Andrew Boyagi: 'AI alone is not a panacea’

AI is the catalyst, but humans make the difference.

Andrew Boyagi, customer CTO at Atlassian, says AI’s impact should never be measured in lines of code, but in human creativity.

AI's rapid development has fundamentally changed the world of work and software development, particularly in recent years. Companies are faced with the challenge of rethinking their traditional processes and making good use of innovative technologies to increase productivity and minimise friction.

We spoke with Andrew Boyagi, customer CTO at Atlassian, about how organisations are successfully making the transition to AI-powered development and what opportunities and hurdles need to be considered.

Andrew is responsible for rolling out products and solutions at Atlassian Williams Racing - technically and organisationally. Previously, he was Head of DevOps Evangelism, where he was responsible for the State of DevEx study.

Andrew, what do you think are the biggest cultural and organisational hurdles for global companies when they want to move from traditional to AI-powered development processes? How are agile approaches such as DevOps or DevSecOps changing as a result of AI?

Companies are increasingly relying on AI in software development to deliver high-quality solutions faster. According to our study, AI developers save about 10 hours per week, but they continue to lose a lot of time due to non-coding tasks and organisational obstacles.

Sixty-three percent of developers feel that their managers don't understand these issues. Companies should therefore focus less on the change in technology, but rather specifically eliminate friction losses in the developers' everyday work.

Only when these challenges are solved can AI unleash its full potential – developers will then use the most efficient tool on their own.

How do leading organisations measure the strategic value of AI in code development? What new metrics may need to be introduced to measure the success of your AI strategy when traditional metrics like "lines of code per day" are outdated?

It is questionable whether key figures such as "lines of code per day" ever really made sense. The DevEx report 2025 shows an interesting picture: Developers save ten hours a week with new tools, but lose this time again due to various inefficiencies. The key metric should therefore not be how much time AI technologies save or how much code is written, but how much friction we have been able to reduce in the entire software development process.

The real question in the AI age is: Has technology helped us remove our biggest obstacles? The point is not to measure how much code the AI writes, but whether it achieves the desired results.

Where do you see the biggest untapped advantage of AI for Dev(Sec)Ops so far?

Our research for 2025 shows that the biggest challenges for developers are not coding itself. Rather, it's factors like finding relevant information, navigating new technologies, and constantly changing context that have long been a challenge for developers.

This is also the biggest untapped opportunity for DevOps: to address these problems with AI and eliminate friction in the entire software development process. Less friction means better flow, which in turn improves the developer experience – and that leads to happier and more productive teams.

A practical example is our product Rovo Dev. This allows agents to be developed that can be deployed throughout the software development lifecycle. These take over the monotonous and time-consuming tasks that developers often face, such as conducting code reviews, creating deployment summaries, or troubleshooting issues in CI/CD pipelines.

When developers are relieved of such activities, they spend less time on handoffs, status updates, and searching for information. This promotes synchronisation within the teams and significantly increases joint performance.

How can organisations best prepare for AI-driven engineering?

An ill-considered introduction of AI solutions can exacerbate existing difficulties and slow down processes. Therefore, a holistic view of software development is necessary.

An effective approach to developing an AI strategy starts with considering the needs and perspectives of developers. Instead of starting development right away, it's crucial to first understand the challenges. Listening to developers is key.

AI should also be integrated as an integral part of existing programmes to improve the developer experience, rather than treating it as a standalone project.

In addition, managers should encourage experimentation and training: Teams that are supported in this process save significantly more time and use AI strategically. AI should be part of existing developer programs and not run as a single project. For example, our 2025 report, The State of Teams, shows that employees whose leaders encourage them to experiment with AI save an average of 55% more time than those who don't.

Employees with such support are more likely to become strategic team members: they don't just use AI as a research tool, they actively use it as a sparring partner and integrate it into their workflows, which improves both efficiency and the quality of results.

In conclusion, acceptance arises automatically when AI offers real added value. If the tool does not help, acceptance can hardly be increased. The goal should therefore always be to align AI with developers’ specific problems, and at the same time to accompany it through information and training, without focusing one-sidedly on acceptance.

Build vs. Buy: How do your customers decide which AI capabilities need to be developed internally, and which can be sourced from third-party providers or partner ecosystems?

The "build or buy" question is particularly relevant in the context of AI solutions. Crucial factors here are the context and the available data that the AI can access - not primarily the ownership of the LLM used.

Atlassian is uniquely positioned in this area through the integration of JIRA and Confluence. Our platform already possesses comprehensive contextual information about collaboration, objectives, and the respective software components of various teams. This allows us to cover the entire SDLC and specifically identify and analyse challenges for developers.

To what extent is the use of “shadow AI” - the uncontrolled use of AI tools by developers - already a problem, and how can companies control and channel this use?

Shadow IT can be problematic in certain situations, but it isn't always. Developers are often eager to try out and use new technologies - a behaviour that has always been observed in the industry. Engineering managers should therefore create pathways to enable the responsible use of new technologies and thus minimise the need for shadow IT. At Atlassian, this is fostered by establishing a company culture that emphasises the sharing of experience.

One example of this is our AI Playground, which allows internal teams to experiment safely and in a controlled environment with various LLMs and AI tools. In addition, we regularly host ShipIt Day, a quarterly 24-hour hackathon focused on innovation, including AI applications; and there are department-specific AI days where teams - including those outside of engineering - receive training and explore AI use cases. For example, a dedicated AI day was recently held for the marketing team.

All AI-related experiments and implementations at Atlassian are subject to strict guidelines for responsible technology and all relevant compliance regulations. For other companies, it is recommended to define clear guidelines for the safe use of AI early on and to establish a culture of experimentation, ideally in isolated, secure environments.

Furthermore, approved, high-quality AI tools should be provided that integrate seamlessly into existing workflows to reduce the need for external solutions.

If you could give other CTOs one strategically important piece of advice, what would it be?

My top tip: Talk to your developers. Research shows that many executives are unaware of the challenges their development teams face, a trend that is increasing in 2024 and 2025. This can lead to inefficient investments. Therefore, systematically build feedback using surveys and data, and prioritise platform and AI investments based on these insights. Use AI strategically where it addresses friction points for developers and improves collaboration.

Team Dog, Cat, or Fish?

I've had both, but I prefer dogs. They're happy to see you when you come home. When I was a kid, I often couldn't even find the cat when I got home.

Famous last words: What's your message to the world?

AI alone isn't a panacea. The teams that will win aren't the ones with the most AI-generated code. They're the teams that combine AI with an efficient developer experience, establish clear lines of responsibility, and have leaders who truly understand developer problems.

This interview was originally published on our sister site, Computing Deutschland.