Recently there have been many debates on future of the developer role, with some saying AI will put developers out of a job and others professing dev teams will always be a necessity. As often is the case, the truth seems to lie somewhere in the middle.
At the recent SAP TechEd conference, TechDay sat down with Casey West, Lead Developer Advocate at Google, to discuss how AI is impacting developer communities, how leadership can support dev teams, the core characteristics of an elite organisation, and much more.
The biggest AI misconception? AI will replace developers
With a career spanning over 25 years in internet infrastructure, application platforms and information security, West currently works as a developer advocate for Google, which has a close relationship with SAP. In this role he's focused on empowering developers with engineering practices. He's also an active speaker and writer, continually engaged in conversations around the future and reality of development.
West says the idea that AI will replace developers is 'the biggest misunderstanding'. He comments, "I know every organisation like SAP and Google have a vested interest in saying, "No, developers will be here," but in my experience this blanket statement that AI will replace developers is inaccurate.
"For two and a half years now I've been helping people build AI-powered systems or agentic solutions. I've been helping dozens of organisations around the world. What I see is that if you're the type of engineer that does vibe coding and thinks that that's production grade, AI could replace you. But if you're considering software engineering from a professional perspective, from an enterprise grade perspective, if you understand the professionalism and rigor that's required to build reliable systems that customers can count on, you're going to be fine. AI can be used as a tool for enablement and acceleration, not replacement."
The ever-evolving role of AI for development
According to West, AI models are 'pretty good' at following directions, and if they have ample context and information they can produce a good result. However, the result is what you might expect from a first year intern. As such, the power of AI is not so much in outsourcing work, but using it to produce quicker results. On the limitations of AI, West encourages coders to do their due diligence, and organisations to work to build necessary guardrails. As the technology improves, this will become all the more important.
He says, "AI coding assistance, overall, has made me faster as an engineer to deliver a solution, but it is still quite tedious because I have to provide a lot of guidance. I have to provide a lot of context. [...] We're still in an era where we're just scratching the surface on being able to leverage AI for really deep and trustworthy business applications, and to do it well requires the rigor that I see SAP bringing to the table."
On the role of responsible AI implementation, he says, "We are at the very beginning of what it means to do responsible, enterprise grade software development with AI, and it still takes a great deal of discipline to do it well."
He continues, "Any software development organisation that's mature has guardrails and systems in place to protect engineers and customers. Traditionally, we've seen these systems be the likes of unit tests, peer reviews and code reviews. The protections and guardrails are shifting a little bit, the needs are different in the age of AI. [...] In non-deterministic systems, and in an AI centered approach, we need to have a different set of guardrails. From a leadership perspective, it's critical to give your teams the opportunity to build those trustworthy safeguards and guardrails in order to keep you safe in an agentic era."
AI for learning, not outsourcing
West emphasises the importance of using AI as a tool and a learning resource, not to outsource all coding.
He says, "If you're not careful, if you're not thoughtful, if you don't look underneath and actually see what the implementation is and make sure you've validated it, you can get yourself into trouble. There have been some unfortunately very public examples where organisations have delivered an AI solution, and that AI solution has done surprising things."
So what is the balance to strike, and what should developers be focusing on? According to West, the problem has to come first, as well as understanding the domain and the building blocks required to achieve a solution.
He says, "You need to fall in love with a problem to solve, not the solution. Domain expertise, domain knowledge in the problem you're actually trying to solve, is one of the most important pieces of the puzzle. Usually that's directly in the knowledge of your customers, the end users, or at least in your product management team and customer service team. Engineering teams that aren't continuously engaging with those folks who actually understand the domain the best are going to fail to deliver high quality AI solutions. The ones that will be the most successful will be those that have the biggest integration within the different areas of business."
On top of this, West says developers engaged in continual learning will be those that come out on top. He says, "Whether you've been in the industry for 25 years or you're just starting, it's important to value learning new things and being able to learn new things quickly. When used correctly, AI systems, whether that's Gemini, ChatGPT or Perplexity, can help you learn deeply, and learn new things well and quickly. Any developer who isn't continuously learning is going to find it difficult to keep their job or get new jobs."
What makes an elite organisation in an AI era?
As technology continues to shift at a rapid rate, leading organisations continue to share similar characteristics. If anything, the speed of technology cycles is only accentuating these characteristics. The qualities of these organisations can offer insights into how all businesses can respond to challenges of our advanced, AI-powered era.
As shared by West, DORA research, the long running research programme that seeks to understand the capabilities that drive software delivery and operations performance, is an excellent insight into the qualities of 'elite organisations'. In 2018, Google acquired the team leading the research and retained them as an independent research entity.
The top four metrics of these leading organisations are: deployment frequency and lead time for changes (throughput), change failure rate and mean time to restore (stability).
As shared by Nathen Harvey via DORA Guides, "DORA's four keys can be divided into metrics that show the throughput of software changes, and metrics that show stability of software changes. DORA's research has repeatedly demonstrated that speed and stability are not trade-offs. In fact, we see that the metrics that the four keys focus on are correlated for most teams. Top performers do well across all four metrics, and low performers do poorly."
As West says, "It's been pretty instrumental in defining what aspects of an organisation make them successful, or make them very elite in software development and delivery, their ability to deliver quickly and accurately, and what holds them back."
Further emphasising this point, West references Ronald Westrom, a sociologist who, in 1988, developed a typology of organisational cultures. The four categories he defined were: pathological (power-oriented), bureaucratic (rule-oriented), and generative (performance-oriented). West says even in today's vastly different world, organisations that are generative, meaning they are more performance focused whilst maintaining stability and understand that this must come above rules or power, are those that can and do ride the waves of change.
West says, "A generative culture is one that embraces failure, that when things don't work it's seen as a learning opportunity, because we can learn from it and avoid making that same mistake in the future. It's a culture that allows people to speak up and say something isn't quite right, we should fix it, and we don't shoot the messenger. Organisations that are more generative have a lot more grassroots autonomy - people of all levels have autonomy to do the right thing in their organisation."
