AI Can’t Cover Up Your Incompetence
You can’t outsource your lack of skills to AI.
Background
There’s been a lot of discussion about how Gen AI is going to impact the workforce. On one hand, there is a lot of fear mongering on how AI will replace you and your job, and that’s why you need to step up and be AI literate — i.e. exploiting the utility of AI in your role. On the other hand, there’s also a lot of rhetoric on how AI will supercharge our abilities so that everyone becomes a polymath (i.e. a person who is an expert in many different subjects, like Leonardo da Vinci). Neither is going to be true.
Everyone has dreams. Those who dream of becoming a writer see the possibility of achieving it with AI as their co-writing assistant. Those who want to be a marketeer see AI as a way to generate, and even execute, rapid ideas. Suddenly, a whole bunch of folks think they can pitch for work for which they weren’t explicitly trained to do, but feel they can be successful in it with the help of AI.
While AI is many things, it is not a replacement. It is not a replacement for learning and mastery. It is not a replacement for critical thinking and decision-making. It is not a replacement for competency. And so I dedicate my 93rd article to a discourse on why AI won’t cover up your incompetence.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
Long March to Redundancy
In August 2024, I wrote an article about why AI won’t be taking your job … anytime soon. To summarise, if your job is about value-preservation (i.e. keeping the wheels moving, being the friction-reducer in a workflow), then AI will likely displace you at some point. But if your job is about value-creation, then it will be hard for AI to replicate because there is a strong component of “problem-defining” in the work that you do, and AI isn’t equipped to tackle that … yet. The reality is that many knowledge workers aren’t really knowledgeable workers.
While it’s unclear the percentage of knowledge workers who engage primarily in value-preservation work, it is safe to say that general mass adoption of AI by industries and individuals isn’t going to happen in the next 10 years. Theories on adoption curves typically suggest a much longer time frame. The current layoffs aren’t really because of AI redundancies; AI is being used as a PR cover.
Can AI Make You More Competent?
But let’s get to the interesting question: can I use AI to supplement missing skills? The short answer is “no”. A tool cannot make you smarter or more knowledgeable. You can’t use a calculator if you don’t understand Maths. A tool can make you more efficient, i.e. get the target work done faster, or it can help accelerate your learning. A tool doesn’t replace skills or competencies, it simply amplifies it. This efficiency will obviously put pressure on those who are less efficient — both learners and workers, leading to the lower performers being pushed out. So there is a genuine case where AI can lead to higher redundancies in the workforce, but not because jobs are going away, but rather because the amount of incoming work is not keeping pace with the improved efficiency brought upon by AI.
To well-exploit a tool requires a deeper mastery of foundational knowledge. For example, you can’t just execute ML algorithms with a passing knowledge of interpreting results and call yourself a data scientist. While AI can be used for writing programming code, it is often error-laden (studies suggest anywhere from 7–30% of AI-generated codes have serious errors), and those errors are notoriously difficult to detect, and correct. This is because these wrong codes are “intentionally” written to look like right codes. And so the time taken to fix them often negates the efficiency benefits that AI brings.
Now, most programming experts will not accept the output of AI as is. Instead, they will refactor, and add on last-mile integrative codes, or edge smoothing for generated images, or improve code security. Experts use AI to accelerate what they already do; what they already know. Beginners, on the other hand, are using AI to learn how to code, or simply getting the AI to do it “blindly”. Some studies have shown that beginners can typically use AI to get to 70% of the target output, but struggle significantly to close off the remaining 30%.
In the (coming) age of AI, the challenge will be whether workers are prepared to do the heavy lifting in investing time to master foundational knowledge. Great that you want to do vibe coding, but you need to really understand computational concepts. Great that you want to use AI to create a website, but you really need to understand website design principles that are anchored on user experience and behaviour economics. A recent academic paper says it all in its title: “The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers”.
We therefore need to a paradigm change in the way we think about AI, and Gen AI, in particular. It’s not about saving time, but re-allocating time. Instead of the current 90/10 ratio of time spent on computation vs cognitive activities, we want to flip that balance the other way around. Time saved on computation must translate into time spent on generating new knowledge (i.e. value creation).
Conclusion
In November 2023, I wrote a piece entitled “The Age of Cognition” where I argued that broad AI adoption will usher in a different kind of challenge: knowledge workers not knowing what to do with the excess time that is being freed up. HR leaders talk a good talk about how AI will give us time back to do value-added stuff, but the reality is that most will not have “value-added stuff” to do, or won’t even recognise what is truly value-additive. This is underscored by the observations that over the decades, technological innovations have given rise to a class of bullshit workers who believe they can get away with sufficient output with extremely minimal input and effort. It will be no different with AI. But in our march forward, we have an obligation to instead consider how AI can transform knowledge workers into knowledgeable workers.
