The latest era of software development explores the potential of AI to shake up DevSecOps processes and software output. Undoubtedly, AI is already enhancing productivity, accelerating development, and, in some cases, improving quality. But despite being hailed as the holy grail for improvement in the software industry, automation seems to shift and reconfigure challenges, rather than eliminating them.

It’s becoming clear that just as AI may enhance productivity, the underlying system complexity remains, often in new and less understood ways. Human expertise, collaboration, and judgement will remain essential to building great software. For this reason, AI will not eliminate the need for experienced software craftspeople.

In the history of software engineering, every major advancement – from increasingly higher-level programming languages to the commoditisation of infrastructure with public cloud, and the progressively sophisticated tooling – has all promised to make building better software easier. Despite this, and even in the rise of AI, we find that complexity in development hasn’t disappeared; it has simply evolved.

The illusion of simplicity

First, we should address the illusion of simplicity that’s emerged around AI. Technological advances such as this have historically been marketed as simplifiers of engineering. For instance, higher-level programming languages abstracted further and further away from machine code, yet software didn’t become simple, it grew more powerful. Similarly, cloud computing relieved engineers from maintaining bare metal, yet software architectures inevitably grew more distributed, requiring more advanced expertise in networking, security, observability, and cost management.

Today, AI is being heralded as the next great simplifier. Tools like GitHub Copilot and ChatGPT can generate code, suggest fixes, and even refactor entire functions. But if history has ever taught us anything, it’s that automation shifts and reconfigures challenges, rather than eliminating them. AI may enhance productivity, but the underlying system complexity remains, often in new and less understood ways.

In an age where viral ideas spread rapidly, executives and industry leaders can all too easily be swept up in AI enthusiasm. However, it’s important to approach AI’s role in software engineering with both ambition and realism, as doing AI ‘right’ often creates more work for humans.

AI as a multiplier

Instead, AI is set to transform software engineering by amplifying a software engineer’s capabilities, not by replacing them. It acts as a force multiplier – enhancing skills, both good and bad, depending on who wields it.

It could make work faster, but not frictionless. AI-generated code still requires human review, debugging, and contextual decision-making. Improved productivity comes hand in hand with increased collaboration. In fact, the best software emerges from diverse, dynamic teams, where creativity, ethics, and problem-solving are navigated together, not automated away.

AI may create more output, but also introduces new work that cannot be skipped. While AI can toil work, it naturally generates new risks around security, explainability, and bias, which engineering experts must learn how to navigate effectively.

As Fred Brooks famously stated in The Mythical Man-Month, “No Silver Bullet” can eliminate software complexity. AI is no exception – it accelerates, but it doesn’t have the ability to bypass fundamental engineering challenges.

AI pitfalls in the real-world

While AI has demonstrated great potential, there are already some real-world examples of AI-generated code leading to poor outcomes in software development. These highlight the risks of relying too heavily on automation without human oversight:

• GitHub Copilot and security vulnerabilities: When we consider GitHub Copilot and security vulnerabilities, research has shown that GitHub Copilot can generate insecure code, inadvertently introducing vulnerabilities such as improper input validation and weak encryption. A study titled “Asleep at the Keyboard? Assessing the Security of GitHub Copilot’s Code Contributions” found that AI-assisted coding can create serious security risks if left unchecked.
• Meeting UI-accessibility standards: In terms of user interface (UI) accessibility standards, AI-driven front-end design has produced software that overlooks the importance of accessibility. AI-generated code may not follow proper semantic HTML structures, which can lead to usability issues for visually impaired users. In addition, poor colour contrast and missing ARIA roles (content to support screen readers) can make interfaces difficult to navigate. All of these examples illustrate how AI often lacks the nuanced understanding necessary for inclusive design.
• AI-caused system outages: Some organisations have deployed AI-generated code directly into production, only to suffer downtime and performance issues. In cases where AI-generated logic introduced unexpected bugs, these disruptions have had significant business consequences. This emphasises the need for rigorous human-led testing before AI-written code is pushed live.

The importance of human input to creating quality software

Ultimately, designing and delivering high-quality software isn’t just about writing code – it’s about understanding real problems, serving real users, and delivering real business value. Context remains elusive even for humans, and real-world experience and judgement is still the greatest superpower in software development.

Undoubtedly, AI’s assistance will be necessary to keep up, but experienced professionals will continue to play a critical role in:

• Understanding and empathising with users – Product managers, designers, and developers are essential to deeply grasp a customer’s needs. While AI may suggest solutions, humans will decide what truly matters.
• Knowing when to stop engineering and start learning – Product discovery, experimentation, and iterative learning remain human judgment calls. AI won’t replace the need to balance shipping imperfect but valuable software.
• Designing intuitive and accessible experiences – Great UI/UX requires human intuition and empathy. AI can generate UI components, but it lacks the nuanced understanding of accessibility, usability, and human behaviour.
• Balancing trade-offs – Engineering is about making decisions between speed, scalability, security, cost, and user experience. AI can provide useful insights but cannot absolutely replace human expertise in weighing up these factors.

It is also worth noting that AI could play a positive role in spaces such as supporting those with certain neurodivergence, or indeed helping developers who are colour blind manage UI design. These are instances where rather than replacing the human, AI serves as complementary tooling.

Power-assisted steering, not autopilot

While it has been suggested that AI might eliminate engineering roles, instead AI is reshaping them. The most productive DevOps teams will be those that leverage AI effectively while retaining strong human expertise. The rise of AI-augmented engineering means AI will handle repetitive tasks, suggest improvements, and enable engineers to move faster – without replacing them.

Software complexity will always remain and continue to rise under the hood, but our ability to manage it is evolving. The smartest companies will recognise that AI is an accelerator, not a human-replacement strategy. The best software will continue to come from experienced teams leveraging AI as a force multiplier.

🔍 Frequently Asked Questions (FAQ)

1. Why won’t AI eliminate the need for human software experts?

AI increases productivity and accelerates development, but it doesn’t remove the complexity inherent in modern systems. Human judgment, collaboration, and contextual understanding remain essential to building quality software.

2. What is the illusion of simplicity in AI-assisted development?

AI tools give the impression of simplifying development, but like previous abstractions, they often shift complexity into new, less understood areas. This illusion can lead to overconfidence in automation, masking hidden technical debt and architectural challenges.

3. How does AI act as a force multiplier in software engineering?

AI amplifies a developer’s capabilities rather than replacing them. It increases output speed but still requires human oversight for code quality, security, accessibility, and contextual decisions.

4. What are the risks of relying solely on AI-generated code?

AI-generated code can introduce security vulnerabilities, accessibility issues, and even cause production outages. These risks emphasize the need for human review and rigorous testing before deployment.

5. How does human input enhance software quality in the AI era?

Experienced engineers bring critical judgment to product discovery, ethical considerations, and design decisions. Human insight ensures the software addresses real problems and delivers meaningful user value.

6. How does BugBot improve code review processes?

BugBot is an AI-powered agent that reviews pull requests, identifying bugs, security issues, and code quality concerns, complementing both manual and static analysis.

7. Can AI-generated code meet UI accessibility standards?

Not reliably. AI tools often neglect semantic HTML, color contrast, and ARIA roles, leading to inaccessible user interfaces. Human developers must ensure inclusive and usable design.

8. What are real-world examples of AI failures in development?

Studies show GitHub Copilot can generate insecure code, and AI-generated UIs often fail accessibility benchmarks. Some companies have also suffered outages due to deploying unverified AI-written code.

9. How should development teams integrate AI responsibly?

Teams should treat AI as a productivity aid, not a replacement. Successful integration requires combining AI tooling with human expertise in review, decision-making, and risk management.

10. Why is contextual understanding important in software development?

Contextual understanding helps developers prioritize features, balance trade-offs, and adapt to real-world constraints. AI lacks the situational awareness needed for this level of strategic input.