When Incentives Go Wrong: What AI Slop Grenades Can Teach Compliance Officers 

I told you so. This situation was entirely foreseeable.  

In April 2025, Shopify CEO Tobi Lütke set off a firestorm in the technology world by announcing that employees requesting additional headcount or resources would first need to demonstrate why AI couldn’t do the work. Using AI was no longer simply encouraged. It was an expectation. 

And guess what happened? A year and a half later, the CEO recently complained that employees were sending what he called “slop grenades” to one another. These were AI-generated emails, code, and other work products that hadn’t been properly reviewed by the people creating them. As a result, colleagues were left with the responsibility of correcting mistakes, revising the work, and trying to figure out what was being communicated.  

Not surprisingly, this also undermined trust between colleagues. 

The incentives were clear. Employees needed to use AI wherever they possibly could. 

People did exactly what they were incentivized to do. They used AI. A lot. Even when the resulting output wasn’t particularly good, created additional work for other people, or would have been better handled by a human in the first place. 

Shopify wasn’t alone. 

At Amazon, an employee-created leaderboard called KiroRank tracked the use of AI tools, encouraging employees to compete based on their AI usage. This led to something called “tokenmaxxing,” where employees reportedly used AI for unnecessary tasks simply to increase their usage statistics.  

Instead of rewarding effective AI use, the leaderboard encouraged greater consumption of AI resources, even when there was little to no business value in doing so. (Read the Business Insider story.) 

Amazon eventually discontinued the leaderboard, emphasizing that employees should be using AI to solve actual business problems rather than simply using it for the sake of using it. 

Many companies have added the use of AI as a factor into bonuses. Not helpful, good, and appropriate use of AI. Just use.

The problem wasn’t AI itself. It was the incentives surrounding it. 

There is an enormous difference between incentivizing employees to use AI intelligently to improve the speed and quality of their work and incentivizing them to use AI as much as possible. 

One rewards results. The other rewards activity. 

As companies grapple with the financial costs of widespread AI adoption, some are introducing token budgets and other controls to encourage more deliberate use of the technology. The conversation is shifting from how much AI employees use to whether that use actually delivers value. 

It’s a lesson in unintended (but entirely predictable!) consequences. It’s also one compliance officers should recognize immediately. 

We need only look back at the Wells Fargo scandal to see just how dramatically incentives can undermine ethical behavior. 

Wells Fargo placed tremendous pressure on employees to sell additional products and open more accounts for existing customers. The drive to increase account numbers created incentives that encouraged employees to open accounts without customer authorization, sometimes using fabricated email addresses or other false information. 

Customers ended up with accounts they hadn’t requested and, in some cases, didn’t even know existed. 

The misconduct ultimately resulted in enormous regulatory penalties, litigation, and extensive regulatory oversight. 

The problem wasn’t that the company wanted to sell more products. The problem was that its incentive structure encouraged employees to achieve sales goals regardless of how they got there. 

Sound familiar? 

Incentives can be incredibly effective in driving behavior. That’s precisely why companies use them. Unfortunately, they can also drive exactly the wrong behavior when they aren’t designed carefully. 

As compliance officers, we need to be looking for these misalignments before they turn into significant problems. 

Start by understanding how people are actually incentivized, particularly within the sales organization. 

What are the targets? How are commissions calculated? What activities are rewarded? Are there accelerators for hitting certain thresholds?  

It’s also important to understand the less formal incentives. What gets someone promoted? What earns recognition from senior leadership? What happens when people don’t meet their numbers? 

The formal commission structure may tell only part of the story. 

You can’t meaningfully evaluate the risks associated with an incentive structure if you don’t understand how it works or what behaviors it is designed to encourage. 

Once you understand the incentives, take the analysis a step further. 

Imagine you’re an employee who is absolutely determined to maximize your compensation or achieve the highest possible performance rating. 

Now imagine that you don’t care about compliance obligations, or that you’re willing to find ways around them. 

What would you do? 

Could you manipulate sales figures? Offer inappropriate discounts? Push products on customers who don’t need them? Misrepresent the capabilities of a product? Circumvent approval processes to close deals more quickly? 

What happens when someone takes the incentive structure to its logical, but unethical, extreme? 

This exercise is valuable because it forces us to consider not simply how an incentive is intended to work, but how it could be exploited. 

Nobody at Shopify wanted employees creating piles of AI-generated work that made their colleagues less productive. And Amazon’s intention wasn’t to encourage people to waste computing resources on unnecessary AI tasks. 

But incentives don’t operate according to intentions. They operate according to what gets rewarded. 

Ideally, compliance should have an opportunity to review and sign off on incentive structures before they’re implemented, particularly in higher-risk areas of the business. 

If that isn’t possible, compliance should at least have a formal opportunity to provide feedback on the potential consequences of the proposed structure. 

And that feedback should go beyond simply identifying the risks. 

Suggest modifications that would make the incentives more likely to encourage compliant behavior. The objective isn’t to prevent the business from incentivizing strong commercial performance. It’s to make sure the incentives don’t inadvertently reward misconduct. 

After all, it’s much easier to modify a commission structure before it’s implemented than to remediate misconduct after hundreds of employees have responded to it. 

No discussion of incentives would be complete without mentioning the importance of incentivizing ethical behavior itself. 

Joe Murphy, the godfathers of the compliance, has written extensively about the role of incentives in effective compliance programs, including his influential white paper, Using Incentives in Your Compliance and Ethics Program. 

The U.S. Federal Sentencing Guidelines specifically address incentives as an element of an effective compliance and ethics program. Section 8B2.1(b)(6) calls for appropriate incentives to encourage compliance with the organization’s program, alongside appropriate disciplinary measures. 

How to design and implement effective compliance incentives is an entire conversation in itself, and one that deserves significant attention. 

However, it’s worth remembering that incentives for ethical behavior shouldn’t exist entirely separately from commercial incentives. 

The lesson from Shopify and Amazon isn’t that companies should stop encouraging employees to use AI. It’s that they need to be careful about what they’re actually rewarding. 

The same is true for sales, productivity, customer acquisition, and virtually every other area of business performance. 

People respond to incentives. Sometimes they respond in ways leadership didn’t anticipate, and sometimes those responses create substantial compliance risks. 

Our job is to anticipate those consequences before they materialize. 

Because whether it’s AI slop grenades or unauthorized customer accounts, the underlying lesson remains the same. If you incentivize the wrong behavior, don’t be surprised when people engage in it.