A drawer of notebooks, and what came next

If we haven't crossed paths, I've spent about thirty years in car rental, most of it on the distribution and broker side. For a good while now I've been heads-down on something I want to walk you through. It's called niftiOS. It's the working system a rental operation can run on. Quoting, bookings, rates and terms, category mapping, supplier connections, complex pricing, the storefront customers see and the console the team works from. Today, this can be used by brokers or intermediaries, flexible for a car rental company directly (especially independent operators who are a bit fragmented), and others. It works whether you're renting cars, leasing, and other products being added (luxury cars, camper vans, chauffeur services, and more).

It didn't start as any of that, though. It started as a drawer full of notebooks- dozens of them, full of the things you only learn by doing this work for decades. Where quotes go wrong. Why a customer walks, or gets turned away, at the counter. The supplier quirks nobody documents and very few understand. The pricing edge cases that quietly cost real money. I always assumed someone with a proper tech team would eventually build the version I kept sketching - at some point I stopped assuming, and built it myself. This isn't to praise myself for that and it's not an article meant to downplay anything that's been built across the industry - there are some incredibly powerful systems built on years of hard work, and I applaud that to the fullest.

This is really just the honest account of how it all went, what the work looked like, how I used AI to do it, what broke, and what I'd tell anyone in our industry thinking about the same leap.

Now the part that gets a reaction : I built it with AI, mostly Claude, and a great deal of Python. I want to be plain about that, because there's a lot of noise about what "building with AI" means, and most of it is wrong in both directions - it's neither the effortless magic the headlines sell nor the useless toy the skeptics claim. The truth is more useful than either.

Here is what the work actually looked like. I'd describe what I needed in plain language, the way I'd explain it if I was hiring someone to build it. It would draft the code, then the real work started. I'd run it, watch it break, work out why, and go back and forth until it held. Some problems took an hour. Some took days. Supplier XML was its own mountain. Getting rate and availability data to come back clean and consistent, every time, in a form I could trust, took far more patience than I expected. Category mapping, so the car a customer picks is the car they actually get. Terms that read the same at 2pm and 2am. Exchange rates, promo logic, source markets, white label, partner portal, a hundred small things that stay invisible until one of them is wrong in front of a customer. I deployed over WinSCP, watched things fall over in ways I didn't predict, and fixed them one at a time. Each was effectively a module that, combined, created a full platform and booking engine. Each one of those modules was dissected and re-evaluated by me over and over again. I wouldn't consider any one finished until I tested over and over again from the customer's perspective to be sure that I'd accounted for everything I could think of that might increase the risk of friction for the customer. Then I'd do it again. And, again.

The rhythm wasn't gentle. I kept writing in those notebooks the whole way through, because the thinking still happens on paper for me. I'd wake at 2:30 in the morning with a fix or a worry and check my notes to be sure I hadn't lost it. There were sixteen-hour days and stretches with no weekends off, at all. No team to hand things to, no one to call at midnight. There were a couple of nights I was too tired to see the screen, and a few where I was closer to tears than I'd admit. I'm not saying that for sympathy. I'm saying it because the "I built an app with AI over a weekend" stories leave it out, and anyone reading this deserves the real picture. While I consider this a finished product, the truth is, I'm sure there are things I haven't considered and I obsess over it. I'm sure today there will be hurdles I have to overcome. But, at a certain point, you have to trust that it's ready and understand and accept that a finished product is never really totally finished - and so the work continues!

So here's what I actually learned, in case it saves you some time.

The domain knowledge is the whole game. Thirty years of knowing what to build, what "right" looks like, and where it blows up, that's the part AI can't give you right now. It will write code all day long but there are many things it has nearly no knowledge of.

AI shrinks the distance between knowing and doing, not the thinking itself. The gap between my notebooks and a working system was a tech team I didn't have. That gap is smaller now than it's ever been. But it only closes if you already know exactly what you want on the other side.

You have to be able to tell when it's wrong. It is confident even when it's mistaken, so you can't hand it your judgment. If you can't read the result and know it's off, you will ship the wrong thing quickly and confidently. Test everything yourself. Trust nothing until you've watched it work. Over and over and over again.

It's not passive - this was the biggest myth to burn off. You don't set it running and walk away, you steer, constantly. The output is only ever as good as the experience guiding it. The more real world experience the better and, even in my case, I could only build it with the experience I have. Others could view this from another angle or add to it, for sure.

And write things down - the notebooks weren't nostalgia. Half the reason the system holds together is that I'd thought the hard parts through on paper before I ever asked for a line of code.

I'm not writing any of this to sell you something or to pretend that I know better than anyone else. This was a learning experience for me and I'm still learning - all day, every day, it seems. In our corner of the industry, the person across the table today is often a partner tomorrow. I have real respect for the people who keep this business running, at the counter, on the phones, in the fleet yards and back offices and on the trade show floors. And I've always had, and continue to have, extreme respect and admiration for the "technicians" in this field.

If you've built something car rental related yourself, if there's a collaboration worth exploring, or if you just want to compare notes over a coffee or a call, I'd genuinely enjoy that. And if you're sitting on an idea you've carried for years and wondering whether someone like you could actually build it, I'd rather answer that honestly than watch another good idea stay in a drawer. Reach out. I'll tell you what worked and what hurt.

It's the most satisfying thing I've built in a long time, and it's only the beginning of what I believe the potential is.

Brett Gould
Founder, Nifti Mobility Group