Beyond the Hype: AI’s True Impact on Software Engineering

From our recent Breakfast Briefings in London and Manchester to our latest webinar, one question dominated every discussion, “How is AI really transforming the way we build software?”

The short answer? It’s early days, but the conversation has already moved beyond the hype. This article explores some of the key themes and takeaways from these events. 

AI Isn’t Replacing Engineers, It’s Redefining the Way We Work

Despite the noise in the media, no one at our sessions believed that AI would replace entire engineering teams. The idea of an AI-driven workforce eliminating human developers was described as a “non-starter”.

Yes, AI is reshaping how we work. It can speed up code generation, automate testing, and even help navigate complex legacy systems. But as one attendee pointed out, “We’ve seen similar revolutions before… from the rise of the internet to cloud computing. AI is simply the next wave.”

For now, many organisations are still in exploratory mode. Most have adopted a ‘wait and see’ strategy, experimenting in low-risk areas before rolling out AI more widely. There’s recognition that wholesale refactoring of existing codebases through AI is impractical, expensive, and risky. 

While AI can accelerate specific refactoring tasks, applying it at scale can easily introduce new bugs, security issues, or break critical functionality that isn’t well documented. Current AI tools often lack full contextual understanding of architecture or business logic, meaning human oversight and rigorous testing are still essential. In many cases, the cost and effort required to validate AI-generated changes outweigh any initial time savings, making targeted, human-guided use far more effective than large-scale automation.

Some participants even noted that companies using AI as an excuse for staff reductions are missing the point. Those layoffs would likely have happened regardless of AI. The real story lies not in replacement, but in augmentation.

AI is helping teams to focus more on design, architecture, and business logic rather than boilerplate code. But the key question remains, “Is AI genuinely moving the dial for our business or are we just experimenting for the sake of it?”

Why Businesses Shouldn’t Have an “AI Strategy”

A clear message emerged across all three sessions, don’t build a separate AI strategy, build a business strategy that aligns to the needs and objectives of the business, and most importantly, the customer.

AI is a tool, not an outcome. The businesses seeing early value from AI are those who started with a clear problem and then asked whether AI was the right tool to solve it.

For instance, some companies are using AI-assisted development tools to reduce time spent on testing and documentation. Others are using AI to help teams explore new programming languages or frameworks. But in each case, success depended on a solid understanding of business objectives, not blind adoption of AI trends.

Without this clarity, many organisations risk falling into the ‘AI FOMO’ trap, adopting new tools without knowing why or how they add value. 

The Key Risks You Should Consider

AI’s productivity boost is undeniable. Tools like GitHub Copilot, Claude, and Cursor are helping engineers navigate legacy codebases, identify bugs, and generate new components quickly. But this speed comes at a cost.

Several engineering leaders highlighted a growing issue,  AI-generated code is often difficult to maintain. It can lack clear structure, documentation, and consistency, making it harder for teams to build on top of it later.

One participant described it as “code wrangling”, like trying to ride a “bucking bronco” where developers constantly regain control over what the AI produced.

Patrick Debois, a leading thought leader in the AI space, recently warned “While AI-generated code and copilots have become commonplace, their role in DevOps and infrastructure remains less defined. The tooling is improving rapidly, but the risks—from non-determinism to lack of visibility—make AI adoption in production a complex, evolving journey.” The industry is producing more code than ever, but not necessarily better code.

The long-term risk is clear: teams that rely too heavily on AI generated code may end up with systems that are faster to build, but slower to evolve and therefore harder to fix in the future.

Three Emerging Use Cases for AI in Engineering

Through our discussions, three main use cases emerged where AI is proving most effective:

1. Vibe Coding
   Using AI to rapidly prototype ideas, visualise concepts, or create proof-of-concepts for stakeholder or investor demos. This use case can be valuable for fundraising or early validation, but the resulting code should never be deployed to production.
   
   
2. AI as a Coding Assistant
   Engineers are increasingly using AI to automate repetitive tasks such as writing boilerplate code, generating unit tests, and explaining legacy code. This can dramatically accelerate workflows when used under proper review and governance.
   
   
3. AI as a Learning Tool
   Perhaps the most promising use case. AI can help developers learn new frameworks, suggest better design patterns, and even guide them toward full-stack proficiency. Several organisations reported that introducing AI learning tools helped new hires become productive within days.

Interestingly, some start-ups are using AI as a way to quickly build credibility. For instance, building and demonstrating rapid technical progress to investors and stakeholders by using AI to build early prototypes or Minimum Viable Products (MVPs). 

Prompt Engineering: The New Core Skill

If traditional programming is about writing code, Prompt Engineering is about designing intent. Prompt engineering is quickly becoming a first-class skill for engineers. Clear, structured, and well-thought-out prompts can mean the difference between functional, elegant output and unusable code.

But as our participants noted, the language of Prompt Engineering is still evolving. More recently, the term “Context Engineering” has been introduced as the next level of detail and nuance around Prompt Engineering. 

The syntax and techniques that work with today’s tools, like Copilot or Claude, may change entirely within months as new models emerge.

For this reason, companies were advised not to get locked into long-term vendor contracts. The goal should be to integrate AI tools that align with your workflows today but can be swapped out easily as the market evolves.

Culture Over Tools

Perhaps the most important insight from our events, AI should never replace engineering culture. AI works best when it enhances collaboration, not when it replaces human expertise. Teams that openly share their best prompts, review AI-generated code together, and continuously refine their usage guidelines will thrive.

Those that use AI in isolation risk creating fragmented knowledge and dependency on the tool rather than the team.

As one speaker said: “AI is an assistant, not a replacement.”

There’s also a talent risk. Over-reliance on AI can lead to the erosion of core engineering expertise. If teams lose the ability to read, understand, and improve code without AI assistance, the organisation’s technical resilience weakens over time.

At Codurance, culture is at the heart of how we work. From pair programming to community-led learning, our approach ensures teams build lasting skills and collaborative habits that AI only enhances, never replaces.

We also host thriving Software Craftsmanship communities across London, Manchester, Newcastle, Cambridge, and Leeds, where engineers and developers come together to collaborate, share ideas, and continuously improve their craft. 

The Takeaway: Build Software That’s Easy to Change

Ultimately, AI doesn’t change the most important rule of software engineering, build systems that are easy to change, maintain, and scale. AI can be a fantastic catalyst for experimentation, a way to test ideas, validate concepts, and learn fast. But not every piece of AI-generated code should survive beyond the prototype phase.

The best teams will continue to listen to their customers, understand their needs, and then decide which tools can help deliver the most value. AI isn’t replacing software engineering. It’s redefining it, and the most forward-thinking organisations will treat AI not as a shortcut, but as a creative partner that supports their long-term goals.

Codurance helps organisations do exactly that through software modernisation, building adaptable systems, empowering teams, and enabling confident experimentation with AI. Our goal is simple - to help you build software that’s easy to change, easy to trust, and built for the future. In addition to this, our Data and AI Readiness Assessment gives companies clear, independent recommendations to evolve their data strategy and kickstart AI adoption effectively.