AI’s True Impact on Coding Skill, Workforce Trends, and Lessons from Tech History
AI will reduce the number of developers most organizations need. But the roles that will be eliminated will primarily come from the bottom, not the top. In many ways, AI is like having a team of enthusiastic assistants; assistants that need code reviews and clear instructions, and who often go off in the weeds. One way to think about it; AI tools tend to operate at about 70% of the skill level of the user. This isn’t a hunch; it’s a practical observation borne out by both modern studies and decades of experience in the trenches.
Framing AI’s Role in Development
Artificial Intelligence has emerged as a transformative force in software development, reshaping how code is written, tested, and documented. While the hype paints AI as an equalizer, enabling less skilled developers (or even non-developers) to keep pace with seasoned engineers, the reality is different. AI is a remarkable force multiplier for those who know what they’re doing, but it isn’t a magic wand for those who don’t. Let’s take a look at what that means, both from first-hand experience and from a broader industry perspective.
AI as a Productivity Enhancer: Skill Amplification, Not Substitution
I would suggest that AI tools operate at about 70% of the skill level of the person using them. This isn’t just anecdotal; it’s reinforced by research. For example, a 2023 Stanford and MIT study found that junior programmers using AI coding assistants like GitHub Copilot could complete tasks faster, but the quality of their work still depended heavily on their own expertise. Senior developers, on the other hand, were able to leverage these tools for even higher productivity while maintaining accuracy and robustness. As Andrej Karpathy puts it, AI tools are “amplifiers”. They boost the user’s underlying skill but never replace the value of careful judgment and deep technical understanding. In practice, AI helps developers write more code, spot errors, and patch gaps in their knowledge. But it always requires oversight; it’s the tireless assistant that, left unchecked, can introduce subtle bugs or architectural missteps that only an experienced eye would catch.
Workforce Impact: Real-World Examples and Trends
The assertion that “AI will reduce the number of developers most organizations need” is already materializing across the tech industry. Companies like IBM and JPMorgan Chase have openly shared how they’re using AI to automate routine coding and maintenance tasks. IBM’s CEO, Arvind Krishna, mentioned in 2023 that the company expects to pause or slow hiring for roles susceptible to AI automation, with thousands of jobs potentially affected over time. McKinsey’s surveys echo this trend, showing that enterprises deploying AI in software development are seeing increased productivity, enabling teams to accomplish more with fewer people.
It’s important to underline that these reductions primarily impact entry-level or routine positions. The high-value work; architecting systems, optimizing performance, and solving complex technical problems still demand expertise. AI isn’t taking away the need for skilled developers; if anything, it’s allowing those with deeper skills to have more impact, while shrinking the necessity for large teams of less experienced hands.
Historical Perspective: Lessons from Past Tech Shifts
This isn’t the first time technology has shifted the ground beneath our feet. When I started as a professional programmer in the mid-1980s, most of my work was in C. There was no internet as we know it today. No Stack Overflow, no Google, no endless YouTube tutorials, no vendor support sites. If you didn’t know something or have a book on your shelf, you had to buy a physical manual, ask a colleague, or figure it out on your own. Frameworks and libraries were rare. You “rolled your own” everything: database, user interface, the lot.
Applications like Lotus123 and dBase III were revolutionary, letting non-technical users build custom solutions. Among those users, there were always a handful with the knack and drive to create surprisingly powerful systems, at least until those systems grew past their skills. I spent a good chunk of the ’80s making a living fixing by finishing half-baked projects that had outgrown their creators’ abilities.
Over time, Lotus123 and dBase III gave way to MS Excel and MS Access. Access, in particular, was infamous for spawning impenetrable, half-completed systems that cost companies far more to migrate or fix than if they’d been built right from the start. I often joke that Access made me more money than any other product, simply by providing a steady stream of projects in need of rescue.
Then came Visual Basic, which made it easy for almost anyone to whip up a UI and bolt some code behind the controls. Customers would show me a shiny VB front end, confidently declare “It’s 90% done!” and ask me to finish it. The reality? Building a robust, performant, multi-user backend is hard. Many didn’t know what they didn’t know, from relational database design to handling race conditions. Each wave of new “easy” tools brought the same cycle: initial empowerment, followed by a reckoning with complexity that demanded professional intervention.
Modern Parallels: Low-Code Platforms and AI Coding Assistants
Fast forward to today and we’re seeing the same patterns with low-code platforms like OutSystems, Mendix, and Microsoft Power Apps, not to mention AI coding assistants such as GitHub Copilot and ChatGPT. These tools let non-developers and junior staff build applications at a pace that would have been unthinkable a generation ago. Gartner predicts that by 2025, 70% of new applications developed by enterprises will leverage low-code or no-code technologies.
Just as with Lotus123 and Visual Basic, these platforms empower users to solve business problems creatively … and sometimes to build solutions that spiral beyond their control. The risk is clear: without foundational knowledge in architecture, security, and maintainability, organizations can end up with costly rewrites and migrations. The lesson from history is as relevant as ever: technology lowers barriers, but it doesn’t erase the need for real expertise.
Impact on the Junior Developer Pipeline
My biggest fear for the youngest generation of programmers about to enter the workforce is that AI will enable a smaller, more experienced staff to function without hiring junior developers. If that trend continues, ten years down the line we could be facing a talent gap with no pool of former juniors to step into mid-level and senior roles as veterans retire or move into management.
Foundational skills still matter. The magic of AI and low-code isn’t in turning a bad programmer into a good one, it’s in making a good programmer faster and more productive. For those just starting out, the path to mastery is unchanged: learn the core concepts, build real projects, make mistakes, and fix them. AI can provide guidance and support, but it can’t replace the value of hard-earned experience.
Navigating Change and Amplifying Human Skill
AI is a powerful addition to the modern developer’s toolkit, but it’s not a shortcut to expertise or a cure all for skill gaps. The real workforce impact is already happening: companies are streamlining teams and using technology to do more with less. Still, the lessons of history are clear; tools may come and go, but the need for deep understanding, adaptability, and experience never fades.
As we embrace AI, let’s make sure we’re also nurturing new talent, investing in foundational learning, and using technology to amplify, not replace, human skill.
