The Software Engineer's AI Reckoning: From Code Jockey to AI Orchestrator

How AI is Forcing Software Engineers to Rediscover Their True Value

Software engineers or developers as we love to call them, are having a tough year. They are being told to use more AI at work to write code because it will magically make their lives and even themselves better. At the same time, they read wailing predictions made by people that used to pontificate about COVID, economic crisis and Y2K doom and have now focused on all things AI. They hear news, doomscroll layoff updates, and see how assorted coding copilots and AI agents get better writing code.

And you might have noticed something disturbingly unsettling: the thing you've been calling "programming" or "software development" for the past decade or two might not actually be programming at all. It might have just been glorified typing for the most part.

Welcome to the great software engineering identity crisis of 2025, where most developers already use AI tools in their daily workflows, yet somehow most of them are still employed. This paradox reveals something fascinating about what software engineers do versus what we think they do, and more importantly, what developers need to become to survive the AI transformation without ending up as digital fossils or nervous wrecks.

Fear not – we will dive into the realities and how to win this round of innovation. Remember to read my post on EPIC cycle https://olegov.substack.com/p/the-epic-cycle-when-winnie-the-pooh and on why most large tech companies are in the state that they are at https://olegov.substack.com/p/the-secret-behind-ai-layoffs-when

The great coding delusion

Let's start with some interesting factoids. Research consistently shows developers spend surprisingly little time actually writing code, ranging from 9% to 24% depending on the study [2], with the rest involving creating software designs, writing and running tests, debugging issues, and collaborating with stakeholders. These are activities that require human judgment, creativity, and communication skills.

This statistic should be liberating, but instead it's unsettling most engineers who've built their entire professional identity around the myth of the "10x programmer" who can bang out elegant algorithms at superhuman speed. Turns out, the real 10x programmer was always the one who could understand business requirements, design systems that won't collapse under pressure, and communicate technical concepts to humans who think "API" is a type of beer.

AI has essentially called this bluff by automating “code typing”. Substantial portions of code are now AI-generated, but skilled developers and architects remain in high demand. It is becoming obvious that the industry has been conflating "typing code" with "software engineering" for far too long. Recent empirical research reveals a counterintuitive finding: experienced developers using AI tools take 19% longer to complete tasks than those working without AI assistance, despite developers expecting a 20% speedup [1]. This behavior is typical for immature industries that require highly specialized people with non-commoditized skills, i.e., software engineers that know syntax of specific language and the infrastructure of specific platforms. Just like car engine machinists in the past before they mostly got replaced by automated assembly lines.

Effect of maturing industry

The data paints a picture that should make every software engineer reconsider the approach to career strategy. I observe the majority of job opportunity declines are affecting frontend engineers more than backend engineers, not because frontend is easier, but because frontend work often involves more predictable, pattern-based coding that AI can handle effectively. At the end of the day, most people want pretty much the same things when it comes to user experience.

Meanwhile, I observe significant increases in software engineering hiring in investment banking and industrial automation sectors. Translation: companies that need engineers to solve complex, domain-specific problems are hiring aggressively, while those that need people to implement standard web interfaces discover they can get by with fewer humans. Large ISV companies also have lots of people in rather code typing roles, plus the pressure of the COVID over-hiring glut.

This isn't the democratization of programming that Silicon Valley promised. It's the professionalization of software engineering, and it's long overdue. Our industry is maturing from an artisanal craft requiring creative problem-solvers to a structured discipline with specialized roles, standardized processes, and clear hierarchies. We will gradually move away from suppressed art majors being software developers. It will become just as all other mature industries, you get highly specialized education and training, then you can get a very specific job.

The industry will stratify into three distinct strata of software engineering roles:

AI orchestrators (the new elite) These are senior engineers who design systems, manage AI workflows, and make architectural decisions. They don't write much code anymore, they direct AI to write it while focusing on the hard problems of scalability, security, and system integration.

Specialized domain experts (safe harbor) Engineers with deep knowledge in specific industries, fintech, healthcare, and manufacturing. They understand both technical requirements and business constraints. AI might generate code based on their directives, but it can't replicate their understanding of why a banking system needs to handle exactly 4,237 edge cases or how medical device software must comply with FDA regulations. Those are people with the "hunch". They can troubleshoot non-trivial problems, predict impact of new conditions etc.

AI-augmented implementers (the endangered middle) Developers who work closely with AI tools for rapid prototyping and implementation. This tier offers moderate value but faces the highest risk as AI capabilities improve. The path forward requires either moving up to orchestration or sideways to specialization.

Software engineers who want to remain relevant need to develop capabilities that complement and use rather than compete with AI:

Comprehensive system architecture and design: Complex system design remains beyond current AI capabilities. Understanding how to build systems that scale, fail gracefully, and integrate with existing infrastructure requires human judgment and experience. Having a 360 view of an entire ecosystem in which specific software will function becomes more valued. The premium will be placed on the ability to create unique value propositions, because AI will catch up with standardized solutions very fast.

Cross-functional leadership: Managing AI-human hybrid teams, translating between technical and business stakeholders, and coordinating complex projects across multiple disciplines. The mythical "technical lead who can actually communicate" is becoming the industry's most valuable asset. This skill is also applicable to the future of Product or Program managers – see my other essay on the future of PM at https://olegov.substack.com/p/product-manager-transformation-ai.

Domain expertise: Deep understanding of specific business contexts that AI lacks. A healthcare software engineer who understands HIPAA compliance, medical workflows, and clinical decision-making processes provides value that generic AI cannot replicate. As industry matures, it will follow the pattern of other industries. For example, chemical engineers do not get to work at Boeing fuselage design, however, to this day software engineers can easily hop across domains.

Ethical and legal AI oversight: As AI becomes more prevalent in software systems, companies need engineers who can audit AI-generated code, understand bias implications, and ensure responsible AI implementation. Making sure the code will not land a company in hot legal waters will become super-valuable and important. Lawyers and paralegals can't do that; you need to understand the code.

Adapt you must

The evidence suggests we're not heading toward mass developer unemployment, but rather selective evolution. Industry projections indicate software development roles will continue growing through 2033, adding substantial new jobs [3]. However, the nature of these jobs is changing dramatically.

Companies increasingly want "AI multipliers" i.e., engineers who can leverage AI tools to build more sophisticated systems faster while providing the human judgment, creativity, and domain expertise that AI currently lacks. The key insight: software engineers who know how to use AI well will replace those who don't.

Warning, there are two trends that will disturb this equilibrium, exponential improvements of LLM, including reasoning, specialization etc., and standardization of software. In other words, most algorithms will be standardized so AI can be productive. Stay ahead of the curve, be creative inventor, not syntax guru.

The entry barrier

Perhaps the most significant change is the rising barrier to entry for new software engineers. The traditional route of learning to code, building some projects, and landing a junior developer role is becoming obsolete. Junior positions will be disappearing as AI handles routine coding tasks that were traditionally assigned to new graduates.

This creates a catch-22: you need experience to get hired, but you can't get experience without being hired. The new pathway likely involves demonstrating ability to work with AI tools, understanding system design principles, and developing specialized domain knowledge before entering the job market.

This is the effect of maturing industry. Getting a first job for software engineers will be as hard as for anyone else in this professional stratum. The days of hiring everyone with a pulse and passing awareness of programming language syntax by busload for six-figure salary are gone.

Back to the future

Despite fears of AI-driven wage depression, software engineers who successfully adapt may see compensation increases rather than decreases. Current market data shows median software engineer salaries remain strong, with elite technology companies offering significantly higher compensation packages. As the industry consolidates around higher-value roles, successful engineers may find themselves in higher demand, not lower. And yes, there will be bizarre compensation packages given to engineers, but that is rather vanity trip for executives than an objective trend.

The key is understanding that compensation will increasingly correlate with AI leverage capability rather than raw coding speed or syntax knowledge. Engineers who can accomplish 10x more with AI assistance will command premium salaries, while those who view AI as a threat will find themselves competing for a shrinking pool of traditional coding jobs.

Engineers must realize that AI is not magic or silver bullet. It can't do your job for you. It can greatly improve productivity when properly understood and applied as a sophisticated productivity multiplier.

The software engineering profession is experiencing its most significant transformation since the shift from mainframes to personal computers. It's a turn of evolution wheel that's forcing the industry to mature beyond the "code jockey" stereotype and embrace the complex, multidisciplinary nature of building software systems.

The task is to embrace this transformation quickly enough to shape what comes next.

References

1. Becker, J., et al. (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity. METR Research Study. https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/

2. Meyer, A. N., et al. (2019). Today was a Good Day: The Daily Life of Software Developers. IEEE Transactions on Software Engineering, 45(12), 1179-1204.

3. JetBrains. (2024). State of Developer Ecosystem 2024. Survey of 23,262 Developers. https://www.jetbrains.com/lp/devecosystem-2024/

This analysis represents observations from current industry data and employment trends. While AI's impact on software engineering is accelerating, individual career outcomes will vary based on adaptation strategies and specialization choices.