The Lieutenant Goodbody Effect
When Lieutenant Ernest Goodbody, in the satirical novel How I Won the War, received orders from British headquarters to collect stool samples from North African locals for military medical research, he didn't hesitate for a second. World War II was raging, but the order was clear: the army needed excrement. And Goodbody, as a model bureaucrat in uniform, immediately instituted tracking sheets, quotas, and strict deadlines.
The result? Absolute, 100% bullshit.
Naturally, the locals refused to cooperate with the British army on such a delicate matter. Goodbody's men, exhausted by the meaningless task, did what any reasonable subordinate backed into a corner by a stupid metric would do: they filled the test tubes with plain mud and animal dung. On paper, everything was 110% fulfilled. The generals in London received their charts, the soldiers were left in peace, and medical research gained a ton of worthless mud.
When corporate management today ponders how to motivate employees to use artificial intelligence, they very often follow right in Goodbody's footsteps.
Bonuses are introduced for "number of prompts," rewards for "hours spent in ChatGPT," or financial perks for running every email through AI. The result of such distorted motivation is structurally identical to the outcome of Goodbody's operation: employees will fill those test tubes for you. But instead of genuine productivity and innovation, you will get the digital equivalent of North African mud—mounds of artificially generated, unread text and unnecessary code fed into the system solely so that bonuses hit bank accounts at the end of the month.
Because when you start rewarding people for activity instead of actual value, artificial intelligence will certainly generate plenty of stuff, but the end result will simply be... garbage.
This is the second article in a series on what a company needs to solve so that AI works in daily operations, not just in a pilot project. In the first article, we looked at why roles need to be adjusted. Now we look at an issue companies address even less, yet one that is at least as decisive: compensation.
Start Thinking About Compensation Differently
Companies invest in AI and struggle with how to get people to actually start using it. Fairly quickly, the question arises: Shouldn't we reward AI usage in some way?
Number of prompts. Active Copilot usage. Number of created use cases. AI targets in personal evaluations. Here, I would be extremely cautious.
If a company already has fixed and variable compensation tied to results, it doesn't need to invent a new reward system just for AI. It needs to check whether the current system rewards the performance and behavior the company will need under the new way of working.
AI changes work. And changed work must be reflected in goals and compensation.
1. Start with What You Pay For Today
Every compensation system creates priorities. We reward a salesperson for revenue, margin, or new customers. A manager for team results. A project manager for project completion. In some places we pay for individual performance, in others for team performance.
Therefore, the first question when transitioning to AI is: What behaviors and results does our compensation system support today, and do they match what we will need from people going forward?
If yes, there is no reason to overhaul the system. If roles, processes, and expected output change, that change must also be reflected in the goals by which we evaluate and reward people.
2. Define the Business Problem, Not the AI Usage
Consider HR. A company can give an employee a goal:
"Create three AI use cases and regularly use Copilot."
Or:
"Reduce new employee onboarding time by 30% while maintaining or increasing quality. Identify where AI can assist in the process."
The second assignment triggers a completely different type of behavior. The employee must understand the problem, describe the current state, find opportunities for change, experiment, measure the result, and potentially implement the new process. AI is a tool.
The same principle applies to sales, finance, marketing, or operations: shorten proposal preparation, improve forecast quality, reduce administrative overhead, accelerate customer request processing, increase conversion rates. These are goals that carry actual business value.
3. Put Change Into Existing Variable Pay
If an employee has a portion of their pay tied to performance, AI doesn't need its own separate bonus category.
For example, one of their targets might read:
By the end of the year, redesign a specific process using AI to achieve a pre-defined improvement in time, cost, or quality.
Or the company can establish a project bonus for solving a specific business problem.
The principle remains the same: rewards are tied to the result of the change, not the consumption of technology.
If you reward a salesperson today for revenue, don't start paying them for using Copilot. If Copilot helps them sell better, reflect that in expected performance. If they need to build a new way of working, set a specific project goal.
4. Measure Baseline, Outcome, and ROI
For such a goal to work, it must be measurable. Return to the onboarding example: today, onboarding takes HR 20 hours of labor per employee. The target is to reach 12 hours while maintaining quality. We have a baseline. We have an expected outcome. We can calculate saved capacity and weigh it against the cost of the change. Only then does it make sense to discuss a reward.
The same applies to processing speed, error rates, conversion rates, margins, output quality, or customer satisfaction. And by the way: an experiment that proves a proposed AI use case doesn't work also holds value. If the experiment was well-designed, you get data, and the company avoids investing in a solution that yields no expected value. Important organizational learning has occurred.
Rewarding only "successful" experiments quickly teaches people to inflate or fabricate results.
5. Don't Turn AI Usage into a Performance Metric
Data on AI usage has its place. A company needs to know how many people use the tool, what they use it for, and where adoption stalls. It is an important diagnostic metric for change. For compensation, however, it is a very dangerous shortcut.
Do Not Reward AI Usage
People adapted to what the company measured. And this is where the old lesson of compensation holds true: you get exactly what you pay for. If you pay for prompts, you get prompts. If you pay for created value, you give people a reason to find ways to create it.
AI Doesn't Need Its Own Reward System
When transitioning a company to AI, I recommend reviewing compensation. I would simply guide that review with a different question than how to motivate people to use AI.
I would ask five questions:
What results and behaviors do we reward people for today?
What is changing in their work with the advent of AI?
What specific business problems do we want to solve using AI?
How do we translate that change into a measurable individual, team, or project target?
Can we tie rewards to created value and learning instead of technology usage?
AI is another tool that changes the way we work.
A company does not need to invent a new motivation system because of it. It needs to update its current system based on the work, behavior, and output it now expects from its people.
Track activity. Manage the shift in work. Reward the outcome.
This text is the second in a series on what a company needs to solve to make AI work in practice: roles, compensation, processes, and leadership. It draws on publicly available research and Kogi's field experience.