AirAsia CTO Bets on AI to Rewrite Legacy Software — and on His Engineers to Trust It

AI Summary

AirAsia's chief technology officer, Wan Yussman Wan Yusof, is using AI to attack one of the most expensive problems in any large organisation — legacy software — while redesigning how his engineering teams work day to day. In an interview with Digital News Asia, he made a point that many vendors skip: the technology is the easier half of the job. The bigger challenge, he says, is building trust, clear accountability, and new skills across his workforce. For Malaysian business readers, this is a rare public case study of a local enterprise treating AI adoption as an organisational change programme rather than a software purchase. The lesson is straightforward — cost savings from AI-assisted software rewrites are achievable, but they only arrive if you manage people as carefully as you manage code.

Key Takeaways

• AirAsia is targeting legacy software costs with AI-assisted development — recurring licence and maintenance fees for old systems are typically among the biggest IT line items for airlines, banks, and large government agencies.
• Wan Yussman frames the real obstacle as human, not technical: employees must trust AI output, someone must be accountable for AI-assisted work, and engineers need genuinely new skills.
• The engineer's role shifts from writing every line to directing and reviewing AI-generated work — a different job, not a smaller one.
• Malaysian enterprises outside aviation face the same legacy software problem, which makes this a transferable playbook for banking, logistics, and the public sector.
• Agentic AI can now draft code, run tests, and prepare system migrations largely end-to-end, but human sign-off remains the control point that keeps this safe.

What Happened

Digital News Asia reported that Wan Yussman Wan Yusof, chief technology officer of AirAsia, sees artificial intelligence as an opportunity to fundamentally change how the low-cost airline develops and operates its technology. His push runs on two tracks. The first is offensive: using AI to challenge the cost of legacy software — the ageing systems airlines accumulate over decades of operations. The second is internal: rethinking how his engineers actually work now that machines can do part of the job.

The DNA report adds the detail that matters most. When asked about the hardest part of this transition, the CTO pointed not at the technology but at the workforce. Trust, accountability, and new skills — those, in his assessment, are the bigger challenges. Getting engineers to rely on AI-generated code, deciding who answers when that code fails, and training people for a role that looks different from five years ago.

It is worth being precise about what "using AI to rewrite software" means in practice, since the source does not detail the specific tools. In industry terms, this usually means AI coding assistants that generate, refactor, and translate code under an engineer's direction — for example, modernising an old booking or operations module into a newer language or architecture, or rebuilding in-house functionality that a company currently pays a vendor to maintain. That reading is my analysis, not a quote from the CTO. What the source confirms is the strategic intent: AI as a lever against legacy software costs, paired with a deliberate overhaul of engineering workflows.

Why It Matters

Legacy software is the quiet tax on every large organisation. Airlines, banks, and telcos run core operations on systems built years or decades ago. They work, so nobody replaces them — but the licence fees, specialist maintenance contracts, and integration costs compound every year. When a CTO publicly says AI can challenge those costs, it signals that AI-assisted code generation has moved from demo territory to a genuine procurement and budget conversation. If an airline — an industry with thin margins and zero tolerance for operational failure — is willing to try, the argument that "our systems are too critical for AI" weakens everywhere else.

The second reason this matters is what it says about adoption. Global experience with generative AI in enterprises has followed a consistent pattern: pilots are easy, sustained use is hard, and the failure point is almost never the model. It is whether staff trust the output, whether accountability is clear when things break, and whether people have been retrained. Wan Yussman naming trust, accountability, and skills as the bigger challenge puts him squarely in line with that pattern — and it is refreshingly honest for a Malaysian executive to say so publicly rather than repeat vendor talking points about instant productivity gains.

There is also a competitive signal here. A low-cost carrier's entire business model is cost discipline. Applying that instinct to IT, with AI as the instrument, suggests AirAsia views software cost reduction as a structural advantage, not a one-off project. Competitors and peers in Malaysian aviation, travel, and logistics will read this the same way. In AI news from Malaysia, most announcements come from vendors or government launch events. A practising CTO describing the human side of adoption is rarer and more useful.

What This Means for Malaysia

This is a Malaysian case study, and that changes its weight. Most enterprises here — banks, government-linked companies, manufacturers — sit on the same legacy software problem AirAsia does. AI automation in Malaysia has often been discussed at the policy level, through MDEC's digitalisation push, the MyDIGITAL agenda, and national AI planning. What has been missing is grounded evidence of a large local company operationalising it against hard costs. AirAsia's effort, if it works, becomes the reference case other Malaysian CTOs cite to their boards.

For talent, the implications run both ways. Engineers who learn to direct and review AI-generated code become more valuable; those who refuse may find the market moving past them. That feeds directly into the AI jobs conversation in Malaysia — the premium skill is shifting from writing code from scratch to specification, review, and judgement over machine output. Polytechnics, universities, and HR departments should note this.

One regulatory thread to watch: PDPA. When AI tools process source code or data tied to passenger records, personal data can flow into external AI services. Any AI-assisted development programme needs a data handling policy from day one — which candidates for AI processing, which tools are approved, and where the data sits. The CTO's emphasis on accountability pairs naturally with this: someone must own not just code quality but compliance.

How Your Business Can Use This

Start with money, not technology. Pull up your software spend for the past two years: licences, maintenance contracts, and the hours your team spends babysitting old systems. Identify one or two systems that are expensive but not safety-critical — an internal reporting tool, a legacy inventory module, an old customer database front end. That is your pilot candidate