Broken software has become a familiar occurrence, as Hamdija Jusufagic, co-founder and CEO of CodeScene, explains. In his view, this familiarity has become so ubiquitous that organizations barely register it as a problem anymore. Slow releases, brittle systems, rising defect rates, and frustrated engineers are often dismissed as the unavoidable cost of building at scale, often leading to technical debt.
From Jusufagic's perspective, the danger is not just technical debt itself, but the way it quietly drains time, confidence, and strategic momentum from the business. In an AI-driven era, that risk accelerates even faster. Each year, globally, poor software quality costs businesses over $2.41 trillion, creating an accumulated technical debt of $1.52 trillion, which poses the biggest obstacle to making any changes to existing code bases.
loveholidays, an online travel agency, refused to accept that trade-off. As a digital-first travel platform operating in a fiercely competitive market, technology, for the company, is central to conversion, customer experience, and growth. Stuart Caborn, a distinguished engineer at loveholidays, describes the original challenge bluntly: "From our perspective, if you can't measure it, you can't understand if it's improving. All we had was qualitative feedback. Engineers felt the codebase was hard to work with, but we couldn't prove where the problems were or whether things were actually improving."
This lack of visibility, Caborn notes, has become an unavoidable turning point. Without measurable insight, he felt, effort was scattered, and decisions were driven by instinct rather than evidence. Seeking clarity, loveholidays partnered with CodeScene to find an answer for its most persistent question: how to measure what good looks like in its codebase? "We needed to support our engineers, so we appointed CodeScene, implementing their hotspot technology to find the bottlenecks," Caborn says.
By embedding CodeScene's metrics directly into their development workflows, loveholidays gained an objective language for code quality. "What CodeScene brought was focus," Caborn explains. "We could show the C-suite exactly where change would be slower, riskier, and more expensive, and we could also show where teams were genuinely healthy." As a result of this, Caborn notes, Engineers gained the ability to self-regulate, leaders gained greater confidence in decision-making, guiding them down the path with the highest impact.
At the core of that clarity was CodeScene's CodeHealth™, an evidence-based metric built from more than 25 factors scanned from the source code. Each source code file is automatically categorised by Green Code (healthy, low-risk code), Yellow Code (problematic code), and Red Code (unhealthy, high-risk code). In doing so, it transforms code quality from a hidden technical issue into a business KPI, enabling companies to deliver faster, with fewer bugs, and more predictable outcomes.
Midway through that journey with CodeScene, AI entered the equation. Its arrival, Caborn emphasizes, forced a rapid reassessment of how acceleration and quality interact. "AI has an amplifying effect," Caborn says. "If your engineering practices are strong, AI helps you move faster. If they're weak, it will destroy you." The safeguards already in place became the foundation for responsible AI adoption.
This is where CodeScene's deeper research mattered. According to a 2025 study, 65% of developers using AI for refactoring say the assistance misses relevant context. Adam Tornhill, CodeScene Founder, built the company's code health metric on that problem. "The patterns that confuse developers also confuse AI," Tornhill explains. "That makes code health an ideal feedback mechanism. The AI generates code, the metric evaluates it immediately, and the AI can self-correct before problems spiral."
Delivered through CodeScene's MCP server, Tornhill likens the implementation to "a USB port for the AI." According to him, metrics plug directly into AI agents without disrupting developers' workflows, which can facilitate frictionless adoption, grounded in Code Health data, with guardrails that maintain readability and long-term maintainability for both humans and AI.
Caborn emphasizes that results have been tangible. According to him, thirty percent of his teams now operate at what CodeScene classifies as elite code health, which means they are among the top tier of the industry. Average code quality continues to rise with a predictable lag, and systems that once declined, one commit at a time, are stabilizing. He highlights that AI-assisted commits now account for roughly 40 percent of changes, yet quality, deployment speed, and recovery times are all improving.
The result, he notes, has also been visible across the company culture. "We're asking engineers to become thought leaders in their systems, not just mechanics," he says. This impact could free engineers from the more mundane tasks that once filled their time, allowing them to focus on architecture decisions that drive long-term business value, but to achieve this, confidence that the AI will write code to the quality demanded is paramount. "They're still going to be engineers, but the nature of their engineering will elevate to a higher level," he states.
Ultimately, CodeScene functions with the belief that poor code health is not inevitable, and AI does not have to trade speed for quality. With validated metrics and business-aligned prioritization, organizations can move past normalization and into deliberate, scalable progress. loveholidays' case validates that the bottleneck many teams may accept today is not a lack of software; it is a choice, and one that the industry no longer needs to make.
