This is a post about how to use AI to make yourself smarter - not dumber.

Everywhere I look right now someone is either talking about how AI is magically granting everyone super powers or catalyzing society’s downward slide into a post-apocalyptic hellscape reminiscent of the movie Idiocracy. Hyperbole like that may be great for clicks and engagement but reality, as is often the case, is much more nuanced. I spent over 12 years in the manufacturing and consulting industries before transitioning to a Machine Learning Engineer role, and experienced the explosion of public-facing AI tools from a front row seat while completing my Master’s in Applied Data Science at the University of Chicago. AI use is more like a choose your own adventure book, where the small decisions we make gradually accumulate into consequences that are not always immediately apparent. The key is to make thoughtful decisions that leverage AI’s capabilities without neglecting to nurture your own.

Learning Debt

Technical debt is a widely discussed topic in software engineering. Dramatically oversimplified, it refers to compromises that are made under the pressure of time constraints. When suboptimal choices are made in the name of speed, you accumulate debt in the form of future work to untangle the progressively more complicated layers of digital duct tape that hold your system together.

Learning debt works much the same way. As we are presented with challenges throughout our personal, professional, and academic lives we are often given the opportunity to apply a quick fix or to dig in and truly learn how to tackle the task at hand. There is immediate-term benefit to the shortcut (including the use of AI), but in the long run too many of these accumulate into learning debt - skills and knowledge that we need but never fully developed.

While the term “learning debt” is a product of the digital age, the phenomena it describes is timeless. Before Wikipedia, there was CliffsNotes. Before Chat GPT, there was that kid in undergrad who monopolized your professor’s office hours getting the assignment spoon-fed to them question by question. Without a doubt, someone in ancient Mesopotamia abused their abacus to avoid learning multiplication tables. There is nothing new under the sun, and as long as there are tools to make intellectual work easier there will be the temptation to use them to cut corners.

I recently weighed in on a discussion on AI in higher education, and stand by my assertion that there is a huge difference in the final output of someone who leverages AI thoughtfully within their workflow and someone who takes their hand off the wheel completely. The former still requires hard work and will produce a student prepared to thrive in the modern workplace, the latter will produce a student indistinguishable from one that coasted through their education the old fashioned way. The only real difference is that you will now converge on one outcome or the other much faster.

The best response I have ever heard to the ubiquitous “Is AI coming for our jobs?!?” question was from University of Chicago professor Nick Kadochnikov - “Of course AI is going to replace more roles as time goes on, but it will also create new ones. You don’t have to outrun the bear, you just have to outrun the other guy interviewing for your next job.” The knowledge economy will always have jobs for people who are intellectually curious and committed to learning new skills, but carrying too much learning debt will make it tough to put someone else between you and that bear. AI is a powerful tool that can help us manage the challenge of learning new things - but you should leverage it to remove obstacles (such as compiling a reading list or generating a study plan) rather than to circumvent your own critical thinking and decision-making processes.

Delegating vs Outsourcing

Rotational’s CTO, Dr. Rebecca Bilbro, often encourages us to view leveraging AI tools as a form of delegation. As someone who has managed projects, teams, and entire organizational departments this resonates with me. When a (good) leader delegates, they are not abdicating responsibility for the outcome entirely - at the end of the day the buck stops with them, and they still own the outcome if things don’t go well. Delegation is not a leader taking their hand off the steering wheel entirely, it is asking someone else to change the radio station or adjust the air conditioning.

This definition of delegation requires a leader to - at a bare minimum - be engaged enough to evaluate how well the task was done. This level of engagement will lead to an improved understanding of the task, and as a leader delegates and coordinates multiple related tasks they may not become an expert in each individual task, but they do become an expert in how those tasks fit and work together. And over time the best leaders develop a strong understanding of these tasks by observing how their team members complete them.

The same idea applies to the use of AI - if you are delegating tasks to a chatbot, coding assistant, or other agentic AI application, you are not absolved of your responsibility to understand what is going on. That is what causes learning debt to accumulate rapidly, and what will make you as an employee disposable and replaceable.

For example, in my role as an ML engineer, I build functionality around tools and code libraries that are constantly evolving. On any given day I may need to re-learn the newest version of a standard software package, adopt a brand new tool that does the same thing better, or embrace a completely different approach to solving what I thought was a familiar problem.

In any of these situations, I could choose to vibe code my way through the process by writing a lengthy description of what I needed to do, asking a coding assistant to implement it in Python, and then feeding any errors back into the chat until the code “worked” (quotes intentionally sarcastic - there is a very good chance this code would embarrass me when it came time to put it into production).

But as an alternative, I could start with an outline of how I think the application should be structured, write the basic code, ask a coding assistant to look up any specific syntax I’m not sure about (instead of manually searching through pages of documentation myself), and then carefully review the produced code to both learn from and critique it.

Compare: "Generate a python script to connect to a Neo4j database, execute the following queries [insert long list of queries], and export the results to a JSON file."

Against: "Review the file Neo4f_connect.py, where I have outlined a python class to manage my application's connection to a Neo4j database. Identify any missing credentials or other required connection information and add them to the instance variables in the constructor method."

"Now complete the stub for the establish_connection() function to correctly format the credentials according the the API specifications."

"In a new file named "query_examples.md" provide 3-5 examples to illustrate the structure and format of a Neo4j query, along with how to execute them using the python SDK. Include a link to the relevant pages of the official documentation."

Both of these approaches leverage AI to get the job done faster; but one abdicates my responsibility to ensure that the work is done well (if opaquely), while the other saves less time in the immediate term but ensures I get progressively better at my job and continue to develop skills that bring value to my team.

By comparison, a lot of the AI use that is being lambasted around the internet is what I would consider outsourcing. Take the vibe coding example above, where I could have chosen to offload the lion’s share of the critical thinking and creative work onto an AI tool. In this case, I am not engaged in the process the way a manager might oversee a task assigned to a subordinate. Instead, this is more like a manufacturing company that places orders for a subcomponent with an external supplier - you are lobbing a request over the fence, and catching what comes back. Even then, you need to have some system in place to make sure you are recieveing what you asked for, but you no longer own or deeply understand the process around how it was made.

Outsourcing is not by itself an evil thing, and there is often a good business case for it. Developing and maintaining your organization’s core competencies and expertise requires an investment of both financial and human resources, and even the world’s largest companies have areas where they rely on someone else’s specialized capabilities. But as a modern knowledge worker, what are your core competencies? Vibe code to your heart’s content when you are working on a side project or experimenting with throwaway code for a quick internal demo. If, however, you are letting AI auto-pilot you through your work week without learning anything new or honing your skills (technical, problem solving, and critical thinking alike) you will not be the one to outrun the AI bear when it comes for your industry or specialization.

The Human Touch is Your Differentiator

“Commodity solutions yield commodity results” has become something of a professional mantra for me lately (and it seems to resonate with the rest of our team as well). In a job market where the next big technological breakthrough could automate nearly any job in any industry, do you want to risk being a commodity? And let’s make no mistake, as shiny and exciting as they are, these tools are already commodities. Being able to prompt Chat GPT into writing a good email (blog post, essay, etc.) or extract usable code from Copilot or Cursor is becoming the modern day equivalent of the “highly experienced in MS Office suite” resume line from the ‘90s and ‘00s.

So what is your differentiator? It is your professional experience, your domain expertise, and all of the critical thinking skills and technical knowledge you accumulate while thoughtfully engaging with these powerful new tools. By delegating as a rule (and outsourcing where appropriate) you can become that leader who not only understands what everyone on their team does, but can even fill in for them in a pinch. You can become greater than the sum of your AI tools because you know how and when to use them, and when to roll up your sleeves and get dirty yourself.

And what if you don’t know enough about something to delegate the way I described in the coding example? Treat your AI like a free tutor - get it to write code for you, but read it, understand it - heck even ask the AI to explain it to you - and then follow that up with a discussion with a real human expert, or vet it against an established credible source (you can even start by clicking through to the sources the AI tool cites). Don’t know any human experts in AI, software development, or IT? Reach out to someone at Rotational! We love to share our knowledge, and we’d love to learn about your experience.

Note: this blog post was neither delegated nor outsourced - it was written the old-fashioned way, and only benefitted from proofreading and feedback provided by a real human. Because this felt like one of those times to just roll up my sleeves and get dirty.

Now if you’ll excuse me, I need to lace up my running shoes…

Photo Credit: Gemini 2.5 Flash Image