GPT-4o, Gemini may soon make certain BPO jobs redundant

A leading engineering university says it is testing an AI product and expects it to be live within the next 2-3 months, resulting in its call centre ceasing operations for routine tasks

With the race to dominate artificial intelligence (AI) heating up among Big Tech companies, the latest versions, such as OpenAI's GPT-4o and Google’s Gemini, might threaten the business process outsourcing (BPO) industry’s traditional business model.

Some analysts also worry that the adoption of these nascent technologies could result in up to half of the industry’s job roles being made redundant.

The launch of GPT-4o (‘o’ stands for omni) by Microsoft-backed OpenAI has created ripples across industries, especially the BPO industry. The latest update to OpenAI’s model, which powers ChatGPT, can reason across voice, text, and vision, and can also have real-time conversational speech.

Earlier, Infosys’ chief technology officer Rafee Tarafdar told Moneycontrol that current applications are already augmenting customer service employees. In contact centres, for instance, AI is enhancing language capabilities and improving responsiveness.

“The shift will happen where they will start focusing on more value creative work, rather than the regular work, which slowly will get eliminated,” Tarafdar said.

Also read: Infosys CTO Tarafdar’s advise to engineering students: Master AI to stay ahead in the game

An HR expert said though cent percent of junior-level executive jobs will not be automated by AI, approximately half of the job roles might become redundant. He added that BPO companies will analyse and invest in employees who have the capability to learn.

“People who are not able to learn, who are not showing the intent to learn, and who are not able to learn, will have to be let go,” Aditya Mishra, MD and CEO of staffing firm CIEL HR, said.

The cost factor

Early AI adopters in the sector could significantly reduce costs and enhance service delivery, thereby gaining a competitive advantage. But companies that resist integrating AI may find themselves at a disadvantage, much like banks that failed to adapt to computerisation in the past, Mishra added.

This points toward the urgent need to reskill employees in newer technologies from transactional to more analytical and decision-making roles, says Kapil Joshi, deputy CEO of Quess IT Staffing. This reskilling is not optional but essential for businesses aiming to balance AI integration with human expertise, Joshi explained.

At the very heart, where students are nurtured, a leading engineering university says it is testing an AI product and expects it to be live within the next two to three months, which would result in its call centre ceasing operations for routine tasks. The move, therefore, will save costs for the university.

“I don't need to put a person there, I just need to put AI, train it on my database and on my information. Queries could be answered by AI better and more efficiently,” said Aman Mittal, vice president, of Lovely Professional University.

This model is scalable across the BPO sector, where AI can manage basic tasks, freeing human agents to focus on strategic, creative, and emotionally complex roles.

Also read: India's youth needs to be AI-ready to avoid huge dislocations, says CEA Nageswaran

Limitations levelled

On the other side of the argument, TeamLease Digital’s chief executive officer Neeti Sharma notes that there are inherent limitations to what AI can achieve. While AI excels at handling routine tasks, it struggles with scenarios requiring emotional intelligence and nuanced understanding, she said.

For instance, calming an irate customer or resolving complex issues often necessitates a human touch that AI simply cannot replicate. This irreplaceable aspect of human interaction remains a significant barrier to the complete automation of customer service roles.

Although the GPT-4o app has a new Voice Mode that allows voice conversations with ChatGPT, the effectiveness is yet to be established by analysts and experts.

Sharma adds that because these large language models originate from the West, there would be many geographic and demographic biases. An example of this is that the colour red is auspicious in south-east Asian countries but the same may not be true for other countries.

“So just imagine a situation where somebody in South Korea uploads a picture and this model says it’s a bad place, aura, or negative energy. So, that bias mitigation has to get built-in and it is not easy to build these into models,” Sharma said.

Is cost saving a myth?

Additionally, the promise of cost savings through AI appears compelling at first glance. However, this perspective overlooks significant hidden costs, another expert argues. “Because it's not a code, which is like a one-time investment. Every time you're putting a query to it, there is a cost to that query,” says Satya Gupta, president, VLSI society of India.

“When you put a question to any GPT engine, whether it's Gemini or ChatGPT, there's a cost to it because it's using the hardware, it's using the energy,” Gupta added.

As AI systems become more sophisticated and power-hungry, the energy costs alone could rival, if not exceed, those of human labour. “Earlier the energy consumed by a computer system was not a huge concern, but the way AI is guzzling energy, somebody has to compute the cost of let's say… how much is the cost of (using AI) versus the cost of some human who is doing the job,” Gupta said.

Nonetheless, some experts are optimistic that the deployment of capabilities such as in GPT-4o or Google’s Gemini will have a larger and wider impact on manpower plans and budgets in the near term. “The impact will be positively felt at employees' desks when they start having GPT-powered digital twins working alongside and undertaking heavy lift/labour-intensive tasks,” said Prasadh MS, head of workforce research at Xpheno.

He adds that the service capabilities of the sector will widen with the inclusion of GPT-4o-like capabilities.

Also read: The importance of ChatGPT 4o