Telcos risk ‘opex trap’ as AI costs rise: Report
Telecom companies leveraging artificial intelligence (AI) to streamline operations and justify workforce reductions are at risk of falling into an ‘opex trap’, where overall costs could rise instead of decline. Research from Bain & Company highlights that while human resources may diminish, AI-related expenditures on agents, tokens, and data could grow, potentially leading to higher total operating expenses.
Bain outlined two potential financial scenarios for telcos. In the first, AI expenses might replace 20-30 per cent of traditional costs, resulting in unchanged total costs. However, a more concerning ‘cost creep’ model suggests that if traditional costs are not sufficiently cut, AI expenditures could simply add 20-30 per cent to overall spending, increasing total operational expenses.
This warning aligns with concerns raised by Gartner, which separately predicted that 30 per cent of employees laid off due to AI might need to be rehired by 2030, likely at a higher cost. Gartner VP Analyst Tori Paulman noted that using AI primarily for cost-cutting risks big and premature reductions, impacting innovation and competitiveness.
Even with strategies like AT&T’s ‘tokenomics’ using open models to control costs, Bain indicated that even cheaper AI models can lead to substantial bills. This is due to employees discovering new uses, power users consuming tokens at scale in areas like network operations and customer care, and teams shifting to more complex, token-intensive models.
To avoid this ‘AI opex quicksand’, Bain advocates a mindset shift. The meaningful economic unit for AI should not be ‘cost per token’, but rather ‘cost per resolved customer issue, network incident, proposal generated, or software release’. The firm concluded that treating AI as a bolt-on tool, where legacy workflows persist alongside isolated AI tasks, is a recipe for cost creep.
