Q: How has software development changed over the last 20 years?
Pedro Carrilho: It’s been extraordinary. Not just an evolution, but a revolution in how we approach building software. We’ve transitioned from 4GL and rapid application development (RAD) tools to low-code and are now entering the era of AI-augmented development. Each shift has been about more than speed; it’s been about rethinking how software is built and how it delivers value. Professional low-code solutions demonstrated that small teams could deliver mission-critical systems at scale. Agentic AI is now taking us to the next level by adding intelligence and autonomy into the software development lifecycle. It’s an exciting time, as we are fundamentally changing the way we work and changing how businesses run.
Q: What was the first major tipping point you witnessed?
Pedro Carrilho: The rise of what is now called low-code 20 years ago. At OutSystems R&D, we observed how abstraction, automation, and service-oriented architecture enabled a handful of engineers to achieve 10 times the productivity. For the first time, visual tools could be used for building software, allowing for faster and more agile application development. Back then, the OutSystems R&D team was small, just 25 to 30 people, but their productivity was staggering. Similarly, this platform and approach applied to other enterprises meant that they could build complex systems in weeks or months, instead of years. I witnessed this revolution in modern software development first-hand. It took a while to spread across Europe, the US, and then the APAC region. Still, the analyst’s expectations—that 70% of new applications developed in enterprises would use low-code or no-code technologies—were eventually realised.
Q: How did low-code prove itself in the enterprise space?
Pedro Carrilho: At first, there was scepticism; people thought low-code was just another rapid development tool. But when CIOs saw it powering mission-critical projects in telcos, banks, and insurers, they realised it wasn’t about cutting corners; it was about amplifying developer productivity with guardrails to deliver real business value. Over time, low-code demonstrated its ability to handle governance, integration, and scale. In 2014, Forrester Research officially coined the term low-code, and it became a recognised force in enterprise software development.
Q: What role did data and analytics play in the software development evolution?
Pedro Carrilho: The data wave a decade ago was critical to what’s happening now. Without this step, we wouldn’t be where we are today. As organisations demanded real-time insights, analytics became as crucial as applications. In 2012, I moved to Australia from Portugal, which is OutSystems’ headquarters, and briefly stepped away from software development to work in this field. I was transforming static reports into dashboards, predictive models and actionable insights for Tier 1 and 2 Australian organisations, helping to bridge the crucial gap between IT and business. This experience in data and analytics enriched my understanding of business language, moving beyond technical jargon to address real-world problems. And it’s directly relevant today, because one of the biggest challenges for AI isn’t the models: it’s the data that fuels them.
Q: Before we get to AI, why did Australian businesses embrace low-code?
Pedro Carrilho: By 2016, the need for speed was undeniable. Organisations wanted to modernise without armies of developers. Low-code provided the answer. This led me to found PhoenixDX, a custom software development company specialising in delivery via the OutSystems low-code platform. In our first engagement, we successfully deployed OutSystems for the leading security services provider, Certis, building operational systems and frontline mobility solutions for 1,500 security guards in just eight weeks. The COVID-19 pandemic further accelerated the adoption of low-code as businesses urgently needed to digitise. Projects like Certis have proven that low-code can deliver enterprise-grade solutions quickly, even in highly complex environments. In 2021, Gartner began publishing significant analyses and forecasts on the rapidly growing low-code market, which solidified its reputation as a serious alternative to traditional coding.
Q: Today, AI, particularly agentic AI, represents another tipping point for software development. Yet, MIT research says 95% of AI pilots are failing. Why do you think that’s the case?
Pedro Carrilho: Just like low-code 20 years ago, AI understandably faces scepticism and is at the peak of the hype, but it’s already proving its value in specific industry use cases. For example, vibe coding is transforming the paradigm of app development with human + AI synergies and Agentic AI is orchestrating models and embedding intelligent agents directly into business processes. This new tipping point is unfolding at such speed that many organisations are understandably struggling – not only with how to approach and leverage this game-changing technology, but even with knowing where to begin. It represents a new paradigm shift, similar to the one we experienced with low-code, and the processes and mindset required to utilise it necessitate rapid experimentation aligned with business problems. Many are implementing AI in an ad hoc way. The key to success, though, is to have a repeatable methodology, playbook and frameworks to implement AI, starting with the education of teams.
Q: But won’t AI replace the need for low-code?
Pedro Carrilho: Not at all. In fact, AI and low-code are much more powerful when employed together. Low-code provides the abstraction and software development lifecycle coverage, while agentic AI adds intelligence and automation across those steps. There are numerous emerging technologies, and we’ve explored and tested them, finding that they have many limitations. Whereas platforms like OutSystems provide the guardrails—security, governance, and compliance— that ensure AI is adopted safely and at scale across the entire software development lifecycle. Combined, they create resilient, future-ready applications. It’s not just faster development; it’s more intelligent software that adapts, scales, and delivers value continuously.
Q: How do low-code and AI work together?
Pedro Carrilho: OutSystems’ Mentor tool goes far beyond using AI for simple “vibe coding”. It applies AI across the entire SDLC – from requirements generation, app prototyping and UX design through to deployment, integration, security and ongoing maintenance and enhancements. OutSystems’ Workbench tool also makes it far easier to embed agentic AI into existing applications. The result? A cohesive and strategic approach to applying AI across your applications, leaving you in total control. Your business can create faster, more innovative, and more resilient applications that can continuously evolve and improve.
Q: What should IT leaders take away from the AI revolution?
Pedro Carrilho: That software productivity is entering a thrilling new era. Just as low-code once transformed delivery speed, AI is now transforming intelligence and adaptability. The real opportunity lies in identifying which use cases and business problems to apply them to and then combining the two technologies in concert. A key message is that you don’t have to navigate this latest revolution in software development alone. That’s where PhoenixDX is focused: helping organisations harness AI to modernise legacy systems, embed agentic AI into critical workflows, and build applications designed to thrive in uncertain times.
