Telecom companies have an AI advantage. Can they convert it into growth?

Telecom operators know what it feels like to power a revolution and watch many of the large, new profit pools form somewhere else. 

Telecoms laid the broadband pipes that made the internet mainstream. They built the mobile networks that put the app economy in everyone’s pocket. They carried the traffic that made streaming, social media, and cloud computing part of daily life. Each of these waves of massive infrastructure investment made connectivity more essential. Each enabled powerful new businesses to grow on top of that infrastructure. And each also pushed more value towards the companies that controlled the customer experience, the software layer, or the platform. 

Telecommunications companies are well positioned to prevent AI from becoming the next version of that story—if they translate their relatively strong AI foundations into real operating and business model changes. 

The pressure to do so is mounting. Telecom operators face a familiar squeeze: data traffic keeps rising, networks require continued investment, and traditional connectivity revenues are growing too slowly to absorb the pressure. AI can help telecoms run leaner, which itself is valuable. But the bigger opportunity is to use AI to help telecoms grow differently. That approach will address what many telecom CEOs see as an existential concern: 55% of industry CEOs in PwC’s 2025 CEO survey said that their company will not be economically viable in a decade if it stays on its current path.  

PwC’s AI performance study, which surveyed 1,217 senior executives at companies around the world, in 25 sectors, helps quantify the stakes. The top 20% of companies capture 74% of AI-driven returns. The companies with the highest AI fitness levels—those building the foundations needed to scale, embedding AI into daily work across the business, and using the technology to fuel growth where sector boundaries blur—generate AI-driven revenue and efficiency gains 7.2 times as high as those of everyone else.  

Telecom operators have the raw material to compete in that top group. For years, they’ve been striving to become more like technology companies. That pursuit may no longer serve them. The AI performance study suggests some of telecoms’ AI foundations already match up with those of tech companies, but they trail the businesses creating the most value with AI in the capabilities that enable AI to scale. This represents the strategic work now before them: build out the foundations that help AI scale, and use those foundations to both run today’s business better and to build the next one. 

The AI conversion starts with foundations—the capabilities that make the tech reliable and scalable. But not all foundations are equal. Telecom companies lead the cross-sector average in the capabilities that get AI started inside an enterprise. But they trail the companies seeing the highest returns from AI (the ‘AI leaders’) in the capabilities that enable the technology to scale.

Telecoms have made significant progress in readying their enterprises to use AI. Compared with the average company, across sectors, they have stronger AI foundations in areas such as strategy, governance, and workforce engagement. That positions them better than most to capture value from their AI initiatives: PwC’s AI performance study shows that when companies with strong foundations increase their AI use, they see twice the returns achieved by those with weaker foundations. 

In some foundational areas, telecoms are on par with the AI leaders. Their governance and risk practices are especially close to those of the leading companies. Some telecoms are already turning that advantage into new business. After becoming the first Canadian telecom to sign the government’s voluntary AI code of conduct, TELUS parlayed that governance credibility into Canada’s first sovereign AI factory—a facility now fully sold out and serving as the infrastructure backbone for the country’s national AI strategy. Deutsche Telekom translated its incumbent status as a trusted domestic operator for German and European industrial customers into a new enterprise business line—the ‘Industrial AI Cloud’—positioning data sovereignty and security as commercial differentiators against hyperscalers. 

Telecoms also have an enviable set of assets in their massive troves of proprietary data. They sit on network, customer, device, service-quality, billing, location, usage, and enterprise data. PwC’s research reflects that advantage: 71% of telecom respondents said their organisation used proprietary data for AI, compared with 60% of AI leaders and only 40% of other companies.  

But telecoms still trail the AI leaders in most areas, particularly in the foundational practices that turn readiness into the repeatability that drives scale. For example, the vast amounts of data telecoms possess is often spread across silos—network systems, billing platforms, product catalogues, care records, field tools, and enterprise systems—built for different eras and different purposes. Indeed, just 46% of industry executives said employees could quickly find and use high-quality data for AI work, compared with 57% of AI leaders. And only 45% said they kept a single trusted record of critical data accessible across the business, compared with 59% of the executives at leading companies. Fragmentation may not prevent a pilot from working, but it does prevent AI from becoming repeatable and scaled. 

The operators that have addressed this challenge are reaping the benefits. For example, SK Telecom, in South Korea, built a proprietary large language model trained on its telecom-specific data assets and embedded it into an enterprise AI agent platform, now targeting over US$3.5 billion in AI-related annual revenues by 2030. 

The constraint, however, is not simply ‘legacy IT,’ as some might presume. Telecoms are close to AI leaders in eliminating outdated and costly systems. The bigger issue is that too many workflows, records, and AI components still behave as if artificial intelligence will be used only in individual projects. Telecoms trail on capabilities that move AI from isolated use cases to repeatable value—especially redesigning workflows and creating reusable AI components, such as data pipeline and integration layers. Only 38% of telecoms said they had reusable AI components centrally catalogued so teams did not have to build tools from scratch, compared with 51% of AI leaders.  

For telecom companies, the next phase of implementation should be measured by whether they can reuse, embed, and expand what works well. That requires trusted data and reusable assets, but also the management discipline to decide where AI should scale, who owns the outcomes, and how work needs to change—the leadership tasks that progress AI from a set of tools to an operating model. 

The telecommunications sector is not short on AI activity. It is short on the practices that direct it. In our study, only 38% of telecom respondents said AI had enabled operating model transformation to a large or very large extent, compared with 66% of AI leaders. 

Telecom respondents reported high levels of experimentation: 93% said they’d participated in AI pilots, above the 88% of AI-driven performance leaders and the 85% of all other companies. These respondents also reported a 20% median return on AI spending within functions, close to the 25% of the leader group. 

However, scale and enterprise-level impact remain elusive. Only about one-quarter of telecom companies said AI was scaled or embedded across their major business functions, versus upwards of 40% for AI leaders. In marketing and sales, for example, 23% of telecom companies have scaled AI, compared with 44% of AI leaders, and in support services, it’s 26% of telecoms versus 42% of AI leaders. 

As a result, telecom operators reported sharply lower enterprise-level returns on AI investment than AI leaders did. Telecoms said that 14% of their enterprise revenues were derived from AI efforts, versus 40% for the AI leaders. Results on cost takeout and efficiencies achieved were nearly as lopsided—telecoms reported that 20% of current efficiencies and 16% of cost takeout could be attributed to AI, while AI leaders reported 43% and 40%, respectively.

A high pilot count and function-level returns can mean an organisation is learning fast, but it can also mean AI is spreading faster than management discipline. PwC’s AI performance study data suggests the latter is happening. Telecoms are reasonably close to AI leaders on tracking AI’s business impact, but they trail on additional practices that turn scattered activity into enterprise change: prioritised road maps, use cases aligned to high-value business objectives, leadership accountability for AI outcomes, and portfolio reviews that force scale-or-stop decisions on AI pilots. The clearest gap is in AI portfolio discipline. Only 38% of telecoms said they conducted frequent portfolio reviews to prioritise AI initiatives, compared with 56% of AI leaders. Without a high level of discipline within the business, AI becomes a thousand experiments rather than an operating model. 

Management practices tell telecoms where to scale AI. Agents help them discover what has to change to make scaling work. The two are complementary: practices direct the portfolio, and agents reveal where the operating model breaks under real workloads.  

The strategic imperative is to deploy agents in bounded, high-value workflows where the task is clear, the guard rails are explicit, and the outcome is measurable. In those settings, agents can surface the practical barriers to scale: missing context, broken handoffs, conflicting rules, fragile integrations, or unclear ownership. A repeated handoff to a human, for example, may point less to a weak agent than to an inconsistent product catalogue, an incomplete customer record, or an integration gap.

The payoff is twofold: learn what must change to allow AI to scale, and capture value from automating and coordinating targeted work now.  

Few industries are better suited to realise that payoff. Telecom companies run many high-volume, repeatable, rules-heavy workflows across fragmented domains. A customer issue may touch network performance, billing, care, field operations, and product. An enterprise order may require sales, service availability, pricing, provisioning, and assurance. Agents can help coordinate actions across those domains that siloed teams often struggle to manage consistently.

However, telecoms haven’t yet deployed agents at the level needed to derive significant value or learn from them. They over-index on AI that assists, drafts, analyses, predicts, and recommends. But they trail at the agentic end. Only 18% said their most sophisticated AI use case could execute multiple tasks within guard rails, compared with 31% of AI leaders. And just 7% said their most sophisticated use case was autonomous and self-optimising, compared with 15% of leaders.  

The difference matters. An AI copilot can help an employee move faster. An agent can change how work gets done. But operating model change is only half the work. The other half is changing how telecoms grow.

Running a telecom company more efficiently matters in today’s environment. The ability to grow, however, is the larger opportunity, and it’s the one most telecoms haven’t yet captured. In PwC’s AI performance study, just 29% of telecom respondents reported that AI had helped them significantly transform their business model, compared with 59% of AI leaders. The gap is the largest of any business outcome the study measured.

The data on how AI leaders are using AI to grow shows the path for telecom companies. Leaders are using AI to sense emerging value pools, unlock new value from ecosystems, and reconfigure their value chains and business capabilities at roughly twice the rate of telecoms. They’re also nearly twice as likely as telecoms to use AI to collaborate with companies outside their own sector. The pattern suggests that AI leaders are not just deploying more AI—they are directing it towards opportunities arising from industry convergence. Where many telecoms are using AI to do the current business better, leaders are using AI to identify, design, and compete for the businesses forming where sectors meet. 

Emerging telecom examples show what these opportunities can look like. India’s Airtel is using AI and network-level intelligence to warn customers when one-time banking passwords may be at risk during suspicious calls. TELUS is applying trusted infrastructure and AI-enabled platforms to healthcare workflows through TELUS Health and its partnership with League. Spain’s Telefónica and Finland’s Nokia are testing agentic AI to make network APIs easier to use for cross-sector problems such as bank fraud. 

The next telecom growth pools will likely form at these sector interconnections. As a result, customers need an integrated capability, and the company best positioned to deliver it is the one that can use AI to orchestrate across boundaries. Data centres are the clearest current example.

Data centres show the opportunity 

AI is turning data centres into one of the world’s most important infrastructure markets. But data centres are no longer just real estate with servers. They’re becoming an ecosystem of energy companies, industrial equipment suppliers, semiconductor manufacturers, capital providers, connectivity suppliers, and other entities. That’s precisely the kind of market where telecoms need to use AI to see across boundaries, orchestrate partners, and move higher up the value chain.

Many are already in the mix: 68% of telecom respondents said they participated in the supply side of the data centre economy, compared with 36% of companies across sectors. But the data shows telecoms’ participation is heavily weighted towards the infrastructure layer: about two-thirds own and operate data centres, and 50% provide networking and interconnection services. Participation is lower in roles that require less capital and provide potentially higher margins. For example, just 38% of telecoms provide security services, and 30% provide operations and capacity services. 

A telecom that treats data centres mainly as facilities, fibre, or interconnection may win volume but miss the larger prize. A telecom that uses AI to forecast demand, optimise power and cooling partnerships, design edge offerings, strengthen cyber services, orchestrate cloud partners, and create enterprise AI infrastructure propositions can play higher up in the value chain. 

This is not to suggest that every operator should become a hyperscaler. But every operator needs to decide whether it wants to be a supplier to the AI infrastructure boom or a shaper of it.

The operating model work and the growth work are deeply connected: they rest on the same underlying foundations. To establish those foundations, companies need to make the following four moves, which are sequential in logic but parallel in execution.

Build for repeatability rather than from scratch

Every other recommendation depends on this one. Telecom executives should focus first on the foundations that enable AI to scale across markets, brands, channels, and functions: trusted data that’s easily accessed, reusable AI components, redesigned workflows, and common integration patterns. The goal is not to modernise everything at once. It is to make the next AI use case easier to build, govern, and scale than the previous one. 

Scale AI by domain, not by pilot

Running a pilot proves that a use case can work. Embedding AI in a domain changes how part of the business performs. The shift requires the management practices that turn scattered activity into operating leverage: prioritised road maps, business owners accountable for outcomes, and portfolio reviews that force scale-or-stop decisions. 

Telecoms should focus AI on a short list of domains where the value is material, measurable, and repeatable. Options include network assurance, field operations, digital care, customer value management, fraud and cybersecurity, enterprise sales, order orchestration, and capital planning. Each domain should have a business owner, a financial target, and a path to scale across markets and functions. If an initiative does not move a clear metric—churn, outages, field technician visits, cost-to-serve, capex productivity, speed-to-market, or enterprise win rates—it should not keep absorbing capital.

Deploy agents to change the operating model 

As mentioned above, agents do two things at once: they capture value in the workflows where they’re deployed, and they pressure-test the operating model in a way pilots cannot. The first justifies the investment; the second compounds it. Telecoms should deploy agents where work is high-volume, rules-heavy, and spread across fragmented domains. Consider areas such as network and IT assurance, complex order capture, churn outreach, serviceability analysis, SIM-swap and porting fraud, installation scheduling, and enterprise order validation. 

Use AI to move up the value chain and create new economics

This recommendation determines whether a telecom answers the supplier-or-shaper question by design or by default. Telecoms should use AI to identify new propositions, pricing models, ecosystem plays, and revenue streams—especially where connectivity intersects with cloud, cybersecurity, edge, data centres, managed services, and enterprise AI. Data centres are the current test. Telecoms are already participating through ownership, operations, networking, and interconnection. The higher-value opportunity is to orchestrate more of the stack: secure enterprise workloads, edge AI, sovereign infrastructure, managed AI services, service assurance, energy optimisation, and ecosystem brokerage.

Telecoms have powered earlier waves of digital growth but haven’t always captured their full share of the resulting value. AI gives them a chance to change that pattern, but only if they treat it as something more than a tool for improved efficiency. The operators that build for scale, use agents to guide AI implementation and change how work gets done, and point AI at emerging value pools will be better positioned to shape their next business, rather than merely optimising the current one.