Let me tell you a story about recruitment.
Once upon a time, when a job ad was posted, one had to wait a few weeks until there were enough resumes to start the interviews. But over the years, the funnel got filled quicker and quicker. At some point, the number of resumes coming in was so high that recruiters could not handle them manually, so ATS was put on top of it.
The result was immediate – just not entirely as expected. Yes, the filtering was easier, and a lot of manual work was gone. But at the same time, the filtering got more strict and less flexible, so more people got filtered out in the early stage more aggressively. And to increase their chances, they started to send out more resumes responding to more roles.
Not surprisingly, the number of resumes grew even faster for each of the job ads, making ATS… more indispensible. So the rules got even more strict. Now, just like on social media where reaching a human being requires bypassing an artificial algorithmic gatekeeper, you build your resume simply to make an algorithm happy.
So the resumes are no longer built to convey meaning but are built to bypass ATS. The overall quality is lower; the number of resumes is higher. But instead of saying, ‘OK, this system doesn’t work as expected; let’s make a step back and change it,’ we keep throwing more effort into it, creating a vicious cycle.
It’s not the first time it happens in human history. Want to see more examples? Here they are.
Corporate compliance post-Enron: Sarbanes-Oxley created compliance requirements, which created a compliance industry, which created compliance-about-compliance, which created chief compliance officers who need to justify their existence with more compliance. Reverting any of it means admitting the whole stack was a ghost.
Social media content moderation: introducing more rules about what is allowed and what is not creates edge cases, which require rules about them, which then create more edge cases. Nobody is measuring whether the original problem (harmful content) actually improved. Success is measured by whether the process is being followed.
Complex IT environments: when a legacy software or security protocol introduces vulnerabilities, organisations rarely scrap it and redesign the architecture from scratch. Instead, they buy another software tool to monitor the flaws of the first tool and a third tool to aggregate the logs of the first two. Eventually, everyone gets drowned in the tech spaghetti trapped in motion, just keeping the stack from collapsing.
Anti-bribery: When a regulation has a loophole or fails to prevent money from being spent wrongly, policymakers rarely say, “Let’s throw this out and write a simple, clear principle.” They draft a 200-page amendment. Businesses then hire entire compliance departments whose sole job is to navigate the complexity generated by the attempt to fix the previous complexity. Everyone is busy, and billions are still being spent (only elsewhere now).
Urban Planning (Braess’s Paradox): When cities experience traffic congestion, the instinctive response is to add more lanes to the highway. But adding lanes lowers the immediate cost of driving, which encourages more people to get on the road, which leads to even worse traffic jams.
This dynamic is called the ‘cobra effect’. It originates from the British colonial-era policy where bounties were paid for dead cobras to reduce cobra populations. But that did not limit the cobras. People started breeding the poor creatures for the bounty instead.
The real problem behind all of those starts when the metric shifts from ‘Are we solving the problem?’ to ‘Are we doing the process?’ And once you’re measuring process instead of outcomes, you can always show progress. You can always add more process. You can never really fail, because failure would require someone to ask whether the process is the right one. Instead, you can keep pretending that the process is tantamount to progress and, more importantly, that you know what you are doing. And interestingly, organisations and people can survive doing the wrong thing. They can’t survive the admission that they have no idea what to do.
So why is ‘I don’t know what to do’ so much more threatening than just doing the wrong thing?
Although it may look like it, this is not just about admitting that we made a mistake and have to revert the changes. Yes, admitting it would be hard on the ego, but that’s not all there is. It is not even the sunk cost that is the reason here. Sunk cost is about loss aversion over past investment. Yes, it’s not entirely pleasant to untangle a not-working system, but it’s doable… as long as you have a plan.
What I’m describing comes down to action bias combined with what I’d call the illusion of navigation.
The bad solution isn’t being kept because of the effort that went into it. It’s being kept because it answers the question, ‘What are we doing?’ If you kill it, you don’t have an answer anymore, and this is the scary state. When you’re doing something – even something that’s making things worse – you have a heading. You know what you’re doing tomorrow. Undoing it doesn’t return you to neutral; it returns you to the moment before you had a direction. That’s what’s unbearable. Not the wasted effort. The void.
When a system is broken, keeping it running on life support or adding layers of complexity creates an illusion of progress. As long as we are tweaking the AI filters, rewriting regulations, or holding meetings to fix the problem, we feel like we are ‘in the game’. Each new regulation is proof of activity, proof of governance, and proof that someone is steering. Rolling back the regulation means admitting the steering wheel wasn’t connected to anything in the first place.
Stepping back, dismantling everything, and reverting to the starting point requires confronting a specific vacuum. It means admitting, ‘We actually do not know how to handle 10,000 resumes manually, and we no longer have a system to hide behind.’ It is terrifying for organisations to stay in the uncertainty of having no solution and survive the need to invent one with no guarantee that it’s going to work. A bad system provides structure; structure provides a false sense of security.
Consciously choosing to go back to Stage Zero and to stay there for a while is a skill, but it’s also tolerance, often on the individual level. It’s the tolerance of being the person in the room who says, ‘We don’t have a system yet’, while everyone else is waiting for directions. In the time when we are told, ‘Do not bring me a problem, bring me a solution’, being the one who makes everyone stay with the discomfort of not knowing is not easy.
And maybe that’s the real problem – that the reward system in our brains is oriented towards quick closure. So we keep looking for a method to escape the discomfort of having to develop one with no guarantee of success and, therefore, no guarantee of closure. Which, if you think about it, might be exactly what got us to having so many of these kinds of problems in the first place.