Your AI Journey: Finding the Use Cases That Matter
Improving operational efficiency with AI is non-negotiable in telecommunications. Discover how targeted AI use cases can deliver immediate value.
Analyst insight
18 June 2026
Author
STL Partner's Managing Director of Research, Amy Cameron, on where AI in telecoms stands today: cost savings, revenue growth and trusted, well-governed AI.
Analyst insight
18 June, 2026
Author
Reflections on Amy Cameron's keynote at the Polystar User Meeting 2026.
Widening the Lens on AI in Telecoms
By the time Amy Cameron, Managing Director of Research at STL Partners, took the stage in Lisbon, the Polystar User Meeting had already heard CPO Thomas Nilsson on where AI fits into the assurance roadmap. Cameron's job was to widen the lens. She came armed with two pieces of fresh research: a survey of operators, vendors, and analysts conducted the week before at FutureNet World, and a structured analysis of every agentic AI announcement made at this year's Mobile World Congress (MWC).
She has spent years analyzing AI in telecoms - long before it became the industry’s defining theme. Her opening observation was telling: the rooms of people working on this topic used to be much smaller.
Customer experience leads. Network deployment lags. Governance is the surprise.
The survey mapped AI penetration across operational domains. The headline pattern was familiar - customer experience and business support systems (BSS) at the top, network deployment at the bottom. Customer-facing systems moved to the cloud first, so AI followed naturally.
What was less expected, Cameron said, was the maturity gap on AI governance. Very few respondents could describe what STL Partners considers optimized governance: clear policies, a centralized inventory of deployed models, regular audits, and defined ownership for monitoring drift and bias. Most operators sit somewhere in the middle, with policies in place but inconsistent oversight.
That matters more than it might sound. Cameron's data showed a stronger correlation between overall AI maturity and governance maturity than between AI maturity and cloud platform maturity. Cloud platforms make AI deployable. Governance is what makes AI trustworthy enough for operators to take humans out of the loop. Without that trust, autonomous operations remain an aspiration.
The MWC Reality Check
For her second data set, Cameron and her team scraped every agentic AI announcement from MWC - roughly 250 of them - then filtered out the ones too thin to analyze. They were left with 64.
The live deployments share a profile: batch data rather than real-time, narrow subfunctions rather than full processes, and enough self-governance that escalation paths are clear. Trials and pilots, by contrast, are pushing into real-time data, multidomain orchestration, and end-to-end automation - the harder territory where Polystar's customers are increasingly focused.
The agent counts from a few large operators tell their own story. Orange has spoken publicly about 18,000 AI assistants in internal use. AT&T claims roughly 400 deployed agents. SK Telecom is reportedly approaching 2,000. Most of these, Cameron suggested, are rules-based agents - useful for pulling and placing data, but a long way from genuine autonomy.
Cost is real. Revenue is the prize.
Where AI is delivering, the savings are measurable. In the survey, 22 percent of respondents reported operational cost savings of more than 10 percent, and 20 percent reported similar savings in customer experience. Cameron noted an important caveat: cost impact in customer experience is harder to quantify because improvements in mean time to resolve don't translate cleanly into dollars saved on churn.
But cost savings, she argued, are the floor - not the ceiling. The chart that landed hardest in the room compared global telco performance against the cloud and tech companies that emerged in the 2010s. Between 2019 and 2024, telco revenues grew 11 percent. The cloud cohort nearly doubled theirs. Over the same period, telco market capitalization declined by roughly a third, while the cloud players grew theirs by around 140 percent. Most of that uplift arrived after the launch of generative AI in 2022.
Her framing of what comes next was sharp. Operators that fail to engage seriously with automation and AI risk being sold off for parts. Operators that embrace the InfraCo model - building and owning the network infrastructure that others use - can deliver steady, on-demand operations. Only a small group will achieve real growth - those that build something new from their unique assets and capabilities, beyond connectivity itself. She pointed to Elisa as one of the operators that has done exactly that.
Why This Matters for Assurance
Cameron's closing point connected the dots. The revenue opportunity in differentiated connectivity is not concentrated in any single killer use case. It is spread across private networks, network APIs, edge, network slicing, broadcast, emergency services, and more - each with its own service-level agreements (SLAs) and its own commercial model, from subscription to usage-based to outcome-based.
You cannot operate that mix manually. The AI investments that operators are making in assurance, optimization, and orchestration are not a side project. They are the foundation that determines whether new revenue models are commercially viable at all.
That is the link between the work Polystar is doing with operators today and the growth they are trying to unlock tomorrow. It was a useful reminder, delivered by someone with the data to back it up.
The STL Partner reports referenced in the article:
FutureNet World survey
MWC: Agentic AI in telecoms: What is live now and coming next?
The telecom industry in the AI era: How telcos can follow the money
