Misusing AI At Work Is Due To Workers That Entail Bad Apples, Bad Cases, And Bad Barrels

AI In The Workplace

The saga of AI adoption is a bumpy road. Companies have been rushing since the advent of ChatGPT in 2022 to adopt the latest generative AI and large language models (LLMs) throughout their organizations. There is an ardent belief that if a company doesn’t move fast enough to use AI, a wily competitor will. An AI-enabled competitor will presumably be a more efficient and effective enterprise, snatching up market share and pulling ahead of non-AI-using firms. You see, no self-respecting CEO wants to fall behind and be accused of having been asleep at the wheel when it comes to implementing AI.

The initial wave of AI adoption was often undertaken in a very haphazard way. Some firms would simply tell employees to go ahead and tap into public-facing LLMs such as ChatGPT, GPT-5, Claude, Grok, Copilot, Gemini, and so on. After a while, top leaders realized that this could create a privacy and confidentiality nightmare for the company, since the AI wasn’t set up specifically to provide those types of protection. Many other issues arose too.

Eventually, firms and AI makers realized that sensible use of AI needed to be conducted on a systematic and planned enterprise basis that would suitably serve the needs of companies. Employees would be issued enterprise-specific logins and credentials to use a designated and secure instance of AI. Leaders were relieved that they no longer had the exposures associated with using public-facing AI.

No Sigh Of Relief

It is not a foregone conclusion that just because employees have access to a company-provided LLM, this will translate into proper and worthwhile use of the AI. Recently, there was a craze over so-called token-maxxing; see my coverage at the link here, whereby employees were trying to maximize how much usage of company-provided AI they could muster. It was crazily assumed that if an employee was racking up massive charges of AI time and usage, they must be doing something productive with it. Not necessarily.

Some employees aimed to perform token-maxxing by instructing AI to repeatedly do the same thing or attempt to do long tasks that had no value to the company. The gutsy aim was to make the AI churn and appear to be doing something worthwhile. Company leaders were certainly culpable in this quagmire. At times, top executives would inspect tallies of AI consumption and then broadcast platitudes to those individuals and teams that seemed to be “maximizing” the use of AI. Nobody looked at what the AI usage accomplished, and only cared that AI usage was trending upward.

The bottom line was that this AI usage had no direct company goal or aim to genuinely aid the company. You might dismiss this as nothing more than a silly exercise. The silliness, though, caused firms to end up with enormous charges for the consumption of AI resources, and had next to nothing of substance in return.

How AI Misuse Arises

Let’s consider the basis of why AI usage goes awry in organizations. In the example of token-maxxing, employees were doing what they thought they ought to do and/or were encouraged by their employer to do. Are you misusing AI at work if you do what your boss or your boss’s boss seems to insist upon? That’s a tough ethical or moral question.

Suppose you realize that an AI-using effort is futile or undermines the company. Should you resist doing what the leadership is telling you to avidly undertake? This can be quite a dilemma for workers. You are presumably to do the thing you are told to do, despite knowing that doing so is either pointless or adversely undercuts the company.

I am going to simplify the twists and turns so that we have two straightforward circumstances:

• (1) Workers involved in unintentional misuse. Here, a worker believes they are properly using AI for the benefit of the organization, but it turns out they are not doing so and have inadvertently fallen into unintentional misuse.
• (2) Workers involved in intentional misuse. A worker is fully aware they intend to misuse AI, mainly hoping not to get caught doing so.

We can perhaps give some latitude to the first case. A worker who sincerely believes they were doing the right thing for the company when using AI receives some grace even if the AI usage turns out to be a misuse. But they don’t get totally off the hook. Meanwhile, the worker with intentional misuse gets no latitude. They are ripping the company off. Period, end of story.

When Workers Go Bad

Shifting gears, let’s momentarily broaden the view about unethical behavior in the workplace. After doing so, we can jump back into an AI context.

A now-classic article on unethical work-related behavior is entitled “Bad Apples, Bad Cases, and Bad Barrels: Meta-Analytic Evidence About Sources of Unethical Decisions at Work” by Jennifer Kish-Gephart, David Harrison, and Linda Treviño, Journal of Applied Psychology, 2010, and made these salient points (excerpts):

• “In this paper, we attempt to provide a clearer empirical and theoretical picture of what we know (and don’t know) about multiple sources of influence on unethical behavior at work.”
• “In this meta-analysis, the authors draw from over 30 years of research and multiple literatures to examine individual (‘bad apple’), moral issue (‘bad case’), and organizational environment (‘bad barrel’) antecedents of unethical choice.”
• “Cumulative data suggest not only multiple sources or facilitators of unethical choice — bad apples, bad cases, and bad barrels — but also the intriguing possibility that these agents work at least sometimes through more impulsive, automatic pathways than through calculated or deliberative ones.”
• “This suggests a need to more strongly consider a new ‘ethical impulse’ perspective in addition to the traditional ‘ethical calculus’ perspective.”

An especially telling aspect highlighted by this research was that workers with particular traits or preponderances are more likely to undertake unethical behavior at work. For example, an employee with a Machiavellian predisposition is likely to engage in untoward workplace behaviors. Also, workers who have a relativistic moral philosophy, ostensibly driven by the circumstances at hand, may engage in unethical behavior and not perceive themselves as carrying out any wrongdoing.

Bad Apples, Bad Cases, Bad Barrels

The crafty imagery of bad apples, bad cases, and bad barrels provides a useful basis for exploring the nature of AI misuse in organizations. I will focus on the instance of workers who have knowing intentions underlying AI misuse.

Let’s illuminate each of the three categories:

• (1) Bad Apples in AI Misuse. Suppose that salesperson Jonathan discovers that several of his customers have not renewed their contracts and aren’t going to do so. To achieve his quarterly quota, he uses generative AI to produce fake renewal documents that look completely realistic and submits them as though they are genuine. His personal dishonesty spurred him to deceive his employer. This is a bad apple.
• (2) Bad Cases in AI Misuse. Samatha gets a call from her manager that she must immediately write and submit a performance review for a subordinate who has been troublesome at work. She uses AI to rapidly generate the performance review. It comes out much harsher than if she had written it herself, but Samatha decides that due to the pressured deadline, she will go ahead and submit it anyway. This is a case where she was semi-trapped in a workplace situation and skirted the moral line when it came to using AI. That’s an example of a bad case of AI misuse.
• (3) Bad Barrels in AI Misuse. Top leadership of a firm decides that the number of reports that each employee turns in every week is an indicator of worker productivity. Workers with the most reports will be put on a fast track for promotion. Employees start using AI to generate the reports, and since the aim is to produce lots of reports, the employees tend not to give the voluminous reports the expected scrutiny. Management ends up with a slew of reports that contain unreliable AI-generated outputs. This is a bad barrel that arose because the company incentivized workers to misuse AI.

An added twist is that the three types of AI misuse patterns can occur at the same time. Imagine a firm where top leadership incentivizes AI misuse (bad barrel) and has situations happening that also instill AI misuse (bad cases). In addition, actual bad apples of employees are misusing AI for various personal gains that are not within either ethical or legal bounds. The advent of bad apples, bad cases, and bad barrels can be pervasive, persistent, and occur simultaneously throughout an enterprise.

Solving The AI Misuse Problem

One path to deal with the woes of AI misuse would be to focus on the human side of the equation. How is it that bad apples are working at the firm? Why do bad cases arise in the organization? Where has leadership missed the boat by incentivizing AI misuse? These are all avenues worthy of apt consideration.

Another angle is that these maladies involve AI as the means or root through which bad behavior can be exercised. If AI were properly governed by the enterprise, sufficient controls and auditing would tend to prevent AI usage from going off the rails. Whether it is the effort of using AI to create fake renewal contracts or allowing employees to willy-nilly generate reports and avoid scrutinizing the reports, these are facets that can be dealt with via proper AI governance.

What has the company done to emphasize that AI usage must entail accuracy, integrity, privacy, safety, and other vital elements? Who in the organization is being held responsible for the responsible use of AI? Are there appropriate means for employees to question or challenge AI practices that seem inappropriate? Not having answers to those questions is a red flag that little or no AI governance is taking place in an organization.

The World We Are In

I mentioned earlier that in the frenetic rush to adopt AI, firms often did so haphazardly. The fad-like urgency was done without thinking through the realities of AI use and potential misuse. Leadership must now open their eyes to the fact that AI needs to be prudently managed and governed. Aim to guide AI usage toward being on the up-and-up. Aim to make AI misuse extremely unlikely or impossible to undertake. Do not let AI become a major vulnerability in the company and become a devilish threat to the enterprise’s survival. Establish a comprehensive AI governance strategy, including a workable set of policies, procedures, and controls, and then make sure it becomes reality.

A final thought for now. Benjamin Franklin famously said that “Idle hands are the devil’s playthings.” If you give employees access to AI and leave them with idle hands, there is a solid chance that devilish inclinations will surface. Be both good and wise about how your enterprise governs AI.