From readiness to successful implementation
In March 2026, Altman Solon published findings from its Generative AI Readiness in Telco survey of 109 senior executives, which revealed a gap between ambition and readiness. Six months later, this follow-up draws on C-suite interviews across six regions to identify what separates successful implementation from the rest.
AI investment is now a given in telecom. The first wave of generative AI deployments delivered productivity gains, but most telcos still struggle to translate, or attribute, these gains to EBITDA. As the low-hanging fruit of cost avoidance is exhausted, the next wave of use cases will be harder to implement and measure.
Real gains, hidden returns
The first wave of AI adoption was the implementation of AI pilots, concentrated on clear cost reduction: chatbots for low-complexity queries, average handling-time reduction in contact centers, and predictive network maintenance. The gains were real. One Latin American (LATAM) telco cut branch back-office headcount cost by 90%; another in Southeast Asia (SEA) reduced average handling time by 25% in nine months. These obvious entry points have now largely been addressed.
The problem is that most telcos cannot translate those gains into clean EBITDA gains. The cause is structural: savings in one function are absorbed elsewhere, by wage inflation, rising technology and licensing costs, and headcount that lags productivity because severance and attrition are slow. One European operator cut its workforce over several years yet saw almost no net change in EBITDA, because wage increases ate the savings. The savings are also slow to surface because telcos run financial governance around discrete, single-owner projects while AI’s gains span many budgets.
“The savings are generated, but they are not returned to the central budget.”
Chief Digital Officer, U.K. fixed-mobile telco
Altman Solon’s research found that telcos estimate an average 2% EBITDA uplift from AI. But what the figure means is less clear. It could suggest AI is underdelivering against expectations, or that, amid revenue pressure and macro headwinds, AI is helping to offset declines that would otherwise have been greater. Both interpretations remain plausible because most telcos lack the data and measurement to isolate AI’s impact.
Four traits that predict success
Four characteristics separate the telcos delivering measurable value. Technology choice barely registers.
-
Focus on a few key use cases. Winners bet on a handful of enterprise-critical initiatives with the clearest line to the P&L rather than spreading resources across dozens.
-
Prioritize selective data. The fastest movers prioritized data only where it changes the outcome.
-
Install dedicated governance and measurement. Winners assign a dedicated C-level owner with authority to set policy, allocate investment, and stand up a board to approve, fund, and kill use cases fast against KPI-backed criteria.
-
Create an agile culture and demand accountability. Adoption is where value most often stalls: one U.S. tier-1 invested early in copilots, but employees resisted, and the tools never became the default.
How winners execute, and how AI pilots die
The success of wave 2 of AI adoption, which we define as full integration and value creation, will hinge upon execution. The most effective operators do two things consistently: they make deliberate, use-case-by-use-case decisions about whether to source capability internally or externally, and they kill what does not work faster than their peers.
The capability sourcing decision is not a build-versus-buy binary but a three-way decision tailored to each use case:
-
Buy commoditized platforms where building adds no advantage
-
Build differentiation by layering proprietary data
-
Partner for speed where gaps exist, then reduce dependency as capability matures
The agentic frontier: a governance reboot
Agentic AI, where systems take autonomous actions rather than generating responses, is both the next source of value and a governance challenge most are not equipped for.
Due to its autonomy, Agentic AI requires stricter governance. First, oversight must tighten: agents that act require stricter control than tools that suggest, with a human in the loop. Process transparency is a prerequisite: you cannot govern an action you cannot trace.
Second, cost control becomes a live operating discipline, not a budgeting exercise. Agents consume continuously, so financial operations that track cost-to-serve against value realized are what keep a use case economical.
Operators who build that maturity ahead of need, before use cases force it, will scale the moment value is proven. The rest will establish governance and cost controls while the window is open.
Funding and cost control
The next wave cannot be funded the way the first one was. Cloud migration budgets are running out, CapEx deferrals have limits, and net new spend is hard to defend. The only durable source is the savings AI itself has generated, which is why operators that cannot attribute those savings will struggle to fund what comes next.
The full-transformation arc runs in years, not the 6–12-month payback vendors pitch. No operator reported material EBITDA impact in under two years, and most pointed to a 36-month horizon before the financial picture fully materialized.
For the C-suite, the question is no longer whether to invest in AI. The real test is whether the organization is structured to capture the value it creates, and whether leadership has the discipline to stop what is not working before it consumes what will.
How Altman Solon can help
Altman Solon helps telecom operators move past the easy wins of the first wave to value they can prove and scale, by clarifying where the returns actually sit, how to measure them, and how to govern what comes next.
-
Find the value and prove it. Isolate AI's real EBITDA contribution from wage inflation, volume shifts, and cost reallocations, and put baselines and attribution in place so the next funding round is gated on evidence, not advocacy.
-
Focus the portfolio and set stopping rules. Narrow to the few use cases with the clearest line to the P&L, assign C-level owners, and apply explicit kill criteria so resources concentrate where they pay back.
-
Design the operating model for autonomy. Establish the decision rights, runtime governance, and cost controls that let agentic use cases scale safely, before the use cases force the question.
About this research
This whitepaper draws on in-depth interviews with C-suite leaders at tier-1 and challenger telcos across the U.S., Europe, India, Southeast Asia, Latin America, and Australia, including CEOs, CFOs, CDOs, CCOs, CISOs, and heads of AI Centers of Excellence. It extends Altman Solon’s November 2025 survey of 109 senior executives.
