Beyond personalization: The shift to outcome orchestration in telecom

Insights

• Telecom operators have become much better at making individual interactions intelligent, but customer value is ultimately determined by whether the operator can deliver the customer’s desired outcome.
• The next evolution of customer value orchestration is from “next-best action” to “next-best outcome.” This requires telecom companies to connect customer intent with real-time signals across network, service, billing and customer-care systems.
• Winning with AI requires a continuity layer, not another point solution. A shared process and orchestration layer can carry intent and context across organizational and technology boundaries, enabling AI agents to act across the enterprise and measure success by customer outcomes.

Telecom companies have invested heavily in AI to make customer engagement smarter. According to GSMA’s mobile economy 2026 report, customer care is the single largest area of telecom AI deployment, ahead of network operations and sales. Real-time decisioning, generative AI chat, agent assistance, and engines that recommend the next-best offer all serve one purpose: understanding each customer well enough to serve them better in the moment.

That confidence in AI runs deep across the industry. According to 2025 IBM research that surveyed 106 communication service providers across Australia, Germany, India, Singapore, the UK, and the US, 75% of telecom executives believe AI will deliver competitive advantages within three years, and generative AI is now live in customer care at 69% of telecom companies. Some real-world deployments show what this looks like in practice. China Mobile's Lingxi assistant has turned everyday communication channels into AI-powered services. Its 5G messaging channel now identifies what a customer wants and acts on it directly, like booking a hotel, checking the weather, or navigating a route, rather than redirecting the user to separate apps. More than 10 million customers use this service. Lingxi also offers online medical consultations and AI emotional-companionship calls tailored by age group, from stress relief for office workers to companionship for elderly users, with more than 60 million customers already using the service.

Another interesting example, unveiled by Reliance Jio at Reliance Industry Limited’s June 2026 annual general meeting, is its call agent. Using this agent, users will be able to bring the assistant into any ongoing call by saying "Hey Jio,". The assistant will transcribe the conversation, summarize it, and even carry out tasks such as booking a cab or a restaurant table without anyone leaving the call. As per the telecom operator’s announcement in June, Jio plans to roll it out to its subscriber base later in the year 2026.

Where intelligence falls short

A telecom company can get every recommendation right; it can provide creative and engaging customer assistants and still fail the customer if the underlying service does not hold up.

For the customer, value is simple to describe: a connection that works, a bill that matches what was promised, and a problem solved without extra effort. For the telecom company, benefit has traditionally looked different: a campaign that converted, a call that was handled quickly, a cost that stayed within budget. The two are not the same, and the gap between them is where customer experience breaks down.

Two problems stand between a smart interaction and a delivered outcome.

Service reliability

Consider a customer traveling overseas tomorrow. A personalized next-best-offer engine recognizes the travel pattern and recommends a roaming pack. But the customer's real objective is to stay connected throughout the trip, and that depends on whether roaming is enabled, whether the device is compatible, and whether a billing issue might interrupt service along the way. A roaming pack sold means little if any of these fail.

The same vulnerability shows up at a larger scale when the network itself fails. Despite real advances toward self-healing systems that predict faults, reroute traffic, and close incidents automatically, when something goes wrong, the customer often notices before the telecom company does. This shows up in a small, familiar way: A call drops, and the instinct is to switch to WhatsApp rather than wait for the network to recover. It has also shown up in far more serious ways. In 2025 and 2026, outages at major telcos disrupted emergency-call access across Australia, prompting parliamentary scrutiny and action.

Whether the cause is a misconfigured roaming profile or a dropped network connection, the customer experiences it the same way: a service that failed to work.

Fragmentation across support systems

At times of disruption or any other service-related issue, even when a customer does reach out to telecom support, the systems behind the scenes often cannot see each other. As per research, only 41% of operators have an effective end-to-end data architecture connecting their network, operations, and customer domains. A network fault that breaches a service-level agreement (SLA) on a priority enterprise account might never reach the system that tracks whether that customer is at risk of leaving. The two signals sit in separate domains, and nothing connects them. This is why a customer troubleshooting a broadband issue through an app, then a virtual assistant, then a contact center agent, often ends up repeating the same diagnostic details three times, where none of them hands off what they know to the next.

Customers experience these two problems — service reliability and fragmented support systems — as one broken journey. They are interacting with a telecom company whose individual customer engagement systems are smart, creative, and surprising, but whose overall response still lags.

Shift from personalization to outcomes

Customer value orchestration (CVO) has evolved from understanding the customer to helping the telecom company deliver what the customer actually needs. The two problems described above are symptoms of a model that stopped evolving at personalization.

CVO 1.0 Segmentation: The question was: who is this customer? Telecom companies grouped customers by demographics, products, and value, and targeted campaigns accordingly.

CVO 2.0 Personalization: The question became: what should we offer this customer? Real-time decisioning and next-best-offer engines make individualized offers possible at scale.

CVO 3.0 Contextualization: The question is: what does this customer need right now? Real-time signals from channels, devices, usage, and network experience increasingly inform the next action, rather than a static customer profile.

CVO 4.0 Intent and outcome orchestration: The question shifts again to: what is this customer trying to accomplish, and how does the telecom company make that happen? This changes the unit of optimization from the next interaction to the customer's objective. China Mobile's Lingxi and the forthcoming Jio Call Agent both point in this direction by recognizing what a customer wants mid-interaction and acting on it directly.

Reaching CVO 4.0 requires an architectural layer that can carry a customer's intent and context across every system a telecom company runs.

The continuity layer

This layer keeps a customer's objective, and everything that has already happened in pursuit of it, available to every system and every person who touches that customer next. From our earlier example, when a customer's roaming pack is active, every downstream system knows the goal is "connected while traveling," not just "roaming pack sold." When a network issue affects a customer, that same context travels with them, whether they come back a day later or escalate to a human agent. Nothing needs to be re-explained, and nothing gets lost at a handoff.

Making this work requires architecture that holds and updates that context in real time. Infosys Knowledge Institute calls this a Process Ontology: a live, shared model of how the business runs end to end. It sits above two more familiar layers — a unified data model that standardizes core entities such as customers, orders, and services, and a domain ontology that links related records and events together. The Process Ontology adds what those two layers cannot provide on their own: a real-time map of active processes, state transitions, SLA timelines, and decision logic across domain boundaries.

This cross-domain mapping is possible because of standards such as TM Forum’s eTOM framework, a reference taxonomy of business processes, including fulfillment, assurance, and billing. It gives telecom companies a common set of process definitions to align their own systems against. When a telecom company does this, an event in one domain, such as a network fault, can be understood by a system in another domain, such as a billing or customer-retention system, without custom integration built for every pair of systems. With this layer in place, AI stops working in silos. It understands where a process stands, what is at stake, and what action keeps the business within SLA, protects revenue, or maintains customer experience.

The results are measurable. When network telemetry shows degrading performance on a core node, a process-aware system identifies every affected customer, prioritizes by service tier, dispatches remediation, and applies proactive credit before the customer notices. In Infosys’ experience with deployments across the EU, the UK, and Asia-Pacific, this approach cut mean time to repair by more than half and reduced SLA breach rates from around 6% to under 1.5%. Some telecom companies are already running live systems built on this same principle: Deutsche Telekom’s RAN Guardian agent, built with Google Cloud, has cut the time needed to manage major network events from hours to around a minute since its launch in November 2025.

The industry is converging on this model from the customer-experience side too. TM Forum's agentic ODA for proactive customer experiences project, a proof-of-concept backed by a consortium including Vodafone, TELUS, Verizon, and Telecom Argentina, outlines a concierge AI agent that would carry usage, network, and journey context forward, so a customer would never have to raise an issue the telecom company already knows about. The project estimates that eliminating custom integrations this way could let telecom companies launch new operational services more than 40% faster. The concept has been demonstrated at TM Forum’s industry events but has not yet been deployed commercially by any of the operators.

What telecom leaders should do next

Building a continuity layer requires deliberate, sequenced investment across five areas.

1. Connect network and customer data in real time

Telecom companies should build a unified, live view that links customer, service, and network events, rather than leaving them in separate systems that never speak to each other. This is the foundational layer of a Process Ontology, and it is what allows a network fault to trigger a customer-facing response automatically instead of waiting for a complaint.

2. Expose core systems through open, AI-callable standards

Telecom business and operations support systems should be modernized and exposed through application programming interfaces built on open industry standards, such as TM Forum's Open Digital Architecture and eTOM process framework. Doing so, lets AI agents discover and act on core systems safely, without bespoke integration for every new use case, and reduces vendor lock-in as agent standards mature across the industry.

3. Govern AI use with clear ownership and oversight

Responsible AI requires more than good intentions. Telecom companies should appoint a dedicated owner for their data and process architecture, establish clear key performance indicators, and embed compliance into the system design itself. Infosys Knowledge Institute's research found that operators who treat this as a permanent, owned product sustain it for years, while those who treat it as a one-off project see it drift within two years.

4. Measure success by customer outcomes, not efficiency alone

Traditional metrics, such as net promoter score, churn, and average handling time, will continue to matter, but they measure components of an experience, not the outcome itself. Telecom companies should track a new set of metrics alongside them that evaluate whether the customer accomplished what they came to do, with minimum effort. A few examples make this practical (Figure 1).

Figure 1. Practical examples of metrics to track for customer intent

Source: Infosys

5. Adopt a single outcome metric across the organization

The customer outcome measures are useful to add for a detailed review, but for an overarching view, it is useful to have a single number to track progress enterprise-wide. One candidate is a customer outcome success rate (COSR), which represents the percentage of customer intents completed within the promised parameters, without avoidable effort or repeat contact. It breaks into four dimensions.

COSR = Outcome achieved x Right first time x Within promise x Minimum effort

Applied to the roaming scenario, a high COSR would mean the customer stayed connected throughout the trip, was charged correctly, and never needed to contact support, regardless of whether a roaming pack was ever sold. COSR is not yet in use at any telecom company and would need a deeper evaluation of how it works in practice at scale. It is proposed here as a direction for the industry to consider.

Autonomous networks and smarter interactions will keep advancing. The next differentiator will be whether telecom companies can understand what the customer is trying to achieve, mobilize the enterprise around that intent, and know whether the outcome was delivered. That is the shift from personalization to outcome orchestration — and from intelligent interactions to customer value.