More than 1 400 Absa developers are now using AI-assisted coding tools, including GitHub Copilot and Anthropic’s Claude Code, the bank disclosed on Wednesday in a statement on its first-half technology performance.
It is the second disclosure of an AI development stack by one of South Africa’s big four banks in a week. Standard Bank named Amazon Bedrock as its platform on Monday and claimed a 20% gain from AI coding tools. Absa has given a headcount but no productivity figure.
Further down the stack, the bank said more than 30 000 employees use Microsoft Copilot each month and that 4 400 staff with Copilot Pro licences are building their own automation agents, with more than 3 800 of those agents now running in production.
“Technology only matters when it changes the experience of customers and the performance of the business,” said group chief information and technology officer Johnson Idesoh, who told TechCentral’s Meet the CIO podcast in May that AI in all its forms is the single biggest topic on his desk.
The statement also discloses that Absa has consolidated customer information from 32 separate systems into one AI-enabled customer master data management platform. That sits alongside a write-down programme the bank has been running for more than a year: a R2.4-billion software impairment in FY2025, and a further R200-million written off at the interim stage. Technology now accounts for about 28% of group operating expenditure.
The customer numbers
In business banking, straight-through processing has cut small business account-opening turnaround from two days to under 30 minutes. Financial spreading for credit decisions has come down from two to five days to four hours.
Amazon Connect has cut contact centre response times by up to 21%, improved handling time 14% per call and reduced call holding time 44%. Absa’s agentic chatbot handles more than 100 000 queries a month from what the bank describes as approximately 1.6 million users — a ratio of roughly one query per user every 16 months, which suggests the larger figure counts reachable rather than active users. The business banking chatbot supports all 11 official languages.
Read: Absa writes off another R200-million in software
In debt review, Absa said AI lifted process effectiveness from 28% to 90%. Absa Rewards added 430 000 members in the half, with participation up 21%.
The bank reported group service availability of 99.97%, which over a six-month period still allows about 78 minutes of downtime, along with zero severity incidents and more than 33 000 credential compromises prevented. — © 2026 NewsCentral Media
