tickets for a holiday.
But imagine applying the same approach to business processes.
An AI agent monitoring supplier availability.
Another identifying missing information in customer orders.
One checking whether a critical document has been received.
Or an agent noticing that two systems are reporting different figures and investigating why.
The interesting part isn't necessarily the automation itself.
It's the ability to recognise when something doesn't add up, investigate further, and help find a solution.
That's the difference between a chatbot and a genuinely useful AI agent.
And Frenchie?
Well, he went straight back to looking after my servers.
Although I'm beginning to think he might have a future as a travel agent.
Just don't tell him.
He's already French. The attitude is built in.
At DCX, I'm exploring how businesses can move beyond simply using AI tools and start integrating intelligent agents into everyday operations. Sometimes the best examples of what's possible come from the most unexpected places — even a holiday booking in Sicily.
