WATCH | This is how FirstRand is using AI to help its 3,000 engineers work faster — and better

FirstRand is using artificial intelligence (AI) to help its estimated 3,000 engineers work faster and improve the quality of digital financial services while keeping human developers responsible for the final outcomes.

Speaking at the Amazon Web Services (AWS) Summit in Joburg on Wednesday, FirstRand Group chief digital officer Kevin Mitchell said the financial services group faced growing demands from customers and needed to give it’s engineers “more leverage” without compromising service reliability.

The summit brings together technology leaders, businesses and developers to explore cloud computing, AI, digital transformation and application modernisation.

Mitchell said early AI coding assistants were useful but limited because they did not understand the full environment in which developers worked.

“We don’t have everything figured out. Every week, there’s a new learning that comes up.” — Kevin Mitchell, FirstRand Group’s chief digital officer

The group therefore trained its AI tools using the same information available to human engineers, including functional requirements, test plans, architecture standards and security controls.

“The quality of AI-generated code depends least on the individual prompt, but more on the quality of the organisation’s specifications,” Mitchell said.

“The speed becomes the steering wheel for both the engineer as well as the AI agent.”

However, Mitchell rejected the idea that AI should replace engineers or remove human responsibility from the development process.

“AI can accelerate the work, but it does not determine the outcome,” he said, stressing that human engineers remained accountable.

FirstRand began rolling out its AI coding tools through practical training and workshops, allowing engineers to test the technology on real projects.

FirstRand Group chief digital officer Kevin Mitchell stressed that AI should not replace human engineers while talking about how the company created a core banking system that would have taken them years without AI programming. @Sowetan1981 pic.twitter.com/mif919FuXZ— Koena Mashale (@Koena_xM) August 19, 2026

He said adoption depended on giving employees time and support to understand the tools rather than imposing them from the top.

“We don’t have everything figured out. Every week, there’s a new learning that comes up,” Mitchell said, adding that the long-term benefit could extend beyond productivity and innovation.

He also said lower costs could enable financial institutions to develop and offer services to more people.

“If the cost of value creation drops, we can also enable greater financial inclusion,” he said.

Jonathan Allen, AWS executive in residence, said traditional workplaces often required employees to move between multiple platforms, databases and applications to complete a single task.

“What you need is scattered — across messaging, email, dashboard … and every tool you add creates more spaces, another login, another data source, another place where context goes to die,” he said.

“The promise was intelligence. What we actually got was a slightly faster search bar.”

Allen said agentic AI could understand an intended outcome and determine the steps required to achieve it.

“You don’t want to be the only person in this tool. You need your organisation to collectively have this connectivity,” he said.