In the late 1950s, Enzo Ferrari had a name for the small British teams that kept beating him. He called them garagistes: upstarts working out of lock-ups, bolting proprietary parts together and daring to race against the grandees. Within a few seasons they were winning championships, and Formula 1 was never the same.
I’ve been thinking about them a lot lately, because loyalty has its own garagistes now. Loyalty program managers, marketing teams and the executives they report to are looking hard at programs that work but bore their members: earn a point, wait, redeem, repeat. They want challenges, streaks, games and surprises layered on top. And instead of waiting a year for a slot on the IT roadmap, they’re sitting down with Cursor, Claude or Lovable and building quest maps, prize wheels and progress bars over a weekend. Some of what they’re making is genuinely clever.
At PUG, we love this. For twenty years, the most interesting ideas in our industry died in budget meetings because the person who had them couldn’t get a development team assigned. That bottleneck is gone. The people with the ideas are finally building, and they’re exactly the people we’ve always wanted to work with.
But it’s worth remembering what made the garagistes possible. They didn’t build their own engines.
The vibe coding trap, loyalty edition
Abul Kalam Azad’s recent Medium piece, The Vibe Coding Trap, is one of the clearest explanations I’ve read of where AI-built software breaks. His argument, in short: AI is excellent at the parts of an app you can see, and quietly dangerous in the parts you can’t, such as database design, concurrency and payment handling. His rule of thumb is to let AI build the presentation and workflow, and to let proven, managed platforms hold identity and financial state.
It’s good advice for any founder. For a loyalty team adding gamification to an existing program, it’s an understatement. In our world, the part you can’t see is most of the product.
In loyalty, almost everything is money
Azad’s sharpest example is a withdrawal button. Two clicks land a few milliseconds apart, both read the same balance, and the customer walks away with twice what they were owed. Every loyalty program has that button. It’s called “claim reward.”
Consider what you’re actually building when you add a game layer to a loyalty program. Points are a financial liability: Starbucks, for example, books deferred revenue for every Star a member earns. Prizes cost real money and come out of a finite inventory. Draws and instant wins are regulated promotions, and a promotion that combines a prize, chance and a purchase requirement can be treated as an illegal lottery under US state law. Game outcomes, leaderboards and streaks all decide who gets something of value.
In a typical app, that’s the financial core. In a gamified loyalty program, it’s nearly the whole thing.
There’s also a difference that makes our version of the trap worse. Users of a contract-review tool stumble onto bugs by accident. Players of a loyalty game go looking for them. Good games invite players to work out the best strategy, and the most engaged members of any rewards program spend real hours on message boards mapping its sweet spots. Bolt a vibe-coded game layer onto a program with real members and real rewards, and you’ve also recruited thousands of highly motivated testers, all hunting for the place where it pays out more than it should.
A third layer: relationship integrity
Azad’s framework protects the data. In loyalty, there’s something else to protect: the relationship.
My colleague Craig Dixon put it well in a planning session this week. The failure that worries him most, when a team bolts a homemade game onto an established program, is a member who feels cheated. A streak that breaks on day 40 through no fault of the player. Odds that feel rigged. A reward that disappears from someone’s wallet. None of those shows up in an error log. They show up in a one-star review, a Reddit thread or a quiet cancellation, and a code fix doesn’t win that member back. Worse, the people affected are your existing members, many of whom have trusted your program for years.
AI can draw a beautiful prize wheel in minutes. It can’t tell you whether the wheel is fair, whether you can afford it at scale, or whether your legal team will let you run it.
Four checks before your game layer reaches members
Azad recommends a short architecture review before launch. Here’s the loyalty version: four questions to answer before a new game layer reaches a single member.
1. The economy check. Do the incentives produce the behavior you actually want, and what does it cost if every player plays perfectly? Generous mechanics compound. On one program, our own team debated giving every member a free play every single day of the year. It sounded harmless until we modeled what those plays meant for prize inventory, cost, and how much each play would still be worth to the player by month six. We capped it, and the program was better for it.
2. The exploit check. Can a player fake a score from a tampered phone, replay a claim, redeem twice or farm extra accounts? Outcomes and balances belong on a server the player can’t touch. And don’t assume the threat only comes from outside: the McDonald’s Monopoly promotion was rigged for more than a decade by the security director at the agency running it, in a scheme worth over $24 million.
3. The compliance check. Is your game one of chance or of skill? Is there a genuine free way to enter? Are the odds disclosed? We’ve worked with a company whose legal team found games of chance especially hard to approve, because every member is a paying subscriber. For another client, we built games of skill specifically to stay clear of contest regulations. Those are design decisions, and they have to be made before anyone writes code.
4. The experience check. Does it feel fair, and does it respect the people who don’t want to play? Someone on a client’s team once asked the best question I’ve heard in a design review: what about the 70-year-old customer who just wants her free coffee? A good program still gives her the coffee. The game is a bonus for those who want it, never a toll booth in front of the reward.
What to build with AI, and what to build on
Azad closes with a guide to what’s safe to vibe-code. Here’s the equivalent for a team adding gamification to an existing loyalty program.
| Build it with AI | Build it on a proven platform |
|---|---|
| Game skins, art and seasonal themes | The game currency ledger and its link to your existing points |
| Quest screens and progress bars | How game outcomes and scores are decided |
| Landing pages and onboarding flows | Prize inventory, odds and fulfilment |
| Mini-game front-ends and animations | Draws, sweepstakes rules and audit trails |
| Copy, notifications and campaign variations | Identity, fraud controls and duplicate-account detection |
| Prototypes to test an idea with real players | Economy tuning based on how players actually behave |
The left column is where your brand gets to be different. The right column is where mistakes cost money, trust or a regulator’s attention.
What this means for loyalty leaders
If you run a loyalty program, your team is going to start adding game layers with AI. Many already have, whether or not it’s on anyone’s roadmap. Banning it would waste the most creative energy your organization has seen in years.
The leadership decision is where to draw the line. Let your people own everything customers see and touch, and let them change it as often as they like. Put everything that holds value, decides winners or carries legal exposure on something built and proven for that job. A broken landing page costs you a day. A broken economy costs you your members’ trust, and that rarely comes back.
Welcome to the garage
This is where I have always worked: helping loyalty and marketing teams turn programs that are easy to run, and just as easy to ignore, into something members actually want to play. The next version of our Picnic platform is built for the way those teams now work. Your team builds the game experience with whatever tools it likes, AI included, and Picnic runs everything of value underneath: the ledger, the game outcomes, the prizes and draws, and the economy design that keeps it all fair and affordable. We’re opening a small beta for loyalty teams already doing this.
The garagistes didn’t win by building everything themselves. They won by building the parts that made them different and buying a proven engine for the rest.
Thanks to Craig Dixon for the garage analogy, and to Abul Kalam Azad for the article that started this conversation.
Steve Bocska is CEO of PUG Interactive [LINK: https://www.puginteractive.com] and has spent more than 25 years in the video game industry, working with Disney, EA, Sega and Ubisoft. He applies AAA game design to loyalty and customer engagement, co-created the gamification curriculum for the Certified Loyalty Marketing Professional (CLMP™) program, and is writing a book on what loyalty can learn from game designers. Loyalty teams building game layers with AI can book a free Game Layer Review.
