Artificial Intelligence in Call Centers: A Cost-Increasing Instead of Cost-Reducing Tool.

The BPO sector has adopted AI technology with the promise of increasing operational efficiency, improving customer experiences, and driving business expansion. However, within South African call centres, the situation may be quite different.

As AI tools become fully integrated into customer engagement processes, many operators are realizing that the true cost of AI extends beyond just the software license—it lies in the infrastructure necessary to support it.

The demands for processing power within modern BPO settings have surged dramatically over the last 18 months. This includes everything from voice neutralization software and real-time call support to AI-enhanced first-line service and live agent coaching.

What numerous providers failed to anticipate was the ripple effect on backend systems.

AI doesn’t operate without costs; it necessitates significant compute power, memory capacity, network throughput, low latency conditions, and increasingly costly infrastructure to operate effectively at scale.

Consequently, many BPOs are confronted with a challenging and costly choice.

One potential strategy is to execute AI workloads directly on endpoint devices.

This implies a transition from standard workstation setups to high-performance machines that can locally manage AI-integrated applications.

In practical terms, this transition is pushing a shift from traditional Intel i5 workstations to a growing preference for i7-powered devices in call centre operations.

AI-enhanced workloads are necessitating hardware updates much sooner than many previously planned refresh cycles.

Alternatively, the second option involves maintaining relatively standard endpoint devices while shifting the AI processing demands to the backend infrastructure.

In this configuration, AI applications and workloads are centrally hosted on servers, alleviating the processing needs on the user devices.

While this option mitigates the need for extensive desktop upgrades, it presents a new challenge—substantially heightened server infrastructure requirements.

This is where many BPOs are beginning to encounter financial strain.

Backend server environments that can handle AI-driven workloads require significantly higher compute density, enhanced storage performance, advanced networking capabilities, and much greater scalability than traditional call centre infrastructure.

The expenses associated with scaling on-premises server infrastructures to support these demands are rising rapidly, especially as global demand for AI-capable hardware continues to escalate.

As per Gartner, global expenditure on AI-optimized servers is surging as organizations strive to accommodate enterprise AI workloads, contributing to the overall global IT expenditure projected to reach $6.15 trillion by 2026 (https://apo-opa.co/4gTlf4e).

A third pathway many organizations are considering is offloading AI infrastructure to hyperscale providers such as Amazon Web Services or colocation facilities like Teraco.

In this setup, the infrastructure is rented rather than owned, with AI workloads managed externally and integrated into the BPO environment via cloud or hosted services.

While this alleviates the burden of substantial upfront capital investment, it introduces ongoing rental and operational costs that need careful management over time.

For certain BPOs, this can provide much greater flexibility.

For others, particularly those operating at scale with stringent latency and compliance needs, the long-term cost dynamics can become more intricate.

What is increasingly evident is that AI is dramatically altering the economic landscape of the BPO industry.

Historically, cost optimization in call centres has centered around labor efficiency.

Today, however, the efficiency of infrastructure is gaining equal significance.

The conversation is shifting from merely how many agents a BPO can support, to how much compute power is necessary to assist them efficiently in an AI-enabled setting.

This is why traditional procurement models are facing scrutiny.

Many operators still attempt to acquire server infrastructure outright through significant capital expenditure projects.

However, in a landscape where AI workloads evolve rapidly, hardware needs are constantly changing, and infrastructure costs remain unpredictable, investing large sums into fixed infrastructure is becoming increasingly precarious.

An increasing number of BPOs are considering leasing and rental models for their backend AI frameworks.

Instead of upfront purchases for costly server environments, providers can deploy infrastructure through operational expenditure models that distribute costs over time while allowing flexibility as AI needs change.

This approach also mitigates the risk of overinvesting in hardware that could quickly become inadequate or obsolete compared to traditional infrastructure cycles.

In an environment driven by AI, scalability and adaptability are proving to be more valuable than ownership itself.

The uncomfortable truth is that AI does not inherently lower operational costs within BPOs. In many instances, it is leading to higher expenses.

The distinction lies in the fact that costs are transitioning from a focus on workforce to infrastructure.

This transformation changes everything since the next competitive edge in the BPO industry may not revolve around the cheapest labor model but rather on who can afford to scale AI effectively.

*This article was written by Sanjay Govender, Head of GBS/BPO Solutions at Qrent. The opinions expressed by Sanjay Govender do not necessarily reflect those of The Bulrushes

The post Artificial Intelligence Is Making Call Centres More Expensive -Not Cheaper   appeared first on The Bulrushes.