What triggers buyer’s remorse? Some people feel it within seconds of driving their brand-new car off the lot. For me, it set in as soon as the honeymoon period with my new flat-screen TV is over, and I realize I shouldn’t have fallen in love with the first one I laid eyes on. We’ve all felt it, but that doesn’t make it less painful.

Fifteen years ago, I invested more than $20,000 in new office software only to watch a completely revamped — and cheaper — release render it totally obsolete just two months later. That still sticks in my memory and makes me cagey about software purchases, especially in a professional setting where buying decisions can have negative career impacts.

The exact dynamics are at play when I talk to customers about AI. There’s plenty of research out there to tell them buying is the right decision — that AI is on the cutting edge of banking technology, and billions are being poured into research annually ($37.5 billion in 2019 alone, according to the International Data Corp.).

But that hype doesn't reassure most buyers; in fact, it makes them more apprehensive. Most can't help but wonder, "Is this AI system going to be obsolete in a year?" Many put off a purchase, figuring a delay will work to their advantage. After all, next year’s version is bound to be better, right?

But that’s the funny thing about AI — waiting is almost always a bad idea. That's because, unlike new cars or flat-screen TVs, AI actually gets more valuable as it ages.

Why the inverse dynamic?

The capabilities of most software are locked in the moment the code is complete. Try doing something the programmers didn't anticipate (like opening an unfamiliar file type), and the results can be ugly — anything from an error message to an application crash. Applications mature, of course, and each new version (usually) works better than the previous one, but you have to constantly play the update game.

AI works differently. It can learn just about anything with few, if any, programmed rules. It can even learn from trial and error the same way humans do.

Let’s suppose you’re part of the KYC team at a bank, solving name-screening alerts generated by onboarding clients, and you work closely with AI to do it. That AI investigates and makes decisions just like the best analysts (except a lot faster, of course, solving more than a million alerts per day). But no analyst is perfect right out of the gate, and, as smart as it is, AI is no different.

The day it started, the AI was solving only around 40% of alerts without help, and, after working at your side for a year, it’s up to 80%. How? AI learns from every single case it sees. The longer it’s on the job, the more it learns, and the better it gets.

To use a non-banking example, think about the recommendations social media makes for you. When I first signed up for Facebook, the paid content it suggested for me was laughably off base (like you, I’ve gotten political ads for politicians I'd never vote for). But, as time went by, the recommendations became more focused, eventually narrowing in on things I actually did find interesting. Sure, there are still some weird items that come up in my social media feed, but now they're the exception rather than the rule.

So, why buy an AI system now rather than waiting two years? Because a system with two years of learning under its belt will do the job far better than a new one that has to learn everything from scratch. It's the same reason someone who's been with your institution for two years works more efficiently than a newly hired employee. Experience matters.

The fact is, the banks that have learned not to think of AI like an old-school software product and regard it, instead, as an extremely skilled and eager-to-learn employee are the only ones truly tapping into the enormous potential of this powerful new technology.

The two big investments banks make in AI are money and time. You might be tempted to wait to make a purchase because you believe it will save money. But, really, you're just losing time.