Software development is being rapidly reshaped by AI-assisted approaches. AI researcher Andrej Karpathy has coined the term “vibe coding” to describe how developers express intent in natural language and then allow AI agents to handle implementation details. You do not write every line of code. Instead, you iterate a conversation with an AI partner.
Agentic AI is accelerating this shift, and the primary consumers of telecom are no longer only human developers but autonomous software systems. As these AI agents become central to how digital services are created, discovered, and operated, the telecom ecosystem will be reshaped in developer platform support, monetization control points, and new value models. Agentic systems may soon identify relevant capabilities, reason about their suitability, and integrate them directly into live services at machine speed. What once took years of ecosystem evolution can compress into months.
This raises a few questions:
- What happens when machines become the primary users of telecom networks?
- What happens when APIs are chosen by autonomous agents that weigh task fit, cost, compliance, and risk, rather than developers reading documentation?
- What happens to ecosystem value when decisions are run at machine speed?
These questions motivated our latest research, as we believe this transformation is happening faster than expected. We interviewed ten experts across AI platforms, telecom operators, and enterprise adopters of agentic AI to understand how autonomous agents could reshape telecom economics and ecosystem roles.
We complemented these interviews with a literature review, scenario workshops that explored alternative ecosystem futures, and hands-on experimentation with AI coding assistants and low-code tools. Rather than speculating about distant futures, we focused on early signals already visible in development environments and platform ecosystems.
Here is what we learned from the study.
-
AI agents are becoming the dominant users of telecom capabilities
For the past decade, the industry has invested in developer experience (DX): documentation, sandboxes, and standardized APIs. But we are now entering a phase where developers are no longer the main users of network capabilities. Instead, autonomous agents embedded in development platforms, orchestration frameworks, or enterprise workflows are increasingly discovering, interpreting, and invoking telecom services on their own.
This marks the emergence of an agentic experience (AX), a world where APIs must be understood by both people and machines. In practice, AX is less about machine-readable APIs and more about machine-delegable capabilities. Agents need to understand not only what a network function does, but under what authority, constraints, and economic responsibility it can be invoked.
This implies that agentic systems should be able to read API schemas, reason about context, and select the network functions required to fulfil a task. In practice, an AI assistant could request a quality-of-service API during a video session, or a fraud-detection agent might automatically query call-forwarding status to flag suspicious behavior.
The implication is clear, the developer of tomorrow may be an autonomous agent acting on behalf of a person or a business. This shifts the core design problem from tooling to delegation governance. Who authorizes the agent, who carries liability, and how outcomes, costs, and failures are accounted for across organizational boundaries in interdependent ecosystems become the central questions.
-
AI-driven telecom API discovery, and why invisibility is a new risk
Telecom APIs used to be discovered through developer portals and traditional documentation. The next generation of discovery, however, is unfolding inside AI ecosystems. Coding assistants (such as GitHub Copilot or OpenAI’s GPT-based models), Model Context Protocol (MCP) registries, and AI marketplaces increasingly determine which APIs are recommended or automatically integrated into generated code.
As AI lowers the technical barrier to entry, virtually anyone can build telecom-enabled applications. Natural language interfaces, low-code tooling, and autonomous coding aids means telecom capabilities are no longer limited to enterprise developers or system integrators.
Today, individuals, startups, and even people without technical backgrounds can create and manage communication services using AI-assisted development tools. This broadens innovation but also intensifies competition for visibility within AI-mediated environments.
Discovery is thus evolving from a neutral technical activity into a new power asymmetry. Agent runtimes, orchestration frameworks, AI marketplaces, and tool registries increasingly decide which APIs are visible or recommended, creating novel control points beyond the network layer. In our ecosystem mapping, orchestration layers and AI-native platforms appear increasingly central in the stack, while telecom API platforms are not yet positioned at those decision-making layers. This turns visibility in AI-mediated environments into a strategic question, not merely a technical one.
Horizontalization accelerates this shift. As network capabilities become reusable AI-native components, traditional vertical value chains fragment. This may reduce the number of dominant infrastructure players in some stack layers while enabling new AI intermediaries to emerge. For instance, upcoming agentic AI marketplaces could allow telecom functions to be exposed, discovered, and negotiated by autonomous systems rather than manually by developers. In such environments, agents may dynamically assess factors like latency, reliability, compliance, and price to select optimal providers programmatically.
Despite this growing automation, developers remain cautious about trusting AI-generated code. Ongoing concerns around security, reliability, explainability, and vendor lock-in ensure human oversight still matters. Oversight moves up a level, from writing code to validation, governance, and architecture control.
APIs that are not machine-readable, schema-described, or present in AI training data risk becoming invisible, even if they are technically robust. Over time, this may create a concentration effect, where autonomous systems preferentially use the APIs they already know from prior training and examples, reinforcing a small set of providers.
In this emerging environment, success hinges not only on API design but also on how effectively a service is interpreted, ranked, and trusted by AI models. Visibility and credibility within AI-mediated discovery systems will become as critical as performance or scalability.
-
Identity, authentication, and validation as new monetization control points
As AI agents begin to act on behalf of users and enterprises, machines will initiate actions, negotiate resources, and commit to transactions. The system must be able to determine who is acting, under what authority, and with what constraints. This puts identity, authentication, and validation to the forefront. In the AI agent ecosystem, these security mechanisms become monetization control points—strategic touchpoints where providers can influence how services are delivered, priced, and governed.
This change shifts the value-capture position. In traditional API-based models, monetization follows usage such as calls, messages, or data volumes. In agentic systems, value concentrates around verified trust: authenticated identities, authorized delegations, validated outcomes, and auditable actions. These become enforceable points where technical correctness meet economic accountability.
This is how an experienced platform architect described the shift:
“When an AI agent invokes a network capability, the technical call is trivial. The real value lies in knowing whether that agent is allowed to act, who carries responsibility, and how risk is contained.”
When autonomous agents make decisions at machine speed, these control points become critical for reducing risk, meeting regulatory requirements, and ensuring reliable outcomes.
Many interviewees emphasized that large-scale agentic automation is not feasible without machine-verifiable trust:
“You cannot scale autonomous systems if trust still depends on human review. Identity and authorization have to be readable and enforceable by machines.”
For telecom providers, this is a natural extension of existing strengths. Networks already operate at the intersection of identity, security, regulation, and real-time assurance. By positioning these capabilities as monetizable services, telecoms can move beyond commoditized API access and instead capture value through trust-as-a-service: charging for verified actions, guaranteeing integrity, and reducing uncertainty in autonomous interactions.
If telecoms do not claim these control points, they will not disappear. They will migrate upstream to AI platforms and cloud providers that integrate identity and trust directly into their agent frameworks. In an agent-driven ecosystem, control over authentication and validation increasingly determines who shapes the rules of participation and who captures value.
From APIs to outcomes and self-optimizing networks
For a long time, the focus in application development has been on exposing network APIs and improving developer engagement. Agentic AI pushes the ambition further, however, orchestration is not a role telecoms inherit by default; it appears only when they control the delegation, trust, and settlement layers that autonomous agents depend on to operate safely at scale.
As applications become dynamic and context-aware, they will require continuous negotiation of identity, policy, service quality, and environmental context. These are all areas where telecom providers hold unique assets.
Across interviews and scenario workshops, one signal was consistent. The shift toward intent- and outcome-based reasoning, regardless of whether respondents expected a fragmented or consolidated AI ecosystem.
The study shows that telecoms possess deep network intelligence, global operational stability, and secure infrastructure. These strengths can support agentic automation across domains such as mobility, mixed reality, internet of things, and digital twins. To stay relevant, these assets must be offered in AI-native ways. That means:
- shifting from raw APIs to outcome-oriented service bundles
- providing machine-interpretable trust and security frameworks
- enabling agents to reason about network capabilities autonomously
In some cases, telecoms may even supply the agents themselves, based on specialized models trained on network behavior, safety constraints, or advanced 6G features to help developers build AI-native applications that fully leverage the network.
The industry has a clear opportunity to use this technology to accelerate network exposure. For example, it can start with secure identity, trusted data, and guaranteed quality of service.
A strategic inflection point for the telecom industry
Agentic AI will not eliminate developers, but it will redefine their role. It will elevate the importance of the ecosystems that agents operate in and the telecom capabilities they depend on. Our research clearly indicates that telecoms to stay relevant must move rapidly from a connectivity-first mindset to a trust- and orchestration-first positioning.
The speed of this transition is the real disruptor. Early adopters will shape how agentic systems discover, select, and rely on network intelligence. Those who do not may find their APIs bypassed by autonomous platforms that optimize for simplicity, semantics, and reliability.
The telecom ecosystem is entering a new era where machines are the integrators. While the ecosystem is still evolving, the consistency across interviews, scenario workshops, and experimentation was notable. Actors across AI platforms, telecom operators, and enterprises pointed to similar structural shifts in delegation, trust, and where value begins to concentrate. So the question is not whether telecoms expose APIs, but whether they control the conditions under which agents can act, transact, and commit on behalf of users and enterprises.
The opportunity is clear: become indispensable to the agents that will build and run the digital services of the future.
It is no longer about who wins DX in traditional sense. It is about who offers the most efficient environment for agentic-driven application development.
