From Calls to Revenue

The market is moving beyond isolated experiments. STL Partners' August 2026 Telco generative AI adoption tracker records 870 publicly disclosed projects across 118 operators worldwide and puts increasing attention on productisation. At the same time, the commercial value of traditional voice is under pressure. Ofcom's Q1 2026 market data shows that UK mobile-originated call minutes fell by 4.6% year on year, while fixed-originated call volumes declined by 14.9%. 

This is not the end of voice. It is the end of minutes as the value proposition. The next opportunity is to turn voice into an intelligent service layer that customers can use without changing devices, downloading another app or learning a new interface. 

That was the focus of the recent Tallence and STL Partners webinar, From Calls to Revenue: How Telcos Can Monetise Network-Native Voice AI. Three conclusions stood out. 

Most consumer AI propositions are built around an app or a specific device. That gives handset and platform providers a strong position: the assistant becomes the interface between the user and every service they want to reach.

Telcos have another route. When Voice AI is embedded in the network, the line itself becomes the interface. Customers can access the service from the phone they already own. There is no separate download, no dependency on a premium handset, and no new behavior to learn.

The difference goes deeper than convenience. A network-integrated service can work with operator-controlled capabilities such as routing, signaling, identity and service history, subject to the applicable permissions, privacy rules and governance. Device-based assistants usually see only an audio stream and the information available on the handset.

This creates three advantages that are difficult for app-based competitors to replicate:

The strategic question is therefore not whether telcos can copy the latest assistant feature. It is whether they can use the network to create a service that an app alone cannot deliver. 

Telecom customers stopped counting minutes years ago. Operators learned to bundle usage into predictable packages with a clear monthly price. Generative AI arrives with the opposite commercial logic: variable technical costs measured in tokens, model calls and processing time.

Passing that complexity to customers would be a mistake. Nobody wants to calculate how many tokens it takes to stop a suspicious call, translate a conversation or create an appointment. Customers want the problem solved and a price they can plan.

The operator's opportunity is to convert technical consumption into understandable products. Depending on the use case, that can mean:

• a flat monthly tariff option 
• a defined allowance or volume package 
• a price per transaction 
• or an outcome-based model, such as one translated call, one blocked fraud attempt or one completed workflow.

This requires more than an AI model. The platform must emit a charging signal for the relevant function, tool call or outcome. The operator's BSS can then turn that event into a sellable product with a clear promise and a predictable price.

Voice AI creates opportunities in both enterprise and consumer markets. For smaller businesses, a network-native voice concierge could answer routine questions, qualify requests, schedule appointments or document calls. For consumers, services could include real-time translation, call summaries, personal assistance or enhanced fraud protection. 

The strongest business cases do not start with the model. They start with a specific customer problem and test four questions:

• Value: Is the outcome important enough that the customer will actively use or pay for it?
• Network advantage: Does operator reach, trust, identity or signaling make the service materially better?
• Integration: Can the agent safely connect to the required CRM, ERP, ticketing, knowledge base or other API?
• Productization: Can activation, charging, governance and operations be designed for repeatable delivery at scale?

That distinction matters. A fluent demo may prove that the technology works. It does not prove that the service can be governed, integrated, billed and operated across millions of customer interactions. Without those elements, Voice AI remains a science project. With them, it becomes a product.

For telcos, network-native Voice AI offers two complementary value paths. In B2B, it can support higher-value software and service propositions that are bundled into an existing customer relationship. In B2C and customer operations, it can reduce avoidable service effort while improving availability and convenience.

The common foundation is a platform that connects the voice network, controlled AI agents, enterprise back ends and charging. It must handle models and guardrails, run within operator-defined governance and convert each relevant capability into an event the commercial systems can understand.

Tallence built THOR Voice AI for precisely this purpose: to orchestrate real-time, network-native workflows and controlled intelligence inside operator environments, with integration into IMS, back-end systems and charging. The goal is not to add another isolated AI feature. It is to give operators a practical route from use case to commercially scalable service.