October 29, 2024
Editor’s Note: This article first appeared in The Current, Jellyfish’s LinkedIn newsletter. You can subscribe for monthly updates and articles like this one here.
Software is a key engine of growth in the global economy. The market research firm Statista estimates that total revenues for the software industry will reach $702 billion in 2024, growing to nearly $900 billion by 2029.
The people who actually drive this industry – software engineers – form one of the most important roles in an organization today. Empowering engineers to do their best work can make the difference between rapid growth and stagnation. A recent analysis from McKinsey found that software companies could unlock an additional $500 billion in shareholder value by focusing on efficiency.
But the role of the software engineer is getting more complicated – and generative AI could make things even worse without proper change management. When implemented correctly, AI coding tools help remove repetitive tasks and allow developers to focus more on strategic work. But without a clearly defined strategy for how and where to deploy AI, an organization risks tool overload. Engineers hate wasting time: if a new tool gets in their way, it’s going to have negative consequences for developer experience.
Last month, I connected with Andrew Boyagi, Head of DevOps Evangelism at Atlassian, for a LinkedIn Live on the intersection of AI coding assistants and the developer experience. That discussion made it clear: implementing new AI tools without the ability to measure impact on developer experience is a recipe for disaster.
Software companies need to invest not just in tools that can objectively measure the impact of GenAI but also in DevEx – the science of identifying and resolving the challenges impeding effective output, software delivery, and business outcomes. As software companies introduce GenAI and coding assistants to their workflows, DevEx will be essential in ensuring these tools drive productivity rather than derail it.
Rolling out GenAI: The right way and the wrong way
Before rolling out GenAI across the organization, engineering leaders should be able to answer three simple questions:
- What are we trying to achieve? Tools are only beneficial if they lead to a specific outcome.
- Why are we using this tool to achieve it? Implementing a new tool needs to be intentional. Engineers won’t take on genAI if they don’t understand what the end goal is and why the organization is investing in it.
- When and where will we use the tool? Different engineers will take different approaches to AI. Some see AI as a religion while others don’t want to go anywhere near it. Leadership should put in place simple guardrails for where they expect engineers to use AI and where they should stick with existing processes.
Intentional planning and goal-setting will make the difference between organizations that succeed with GenAI and those that end up wasting their money. You can’t just plug GenAI into your toolset and expect a 20% increase in productivity.
Tools are only useful if they solve a problem. For example, engineers don’t typically enjoy creating documentation; GenAI is good at creating documentation, or at least taking care of the bulk of the most tedious work. By applying AI to existing problems, engineering leaders can show why they’re investing in it and foster buy-in among their engineers.
AI coding tools are evolving quickly, and it’s important that leaders and engineers have an opinion on how they engage with AI. An engineering organization shouldn’t blindly pick up an AI solution, nor should they blindly push them away.
A practical approach to GenAI and developer experience
DevEx can help engineering leaders understand at every stage whether they’re on the right track with GenAI – as well as other key factors impacting productivity and developer satisfaction.
The first step to establishing a baseline for DevEx is to gather feedback from the engineers themselves. If you ask the CTO of a large company how they build software and what challenges they face, the CTO’s answers will often diverge significantly from what an engineer would say. The goal of improving DevEx is to bridge that gap to help leaders better understand the challenges facing their individual contributors.
Speaking to the engineers themselves is necessary, but it’s not sufficient to gain a complete understanding of the situation. Engineering leaders need to look at objective data to validate and provide context for the signals they’re getting from the developers themselves. Engineering data will reveal how much time is being spent on specific types of work – whether that’s building new features, fixing bugs, or writing documentation. Keeping an eye on those metrics will make it easier to evaluate the impact of GenAI. Has a team been pushing new code more frequently since introducing an AI coding tool? Are they spending less time on documentation than they did last quarter? The differences may not seem significant at first, but even small improvements stack up over time to deliver outsized value.
Of course, it’s not enough to just identify a problem with DevEx; you have to actually do something about it. You may have an inkling about a problem from the results of a survey: a few developers may be complaining about code quality since introducing GenAI. That signal should prompt you to dig in more with that team and determine how they’re using the tool – maybe a small correction to the peer review process could improve code quality while maintaining an increase in speed.
The most important thing is to act on the information you’re getting from your engineers – the only way to guarantee a poor developer experience is to let problems go unaddressed.
New tools should make life easier for engineers
It sounds simple, but it bears repeating: the best tools are the ones that help engineers do their jobs. The development environment is complex, and engineering leaders need to think carefully before adding another tool to the mix. A useful tool will be one that makes life easier for the engineer by reducing their cognitive load.
Depending on the implementation, GenAI will either reduce complexity or contribute to it. The stakes are high. An organization that gets their deployment right is going to move more quickly than the competition; if they get it wrong, they’ll be stuck in place – or even moving in the wrong direction.
Investing in DevEx gives engineering leaders the visibility they need to track whether new tools like GenAI are working as intended. DevEx surveys – paired with objective engineering data from Jellyfish’s Copilot Dashboard – provide valuable signals around potential problems, allowing engineering teams to introduce a solution before the issue spins out of control. As engineering becomes more complex, that early warning can be a lifesaver.
For more on DevEx and AI coding assistants, watch Atlassian’s full LinkedIn Live recording with Andrew Lau and Andrew Boyagi.
