AI-Driven Development: Will It Turn Airlines into Software Builders?

Aviation is hardly new to AI coding tools. airBaltic had already adopted GitHub Copilot to improve development efficiency in 2023. By early 2024, Lufthansa Systems had formally rolled it out to its tech teams, while Air India said every software engineer in the company was working with the platform. Air France-KLM, Cathay Pacific, and Virgin Atlantic were among the other early adopters of Microsoft’s coding assistant.

Just two or three years on from those early experiments, the technology has moved far beyond inline suggestions and autocompletion.  The AI-driven software development lifecycle (AI SDLC) is becoming a familiar concept, describing the use of AI end-to-end, from planning and design to coding, testing, deployment, and maintenance.

That depth of adoption matters: the more closely AI is integrated across the lifecycle, the greater the potential efficiency gains. At AltexSoft, depending on the project, an AI-first approach helps our teams accelerate delivery by up to 70 percent and bring new features to market in days rather than months.

Getting there, however, takes more than simply buying additional tokens.  It means rethinking development processes, roles, governance, and, ultimately, the way software teams work.

This article looks at how far AI-driven SDLC implementation has progressed across airlines and aviation technology providers. We draw on public information, AltexSoft’s internal experience, external experts, and airline questionnaires to show what companies are already doing and where the main challenges remain.

Copilot is no longer evidence of real transformation

GitHub Copilot was one of the earliest AI coding tools to gain traction in aviation and remains among the most widely used. But the market has broadened quickly, with airlines now experimenting with Claude Code, Cursor, Kiro, and other agentic development environments.

Still, adopting a particular platform, even company-wide, says little about how deeply AI is embedded in an airline’s software development lifecycle. In many cases, it remains limited to coding assistance.

Victor Alzate, CPO at Ink Innovation, an aviation software company focused on passenger service delivery, describes the current state of AI adoption this way: “Airlines are often waiting for the major players to make the first move. Once they see that it works at scale and the risks are manageable, others follow. But the shift is already underway. We’re having discussions with some large airlines that are already going through this transformation internally.”

Some publicly documented airline initiatives show AI extending beyond coding assistance into testing, security, code review, modernization, and longer-running agentic tasks. 

Cathay Pacific, for example, rolled Copilot out to more than 1,000 developers and moved beyond code completion to agent mode, which can execute multi-step tasks across a codebase. The airline reports a 40 percent year-over-year improvement in fixing technical debt (addressing accumulated code issues in less time) and 63 percent faster vulnerability resolution.

SunExpress aims to automate virtually every stage of the SDLC with AI. "This includes collecting and enriching requirements, creating user stories, development, acceptance and vulnerability testing, environment provisioning, and deployment," says the airline's CIO, Mustafa Egilmezbilek. The team uses Claude Code with MCP and custom skills. It applies the same AI-driven approach for enterprise app development, including Salesforce Apex and SAP ABAP environments.

In day-to-day work, the story point completion rate (the amount of estimated development work the team finishes within a sprint) has increased by 64 percent. In focused experiments, the gains have been much larger. "In a five-day workshop, a two-pizza team completed four greenfield and two brownfield projects. Development was about eight times faster for greenfield and five times faster for brownfield (modifying, upgrading, or redeveloping existing systems. — AltexSoft). The output equals about 135 person-days of traditional work. That was a real eye-opener for us."

At Ink Innovation, AI is also run across the end-to-end lifecycle. “Before a single line of code is written, AI challenges the request,” says the company’s CTO, Jose Santamaria. “It maps it against our domain model, surfaces hidden dependencies and, when needed, rescopes it. In aviation software, the most expensive defect is the one written into the requirement. The level of autonomy scales with the engineer’s seniority, but AI supports 100 percent of our lifecycle, from refinement to release.”

The real breakthrough is a shift to spec-driven development and end-to-end automation