Beyond the Pipe: Telcos Need a New Revenue Playbook for the AI Era

For two decades, the telecom growth story has followed one arc: build the network, sell the connection, defend the ARPU. That arc is flattening. Wireless connectivity is now a mature, commoditized product in most developed markets — and capital intensity keeps climbing while the revenue line struggles to keep pace.

At the same time, AI is creating an entirely new category of economic value, spanning compute and application — measured in tokens, inference cycles, and autonomous agent interactions — growing faster than connectivity revenue has in years.

"The question every telecom leadership team should be asking is whether they'll help define the business models for this new value chain, or simply carry the traffic for someone else's margin."

This isn't a call for telcos to become AI labs. It's a case for revenue diversification built on assets telcos already own: infrastructure, identity, and the most durable customer relationship in consumer technology. Below are three near-term opportunities that hold the promises — where the money actually is, what it takes to capture it, and the hurdles that will need to be overcome to achieve the transition.

The Diversification Imperative

Revenue realization per gigabyte has been falling for years as consumption outpaces pricing power. Meanwhile, network capex – 5G, fiber, now AI-ready infrastructure – keeps rising. This creates a margin squeeze that operational efficiency alone can't fix.

Key stat: Enterprise AI spending is now the single largest category of enterprise digital-transformation investment through 2030 – larger than 5G, cloud infrastructure, or IoT combined.

"For telcos, AI purely as an internal efficiency tool will capture a fraction of the value on the table. Treating AI as a new revenue architecture is where telcos can redefine their growth trajectory for the decade ahead."

Three opportunity areas stand out as the clearest, most actionable starting points.

1️) Real-Time Journey Prompts: One Intelligence Layer, Omni-Touch Experience Optimization

This isn't a single chatbot – it's a shared intelligence layer sitting underneath three different frontline experiences at once: the digital customer self-serving on the website or app, the retail agent helping a walk-in customer, and the care agent on a live call. All three run the same underlying mechanism in real-time: continuously synthesizing the knowledge base into ready answers, monitoring the live interaction for hesitation or friction signals, and surfacing the next-best-action before the user has to ask.

Key stats:

·       Real-time agent-assist AI reduces average handle time by 27% by eliminating manual knowledge-base searching mid-interaction

·       Gartner projects "Connected Rep" / expert-assist technology will lift contact-center efficiency by up to 30% by the end of 2026 — without replacing a single human agent

·       Acting on hesitation signals before abandonment generates roughly 30x the ROI of traditional post-abandonment recovery, which sees open rates below 20%

"The mechanism is identical whether it's a customer stuck entering a credit card or an agent mid-call with a frustrated customer: watch the live signal, synthesize the knowledge base into an answer, and deliver it before the moment of friction becomes the moment of abandonment."

The real frontier, though, is bigger than any one of these tools. Most "unified customer journey" conversation today is about marketing personalization – stitching together browsing data for better targeting. Almost no one is solving the harder problem: real-time resolution continuity, where a customer's struggle on the website is instantly solved, or (if the customer chooses to) visible to the retail agent or care agent they talk to next – no re-explaining, no lost context. That's a genuinely greenfield space, gated less by ambition than by two real constraints: the technical maturity to unify live interaction data across systems never built to talk to each other, and the privacy discipline to do it with clear customer consent. It's a bet that is increasingly gaining relevance with AI advancements, and governable data and privacy constraints.

Takeaway: The winning move isn't better chatbots or better agent scripts – it's collapsing customer self-service, retail assist, and care assist onto one real-time intelligence layer, closing a resolution gap the industry hasn't even started calling by name yet.

2️) The Token Economy: Monetizing Compute, Not Just Connectivity

The least obvious opportunity, and arguably the most important. Telcos have spent decades monetizing bits. AI introduces a new unit of value entirely: the token, the basic unit of computation behind every AI query or agent action. This shift is no longer theoretical.

Key stats:

·       Operators in leading Asian markets have already launched commercial consumer plans priced by token volume

·       National-level daily token consumption has grown more than 1,000x in under two years in the fastest-moving markets

·       Global token consumption is forecast to grow more than 20x by the end of the decade as AI agents move from novelty to default infrastructure

"This isn't about telcos building foundation models. It's about position in the value chain – producing inference near the edge, transporting it efficiently, and packaging it into services, rather than carrying someone else's AI traffic at commodity rates."

The asset that matters: Physical network infrastructure – central offices, fiber, edge real estate – is exactly the distributed compute footprint the AI industry increasingly needs, closer to users than hyperscaler data centers.

Takeaway: The pricing model and technical standards for AI-token services are being set right now, by whoever moves first. Waiting doesn't preserve optionality – it cedes the standard to someone else.

3️) Super-App Bundling: Wireless as the Gateway, Not the Product

Asia's leading operator-led "super apps," where a high-frequency core service – connectivity – became the gateway to cross-sell commerce, content, and financial services under one roof.

Key stats:

·       One major Asian operator's digital-services revenue is growing meaningfully faster than its core connectivity revenue

·       ARPU has climbed steadily as customers adopt bundles combining connectivity with entertainment, cloud storage, and AI-powered features

"Most digital products spend years and real marketing budget earning a recurring, trusted relationship with users. Telcos already have one – mandatory, monthly, multi-year, touching nearly the entire customer base. The opportunity lies in converting share of a wallet relationship a competitor would have to build from zero."

Takeaway: The bundling playbook isn't unproven – it's a repeatable model that needs a rethink on the new bundle strategies and how to build partnerships to scale it as a viable business model.

The Hurdles That Will Separate Movers from Watchers – No credible strategy skips the hard part:

·       Legacy technical debt – billing and network-ops systems weren't built for real-time AI orchestration or token-level metering

·       Organizational silos – network, product, CX, and pricing functions must co-design one customer-facing product, not four separate ones

·       Trust and regulatory exposure – every play here leans on data and identity sitting inside tight privacy and telecom-secrecy regulation

·       Cultural resistance – reframing the core product from bandwidth to compute/intelligence requires a genuine mental-model shift at the leadership level, not a bolt-on

·       Competitive timing risk – some markets already have commercial token products and super-app ecosystems; waiting cedes the model to whoever moved first

The Bottom Line

The telecom industry has weathered commoditization before – long distance, SMS, data plans – and the operators that came out ahead each time found a new axis of value before the old one fully collapsed. AI is that axis now.

"The operators willing to treat infrastructure, identity, and customer relationships as a platform – not just a utility – will define the next decade of industry economics. The ones who wait for the business case to become obvious will be buying their way into a market someone else has already priced."

The strategic frameworks above are directional – the real value is in translating them into a sequenced, capital-efficient roadmap specific to a given operator's assets, regulatory footprint, and competitive position. If you're a telecom leader weighing where to place these bets first, that's a conversation worth having and you are ahead of the curve – for now.

Portcos Are Being Scored on AI Maturity. Get Ready for the Conversation.

My earlier post on AI Maturity made the case that AI maturity is now priced into exits. This one dives into what buyers are specifically evaluating, and what it actually takes to build a position that survives the scrutiny.

In 2025, 28% of global M&A activity was AI-related, 76% of buyers are already using AI in their own due diligence, and 83% expect it to improve post-merger integration. The buyers across the table from your Portcos at exit are running structured AI maturity assessments. The question is how to prepare your companies for that conversation.

What the Assessment Actually Covers

A structured AI maturity evaluation scores a company across several key dimensions: AI strategy alignment, use case deployment depth, governance and decision rights, data infrastructure quality, talent and capability, technology architecture, and ethical and regulatory compliance.

The distinction that matters most is that AI maturity is not a measure of software purchased. A company with three well-run, well-governed use cases embedded in the revenue model is well ahead of one focused solely on deploying the latest technology. Maturity measures whether AI is changing how the company works and showing up in results. Buyers know how to tell the difference. And they are pricing it accordingly.

Why Mid-Market AI Programs Stall Before They Can Be Scored

Enterprise organizations institutionalized AI discipline through Centers of Excellence — dedicated teams with cross-functional mandates, executive sponsorship, and the organizational staying power to move from pilot to production at scale. That is the north star model: a centralized, accountable function that owns prioritization, governs execution, and measures outcomes against business ambitions — not just technology adoption.

For $20M–$500M PE-backed companies, that discipline is rarely formalized. The operating partner is stretched across a portfolio, and the portco CFO is under pressure to show AI progress without a clear path to begin. The result is point-solutions adopted function by function, without a connecting logic or shared accountability — progress that doesn't compound and doesn't tell a coherent story at exit.

The answer is not to build a COE. It is to execute with COE discipline: a smart, sequenced AI roadmap with clear ownership, governed delivery, and a measurement framework calibrated to the company's EBITDA goals. That is an achievable version of the north star — and precisely the kind of structured approach that translates enterprise-scale AI experience into mid-market results.

Four check-points for a Credible AI Roadmap

For a company heading toward exit in the next two to four years, the real question is if AI investments create a narrative with and EBITDA attached. What builds that credibility can be broken into 4 buckets:

• A scored baseline. A structured AI maturity score to evaluate, benchmarked against industry peers, and can outline the before-state of the value creation story – a verifiable starting point.
• A prioritized use case portfolio. Ranked by business impact, not technical novelty, with ROI modeled at the workflow level before any build decision is made – workflow level models, not category-level estimates, survive diligence.
• A named owner with governance that holds. The use cases that reach production are the ones with a named owner, pre-defined success criteria, and accountability that outlasts the launch energy – the structural gains are easy to show and defend.
• An exit narrative connecting AI to EBITDA. This needs to be built over the hold period, as in a claim that AI drove margin improvement means something supported by two years of documented measurement. 

A Note on Where I've Seen This Work

Much of what I've described above is the framework I've been applying through kriAItiv — starting with a structured AI maturity diagnostic benchmarked against industry peers, translating that into a prioritized roadmap with EBITDA impact modeled before a dollar is spent, and staying engaged through execution until the gains are measurable and the exit narrative holds up.

The firms that get this right, start at acquisition, so there is a recorded baseline that anchors the value-creation story at exit. 

AI Maturity Is Now a Valuation Variable. Most PE Portfolios Aren’t Ready.

Companies at the highest "AI maturity" levels traded at a median revenue multiple of 31x vs., 13x for those at the lowest (according to McKinsey’s 2026 study on 471 PE-backed companies across 31 industries). The report also states that AI maturity has to extend beyond productivity (level 1) into revenue maximization (level 3 and above) to truly make an impact – wider embrace of AI results in more than twice the median revenue multiple when using AI only for productivity. 

The study has wide implications for the mid-market companies and PE investors that are shifting focus to AI as a valuation criterion, both for investments and exits.

Productivity AI Does Not Move the Multiple

The most common AI applications in PE-backed companies today — automating invoices, summarizing reports, drafting communications — are level one and level two activities. They reduce cost. They do not move the multiple, and per McKinsey, the valuation premium only becomes material at level three, where AI is embedded in the product or service itself. To take advantage of AI, companies need to think Revenue AI - transformations for growth, not just for efficiency.

The question I ask every Portco leadership is simple: which workflows in your business touch revenue directly, and what would it take to put AI there? In telecom, AI-driven churn prediction models are identifying at-risk subscribers and triggering personalized retention offers, resulting in 10% - 25% improvements across churn and lifetime value. In retail, Zara's use of AI embedded into production and distribution decisions, using real-time sales and social trend signals, resulted in faster inventory turn-rate and lower markdowns than competition. Both are using AI embedded in the revenue cycle, directly improving the margins.

Why Mid-Market Companies Are Especially Exposed

Large organizations have been building the foundations – clean data, insights & use cases, cross-functional impacts, governance and orchestration mechanisms – for such scaled AI applications in the revenue cycle. Most mid-market companies have none of that. Fragmented systems, inconsistent data, manual workarounds are just some of the foundational problems that mid-markets face in AI adoption and execution.

Three success points, common across industries, can help reverse the trend:

• Insights on existing data used to scope use cases based on business outcomes.
• ROI modeled at the workflow (use case) level.
• Named point of accountability to orchestrate roadmaps, governance, delivery and measurement.

I have implemented these at enterprise scale – targeting top customer segments for personalized offers, preferred call routing, digital self-service that saves a support call, are just some of the targeted use cases that drive revenue and engagement KPIs for any customer-focused business. By applying these success points to the top performing core areas of their business, mid-markets can exploit the AI value at level 3 implementation (margin driver per McK study).

The Hold Period Is Not Forgiving

The implied capital cycle for buyouts is now approximately seven years. Assessing AI maturity early at acquisition and adopting “success points” can reduce that capital cycle and provide the runway to accumulate the benefits of implementing an AI program – scored baseline, prioritized use cases, governance structure, and a documented track record a buyer can verify. The firms that start this work pre-acquisition/year-1 are better positioned to drive margin improvements and compress the hold period, without sacrificing the investment goals.

The single most expensive mistake I see is treating AI maturity as an amalgamation of AI productivity tools adopted during the hold period. For example, building a real-time customer journey support engine (for digital, retail and support channels) mandates diligent prioritization, model training, content targeting and governance framework. Revenue margin use cases require development and consistent proof-points over the hold period.

An AI maturity baseline taken at acquisition – scored against industry peers, with the highest-impact use cases prioritized and resourced from day one – is what defines the roadmap for the operating partner to build and scale inherent value through the hold period.

What the Market Is Actually Paying For

BCG's AI adoption research puts the performance gap plainly: companies in the top quartile of AI maturity are 2.3 times more likely to hit their value creation targets than lower-maturity peers. And only 11% of PE investors explicitly link AI progress to exit narratives – that is real value left on the table.

The narrative gap is the most solvable, yet most overlooked. In my experience, I have seen that data exists – it just needs to be organized, developed into use cases, and executed at the workflow level in order to realize the upside. I have helped structure this value-chain through: an AI maturity score at acquisition, workflow-level ROI, and an exit narrative that connects AI activity to verifiable EBITDA lines.

The data is clear on what the market is pricing. The question for every PE operating partner is whether AI maturity has been baselined at acquisition – and if the remaining hold period can accommodate the required AI transformations.