Statistics about return on investment (ROI) from AI vary, but one thing is clear: businesses are struggling to turn a profit on the tokens they spend.
According to a summer 2025 MIT report, 95% of generative AI pilots were failing. A Google Cloud survey published in July 2026 found that 26% of respondents are seeing returns accelerate year on year.
That’s problematic, experts say. “Telecom companies need to account for every cent,” pointed out Giancarlos Soares, Field CTO Latam at BMC HELIX, during a panel at Futurecom. “And they also need to think of the end user, who is counting on a good experience.”
The Google survey points to a potential solution to this problem. Respondents elected AI agents as key tools for achieving ROI in this field, with 94% reporting that AI agents are contributing to cost savings and revenue, with a modestly higher weighting towards savings.
Playing with AI Agents
A good example is Brazilian Operator Algar Telecom. They now have 142 large-scale AI agents running, most created by employees to optimise their own workflow.
“We see AI as a process accelerator,” explained Enock Cabral, Algar’s Head of Automation and AI. “We believe this is much more about a cultural transformation combined with the acceleration brought by technology [rather than the technology for its own sake],” he added.
Instead of the common “trial-and-error” approach, Algar decided to vet each AI tool, from Canva to Claude, to ensure enterprise-level security and let the employees use them more freely. But not without limits.
“We established that every AI agent should be created under a positive business case project [that would likely generate revenue],” Cabral shared. “It was a hard endeavour, but otherwise people would make these agents anyway in a Shadow AI environment.”
Joining Forces
One common position among experts is that people and internal processes matter more than the technology itself when it comes to ROI using AI.
“What we see is that bringing the tools and the IT team closer to the business area is important. However, we need to work with positive business case KPIs,” observed Gustavo Duarte, VP Americas at Etiya.
According to him, this approach compels employees to create agents more likely to succeed and, ultimately, generate ROI. Like Fizz, a Canadian operator that works with Etiya.
“Although they adopt several agents, all are very well thought out, planned, and measured,” Duarte explained. “IT and the business area working in tandem make all much easier.”
Towards an Autonomous AI Era
The AI agents boom will drive us to the era of autonomous networks. At least, that is what Giancarlo Soares believes.
“When we talk about AI, we talk about an ideal world of autonomous networks,” he said. This world may still be some distance away, but the advantages for companies promise to be substantial.
“An autonomous network lets you be more agile, efficient, and deliver a better experience to the client.” By the sound of it, it seems that the AI agent era will be welcome, especially if it comes with some good ROI.
