Bupa’s mobile app modernization reads like a case study in what happens when you stop treating legacy systems as a maintenance problem and start treating them as a business liability. The health insurer rewrote its My Bupa application from Xamarin to native Swift and Kotlin, and the results are the kind of numbers that make CFOs pay attention: app store rating jumped from 3.7 to 4.7, while user-perceived crash rates fell by nearly 24 percentage points on Android and eight points on iOS.
Asifa Sherazi, Bupa’s CIO of health insurance, frames the urgency bluntly. “The end-of-life technology is a risk that compounds quietly, and then arrives all at once,” she said. That’s the trap most enterprises are in — legacy systems don’t fail dramatically, they degrade slowly until a moment of crisis exposes them. For Bupa, the stakes are particularly high: seven million healthcare customers across Asia-Pacific, many reaching for the app during genuinely stressful moments.
Sanjeev Tripathi, senior VP at Infosys, argues AI has fundamentally shifted the economics of these projects. Using AI-assisted reverse engineering paired with forward engineering, Bupa delivered the transformation in roughly 60% less time than the pre-AI era would have allowed. That’s a meaningful claim, because modernization timelines have historically been the thing that kills these initiatives before they start. If AI compresses a multi-year migration into something closer to a year, the risk calculus changes entirely.
There’s also a human dimension that gets lost in modernization conversations. Sherazi and Tripathi emphasize preserving institutional knowledge and giving teams room to surface problems early — the unglamorous work that determines whether a rewrite actually sticks. Tripathi’s endpoint is clear: modern platforms become “the base for far more intelligent AI-driven ecosystems,” where AI is built into design and operations rather than bolted on afterward. For Sherazi, the more telling shift is cultural: the question used to be “can our platform support that?” Now it’s “is that the right thing to do for our customers?” That’s a fundamentally different conversation, and it only becomes possible once the platform stops being the bottleneck. The open question is whether other enterprises will wait for their own compounding risk to arrive all at once — or act before it does.
