- Telcos and enterprises alike are still sorting out if AI is worth the money
- For AT&T and Windstream, the tech is already delivering plenty of ROI
- Incorporating reusable parts - like APIs - is helping AT&T make the most of its AI investments
Two years into the mainstream artificial intelligence (AI) craze, one key question still looms large for many enterprises dabbling with the technology: is it worth it? That question is especially relevant for telcos, which have been looking for ways to cut costs and boost revenue in a tough market environment. The answer, at least according to AT&T and Windstream, is yes.
“While it is still early, our GenAI integrations have delivered exceptional results. So far, we’ve achieved an in-year return on investment for GenAI of more than two times,” Andy Markus, chief data officer at AT&T, told Fierce. “With 55 use cases currently in production, we also expect GenAI to contribute to our recently announced $3 billion in run-rate cost savings' goal over the next three years.”
AT&T previously told Fierce it’s using generative AI for a range of applications, leveraging it for everything from document summarization to copper retirement planning and automation. So, where is it seeing the most returns? Software development and customer care, Markus said.
“These areas have significantly benefited from the efficiencies and enhanced capabilities provided by generative AI,” he explained. “Additionally, we are beginning to see returns in contract management, network optimization and sales coaching insights.”
According to Markus, there aren’t any areas where its GenAI investments have “truly underperformed.” But there are projects that, due to their nature, take a bit longer to yield their full potential, he added.
Meanwhile, Stephen Farkouh, CIO at wireline telco Windstream, indicated it is seeing returns of anywhere from 9x to 12x on use cases on WinGPT, a specialized tool it developed based on the large language model that powers ChatGPT. And like Markus, he said it hasn’t really seen AI underperform ROI expectations.
In terms of applications, Farkouh said Windstream is working on applications spanning the employee experience (EX), customer experience (CX), field operations, network optimization, software development, product differentiation and predictive analytics.
While it sees “clear benefits” across all of these, he added “the impact of Generative AI on the employee experience — which, in turn, directly enhances the customer experience — has delivered the highest return on investment.”
And, Farkouh said, things are only expected to get better from here. “I anticipate that our current applications of Generative AI will achieve a twofold improvement in ROI” in 2025, he noted.
ROI insurance
According to Ishwar Parulkar, chief technologist for Telecom and Edge Cloud at AWS, operators spend anywhere from thousands to millions of dollars training AI on its infrastructure depending on what kind of workloads they’re running. Fine tuning is on the lower end of that range, while pre-training is on the opposite end of the scale.
That matters because while telcos can certainly get started with AI at a lower cost using general purpose large language models, Parulkar said that as they progress along the maturity curve, they really ought to be fine tuning using their proprietary data to better apply the tech to high-value workflows.
So, after spending big how exactly can operators ensure that they’re getting the ROI they need from AI?
Parulkar argued that by “prioritizing the application of fine-tuned models to high-value workflows, telcos can maximize their GenAI investments and achieve transformative results.”
Markus said from AT&T’s perspective, it’s all about using – and reusing – a common framework or platform. “This allows us to reuse components such as APIs and patterns and turn common use cases into self-service solutions,” he explained. And having a mature application evaluation framework doesn’t hurt either when it comes to making sure it’s chasing the right use cases.
You can read more on how AT&T is tackling that, here.