The revolutionary role AI is playing in app modernisation

Joel Macfarlane, Datacom’s Director - Data & Engineering, and a 28-year veteran of software development, describes the role artificial intelligence is now playing in software development as “the largest change management exercise in the history of IT”.

While AI tools and copilots have been widely deployed in software development with the rise of large language models (LLMs), the technology is particularly applicable to app modernisation, the process of updating legacy applications to take advantage of newer technologies and platforms.

“The developers aren't developing to the degree they used to, the testers aren't testing as much. What they're increasingly doing is wrangling large teams of AIs to do those jobs, or they're sitting down trying to crack challenges with them," says Macfarlane, who oversees a 300-strong team of engineers at Datacom.

Many of Datacom’s customers are in the midst of refreshing legacy systems, but cost is a major barrier to modernisation plans. With AI agents now able to handle many of the tasks traditionally performed by human developers, business analysts and testers, New Zealand's ageing IT infrastructure can be rapidly refreshed at much lower cost.

"The challenge for organisations still using legacy systems is that often they have become too expensive, slow and difficult to upgrade. Rather than taking the usual developer-led approach to modernising these systems, we’re taking an AI-led approach that enables our team to deploy AI agents to write up to 70% of code, leading to cost savings of between 30%-50% for our customers - in some cases even more - and a significant reduction in time spent to complete the project."   

The cornerstone of Datacom's approach involves treating application modernisation as an assembly line process. Rather than manual code translation, AI agents developed by Datacom and drawing on a series of LLMs, perform discrete tasks allowing for dramatic efficiency improvements, including:

• Automated documentation analysing 20-year-old systems and generating large specifications in hours instead of months.
• Self-managing developer teams where AI "tech leads" review code commits and maintain consistency across distributed work.
• Continuous validation through AI testers that execute hundreds of test cases across old/new systems simultaneously.

“We emulate the way that human teams work,” Macfarlane explains. “We give the AI agents access to things like code repositories, documentation, and project management software so they can collaborate together and break up challenges,” he says.  

It’s all done in a way that human software engineers can oversee, monitor and understand. The AI agents even update the status of their work on a digital kanban board just like their human colleagues.

For businesses struggling with systems dating back to the 1990s, the implications are transformative. Macfarlane and his team are witnessing major productivity gains in areas like documenting app modernisation requirements and migrating code from legacy systems to new platforms.

For years, the success of IT projects has been judged by the famous Iron Triangle, which tries to balance three priorities: time, cost, and quality. Achieving two of those on a complex project is usually considered a success. But the new AI-powered tools available to software engineering teams dramatically improve the likelihood of goals in all three categories being reached, says Macfarlane.

“For years, app modernisation projects have been plagued with failed implementations, delays and cost overruns,” says Macfarlane. “It doesn’t have to be that way anymore.” 

The AI revolution isn't eliminating software development roles at Datacom, but it is radically reshaping them: "You're either communicating with humans or corralling AIs - there's no middle ground anymore," says Macfarlane.

Senior developers are coding less and employing AI orchestration tools to a greater degree. Graduates enter the workforce as "AI engineers" rather than traditional programmers. Project managers oversee AI agent workflows as well as human teams.

Datacom is addressing this evolution through intensive reskilling programmes, but Macfarlane acknowledges that the cultural shift is as important to get right as wrangling the new AI agents. 

Macfarlane says the app modernisation efficiency gains it has documented are impressing customers who see the appeal in months-long feature development programmes being compressed to weeks.

The use of AI in software development can also improve cyber resilience by rapidly rebuilding vulnerable systems on secure platforms and strengthening regulatory compliance by automating documentation to ensure audit readiness.

“For businesses concerned with competitive agility but weighed down with dated technology, this capability couldn’t emerge at a more critical time.”