By Mervyn Pretorius, Group CTO, CCI Global and Cobus Pretorius, SVP: Development, CCI South Africa

Little more than a year ago, AI was widely heralded as the absolute gamechanger of business process outsourcing (BPO) – and as a potential wrecking ball for employment in the industry. AI was viewed as a universal solution poised to revolutionise every facet of our industry, but this was balanced by concern that such a powerful technological step forward would displace millions of agent employee positions, resulting in job losses across the industry.

Today, the narrative has matured. Organisations have moved beyond the hype, adopting an approach to AI grounded in pragmatism, focusing on targeted, high-impact use cases and recognising that AI’s true value lies in solving specific business problems, not in sweeping transformation.

Lessons from public AI deployments

Recent months have brought some high-profile public failures in the LLM landscape – “meltdowns,” hallucinations, and visibly rocky transitions between AI model versions. These incidents have exposed the risks and limitations of large language models (LLMs) in uncontrolled environments. Attempts to impose stricter guardrails have sometimes backfired, revealing new complexities. While AI continues to develop and improve at an accelerated pace, these setbacks remind us that there is still much work to be done to develop truly practical, reliable AI at scale.

The visibility of AI’s shortcomings has transformed expectations. The industry has learned that AI will not instantly solve long-standing problems, and that every deployment must be rigorously tested in real-world conditions. The gap between promise and practical reality is now front and centre in every strategic conversation.

As organisations try to reckon with finding value in AI outputs, the unintended consequences become clear. Overly restrictive controls can stifle innovation and even introduce new risks, as seen in recent efforts by LLMs to manage messaging and engagement. The lesson is that control must be balanced with flexibility, and that human oversight remains essential, especially when the stakes are high, such as in financial transactions or regulated industries. What’s clear is that, certainly for the foreseeable future, AI is an enabler – not a replacement – for human intelligence.

AI in the enterprise: a different landscape

Enterprise AI operates in a fundamentally different context. Here, the scope is narrower and the environment is more controlled, allowing for reliable, impactful implementations that offer significant benefits to business. In BPO, the greatest value is realised when AI is applied to well-defined, repetitive tasks where accuracy, efficiency, and consistency matter most. Human agents are not substituted by AI – instead, their output is augmented by carefully-judged automation of rote or repetitive tasks.

Consider after-call summaries, a mundane but critical task. Historically, agents spent valuable time documenting calls, a process fraught with inconsistency and inefficiency. AI-driven automation has revolutionised this workflow, delivering accurate, standardised summaries and freeing agents to focus on customer service. The impact is significant: improved operational efficiency, reduced costs, and enhanced customer experience. This is a prime example of AI as a focused problem-solving tool, not an all-purpose panacea. In this workflow, humans assume supervisory and quality assurance roles, leveraging creativity and judgment where automation may fall short.

We will increasingly see AI technology being leveraged to simplify and streamline back-end processes, with Agentic AI being a key driver of this retooling of workflows. This is not something the customer ever sees, but not only does this automation bring tremendous value to the business, it frees up human agents to deliver high-value customer service, driving tangible benefits for clients in terms of brand loyalty and frictionless customer experience.

Data, data everywhere

One of the outcomes of deploying AI in the BPO industry is a vast surge in the capturing of data. AI-driven outputs such as call transcripts, translation tools, sentiment analyses and many other processes are massively increasing the accumulation of data where previously, the speed at which humans work was a limiting factor. This is the next big challenge for the industry.

Approaches to data management vary widely across the BPO industry, with some providers adopting mature, privacy-focused practices, and others lagging behind with unnecessarily large data lakes. As we move forward, ethical considerations around data privacy and management, along with regulatory compliance requirements will require organisations deploying AI to ask themselves critical questions about the data they are capturing, and how it is sorted, stored and used.

There are valid concerns about the commoditisation of customer data, and there is a need to distinguish between data that genuinely improves the customer experience, and that which is exploited for commercial gain. It will require some different thinking for companies to use their data responsibly and ethically, for the purposes of truly improving the customer experience. In pure business terms, regardless of how it’s acquired, holding onto too much data introduces risk with no upside, and this is something the BPO industry must contend with in the near future. 

Embracing AI as a powerful, targeted tool

The journey from hype to practical implementation has been instructive. For BPO leaders, the message is clear: embrace AI’s strengths but remain vigilant to its risks. Success lies in identifying specific problems, applying technology intelligently, and supporting human activity – not in “boiling the ocean” or expecting any technology to replace human activity wholesale with no oversight. The most powerful AI deployments will be those that deliver tangible value, build trust, and enhance – not replace – the human element.