JOHANNESBURG/South Africa: The growing use of artificial intelligence in South Africa’s business process outsourcing (BPO) industry is increasing pressure on call-centre operators to invest more heavily in computing infrastructure, according to a technology industry executive.
The warning was given by Sanjay Govender, Head of GBS/BPO Solutions at Qrent, who said the financial implications of AI adoption were extending beyond software licences to computers, servers, networking, storage and other infrastructure required to operate AI-enabled customer service systems.
Govender said the increasing deployment of tools such as real-time call assistance, voice processing, AI-supported first-line customer service and agent coaching was changing the infrastructure requirements of BPO operators.
According to him, some call-centre operators are now being forced to reconsider the specifications of computers deployed to frontline agents as AI-assisted applications place additional demands on processing capacity.
He said this could accelerate a move away from conventional workstation configurations towards higher-specification devices, including computers equipped with more powerful processors.
Another option available to operators, he said, is to keep relatively standard computers at agent workstations while moving AI processing to central servers.
While that model can reduce the need to upgrade thousands of individual computers, Govender said it could increase pressure on backend infrastructure.
AI-enabled operations require greater computing capacity, faster storage, stronger networking capabilities and infrastructure that can scale as workloads increase, he said.
The infrastructure challenge comes as global spending on technology continues to be driven partly by demand for AI-related systems.
Gartner’s February 2026 forecast projected worldwide IT spending at $6.15 trillion for the year, representing 10.8 per cent growth over 2025. The research firm also forecast server spending to rise by 36.9 per cent and data-centre spending by 31.7 per cent.
Govender said the increasing cost of AI-capable infrastructure was forcing BPO operators to consider different approaches to financing technology investments.
One option is to continue purchasing and maintaining infrastructure internally, while another is to move some AI workloads to cloud or colocation facilities.
Under the latter model, operators can obtain infrastructure from external providers rather than making large upfront investments in their own server environments.
However, Govender noted that outsourcing infrastructure does not eliminate costs, as companies would still incur recurring expenditure for hosted or cloud services.
For BPO companies handling large volumes of customer interactions, he said the choice between owning and renting infrastructure would increasingly depend on factors such as workload, scalability, latency, compliance requirements and the speed at which technology requirements are changing.
The development is also putting pressure on the traditional capital expenditure model used by many BPO operators.
Govender said buying infrastructure outright could become more difficult to justify when AI workloads are changing rapidly and hardware can become inadequate sooner than expected.
He suggested that leasing or rental arrangements could provide operators with greater flexibility by spreading infrastructure costs over time and allowing them to adjust capacity as their requirements evolve.
The argument reflects a broader change taking place in the economics of the BPO sector, where technology is increasingly becoming a significant component of operating costs alongside labour.
South Africa has established itself as an important destination for international business-process outsourcing, with operators providing customer service and other business functions to clients in overseas markets. PwC South Africa said the country’s BPO sector was experiencing continued development as providers increasingly integrate AI, analytics and automated workflows into their operations.
Govender said the growing adoption of AI meant BPO operators would need to consider infrastructure efficiency alongside labour productivity when assessing their competitiveness.
He argued that the question for operators was increasingly shifting from how many agents they could support to how much computing capacity was required to support those agents effectively in an AI-enabled environment.
The development could also make technology procurement a more strategic issue for BPO companies as they balance the need for higher computing capacity against pressure to keep operating costs competitive.
According to Govender, the growing infrastructure requirement means AI should not automatically be viewed as a mechanism for reducing the overall cost of operating call centres.
Instead, he said, part of the cost burden could simply be moving from human labour to technology infrastructure.
