On this episode of ITWeb TV, news editor Admire Moyo talks to Robert Benvenuti, CIO of data and applied AI at Absa, about how the bank is making use of artificial intelligence to boost operations and customer experience. #AI #Absa #Banking #CIO
Absa has reduced the time taken to process debt review applications from four days to one day using artificial intelligence (AI), as the bank moves beyond experimentation to deploy the technology in live operations.
The development is among several AI applications already in production at the financial services group, which is increasingly using the technology to improve customer experiences, automate processes and strengthen fraud detection.
Speaking to ITWeb TV, Robert Benvenuti, chief information officer for data and applied AI at Absa, said the bank is beginning to see tangible returns from its AI investments.
“If you think about customers who are in financial distress, they apply for debt review, they have to submit lots of documents,” he said.
“Previously, employees had to open e-mails, read between 20 and 30 attachments and manually extract the relevant information, making the process time-consuming. That process is down from four days to one day.”
He said the improvement was not only about increasing employee efficiency but also about helping customers who are already under financial pressure avoid unnecessary delays.
Absa has achieved the improvement partly through AI-enabled optical character recognition, which allows the bank to extract information from documents more effectively.
The technology is also being piloted in customer onboarding, where it can read identity documents, passports, electricity bills and bank statements from other financial institutions.
The aim is to reduce the amount of manual document processing required when customers apply for banking products.
Robert Benvenuti, chief information officer for data and applied AI at Absa. (Photograph by Lesley Moyo)
AI to augment, not replace employees
Despite concerns that automation could eliminate jobs, Benvenuti said Absa’s approach is to use AI to remove repetitive tasks and allow employees to focus on activities requiring human judgement.
“AI is there to enable the human, not replace the human,” he said, referencing Stanford University professor Fei-Fei Li’s approach to human-centred AI.
“What we really want is our people to be doing the thinking, the decision-making, the judgement calls and the dealing with customers.”
The bank has also invested in building AI skills among its workforce. According to Benvenuti, more than 10 000 employees have completed immersive AI training, while internal initiatives teach staff how to use AI tools more effectively.
He added that Absa has established an AI Academy offering learning journeys that employees can complete in their own time.
Guardrails against shadow AI
As employees increasingly experiment with generative AI tools, financial institutions face the challenge of preventing sensitive information from being exposed through unapproved applications.
Benvenuti said Absa wants employees to use AI tools but within a controlled environment.
The bank has established an AI gateway through which more than 120 large language models have been made available to staff.
The gateway incorporates more than 40 guardrails governing how employees use the technology, including controls around data protection, bias, responsible AI use and cyber security.
“We’re a bank, right? At the end of the day, trust is everything. Security is everything,” he said.
Benvenuti added that making AI tools accessible through a governed environment helped reduce the risks associated with shadow AI.
AI tackles financial crime
AI is also playing an increasingly important role in Absa’s efforts to detect financial crime, including money-laundering, human trafficking and terrorist financing.
Benvenuti, who previously worked in the financial crime space, said the bank is using machine learning to identify suspicious activity and help investigators narrow down potential cases.
In 2024, he noted, Absa received an award at the International Compliance Awards in London for its work using AI to detect emerging financial crime risks.
The bank has also deployed an AI agent to scan e-mails for phishing attempts, identifying messages designed to trick users into clicking malicious links or surrendering their credentials.
Benvenuti pointed out that AI is accelerating both cyber attacks and the industry’s ability to defend against them.
He noted that AI models are also helping identify previously undetected software vulnerabilities, enabling organisations to address security weaknesses faster.
Caution over autonomous AI agents
While Absa is exploring AI agents for internal applications, Benvenuti warned against giving autonomous agents unrestricted access to customers’ bank accounts.
The bank does not currently permit such access and does not encourage customers to give agents control of their banking credentials.
“We tell all of our clients don’t give anyone access to your bank account, whether it’s another human being or whether it’s an agent.”
However, he expects agent-to-bank communication to become more common in future, provided appropriate safeguards are established.
He said customers need to retain control over what an AI agent could do within their accounts.
AI is not a separate IT budget
Benvenuti said Absa did not treat AI as a standalone technology investment or allocate a separate budget to it.
Instead, the bank evaluates technology investments according to the business problems they are intended to solve and the value they are expected to deliver.
“We don’t see AI as a separate bucket or a separate tool,” he said.
The bank’s broader ambition is to combine cloud infrastructure, trusted data and AI to deliver more personalised financial services across its African operations.
Benvenuti said AI could make it more affordable to customise banking services according to customers’ preferred languages, products and geographical locations.
“We see AI as the great multiplier for financial services in Africa,” he said.
