The most questioned and most intriguing question in our industry today is: will AI replace all engineers? Will AI make companies fire most engineers since fewer engineers can do more work?

My answer is most certainly not. Counter to popular fear, companies will want more engineers, not fewer. Here's my line of thought:

Let's say Company X has features and bugs that span 2 quarters ahead. Why is this load calculated to be 2 quarters ahead? Because that's what humans right now can deliver. Why aren't we getting a bigger stream of requests from our customers? Because it takes time to deliver.

What will happen when dev teams deliver faster? Faster iteration on product needs, more delivered features, and more bugs to fix. Simple as that.

Companies are driven by financial measures. Nowadays, most companies struggle to move fast, struggle to pivot or spread to more business opportunities. The hunger is there, but management needs to compromise based on current delivery rates. What I think will happen? A surge for more engineers will emerge.

This pattern isn't new. ATMs actually increased bank teller employment from 500,000 in 1980 to 600,000 in 2000. Why? ATMs reduced costs per branch, allowing banks to open more locations. The same principle applies here - faster development enables more business opportunities, requiring more engineers to capture them.

Research consistently shows that productivity-enhancing technologies expand markets faster than they displace workers. Manufacturing productivity doubled between 1995-2015, yet employment stabilized after initial displacement.

The numbers back this up: software developer employment is projected to grow 17% through 2033, while 87% of companies face current or expected talent shortages.

Now let's talk about what dev teams are struggling with right now, and for all the dev teams that just started or are going to start using AI - let me give you a glimpse.

First and foremost, we have adoption gaps between team members. This is reasonable. Even for companies that invested in learning, giving engineers dedicated tools, instructions, and time to fail and succeed, I still see many engineers who refrain from using AI to assist in their coding. Some refrain totally, and some use it on the low side.

Now the problem arises: the power users of AI-first coding are delivering massive amounts of features and fixing bugs at a rate that makes the other devs unable to cope with the output. Suddenly, if the non-AI coder was doing refactoring for a class for a week, the other dev is refactoring a module or a service. These two personas can't work together all of a sudden since the mission of the non-AI guy is barely a single prompt in the AI dev's PR.

The data supports this observation. 82% of developers use AI coding assistants daily or weekly, but only 6% of engineering leaders report significant productivity boosts despite widespread tool deployment. GitHub research shows developers complete tasks 55% faster with Copilot, creating massive productivity gaps between AI adopters and non-adopters.

Let's talk PRs and code reviews. The current workflow is designed for a slow pace like humans deliver. Meaning that a tech lead or team lead has a steady state of PRs coming in, and the PRs, as we were all used to, are small and manageable for a human to review.

With AI-first coding, PRs are getting big. Doing a PR for small naming changes seems like a waste of time and CI/CD time. Fixing a single bug seems slow since a dev can fix 5 of them in a single PR. Doing a small UI task feels like a small task - let's create a whole new view, create bigger apps, expand our capabilities.

So now engineers, if they are not writing with AI, need AI to assist them to review the code. Teams are experiencing 50% faster code merging but struggling with review processes designed for smaller, human-scale changes.

This workflow transformation mirrors what I explored in my analysis of why open agents outperform rigid workflows - the traditional PR process is a rigid pipeline that can't adapt to AI-generated code volumes.

Next up is coding conventions. I have grown up on SOLID, clean code, design systems, best practices, and so on. Let's start with a bold claim: LLMs have been trained on the entire GitHub open source. You can logically understand that these open source projects have all the perfect conventions in the world. After reviewing tons of AI-generated code, I can assure you that this code is probably high above average.

And after this claim, I want to add my thoughts: what matters most is the ability to deliver. If I'm delivering at a high pace using AI and the AI needs to make adjustments so humans can cope with its delivery rate, we are missing something.

I don't think that enforcing all the things that humans have enforced to make code more readable will make AI deliver faster. For example, AI prefers long files with all the context in one place. Fracturing the files is actually making it slower.

The research backs this up. GitHub data shows AI-generated code improvements in readability (3.62%) and maintainability (2.47%). However, GitClear found a 4x increase in code cloning and duplication when using AI assistants, suggesting we need new conventions optimized for AI-human collaboration rather than purely human workflows.

This paradigm shift reflects what I discussed in The Assembly Code of AI - we're in the "assembly language era" of AI development, where the tools require expert knowledge but are rapidly evolving toward more accessible systems.

Companies are discovering something crucial: McKinsey's research shows top-quartile companies in developer velocity achieve 4-5x higher revenue growth and 55% higher operating margins. This isn't about replacing engineers - it's about enabling them to capture business opportunities faster.

The market pressure is enormous. 65% of enterprises with 500+ employees have at least 10 applications backlogged, with 15% having over 100 applications waiting. Technical job vacancies take 50% longer to fill than other roles.

AI companies received over $100 billion in global VC funding in 2024, while the AI-augmented software engineering market is projected to reach $26.8 billion by 2030. This investment reflects market confidence that AI will expand rather than contract engineering opportunities.

The transformation is happening whether we like it or not. Teams that adapt to AI-first development while maintaining human oversight and judgment will dominate their markets. Those that don't will struggle to keep up with the delivery pace that AI-enabled competitors can achieve.

The question isn't whether AI will replace engineers - it's whether you'll be an engineer who can effectively leverage AI to deliver at the new market pace. Companies will desperately need engineers who can work at AI speed while maintaining the architectural thinking, system design, and business judgment that only humans provide.

The productivity gains from AI aren't just incremental improvements - they're market-expanding capabilities that create entirely new business opportunities. And those opportunities will require more engineers, not fewer, to capture them.

For teams looking to make this transition, I've written practical guides on migrating from traditional tools to AI-first development environments and understanding the architecture behind modern AI coding agents.