Overcoming the value-stream illusion in OSS transformation

Over the last two decades communications service providers (CSPs) have invested heavily in modernizing their operational support systems (OSS) to improve automation, agility, operational efficiency and customer experience. These efforts enabled the digitalization of workflows, resulting in faster order processing, more intelligent fault correlation and a more structured approach to provisioning. But a deep examination reveals that substantial work remains manual inside planning, engineering, testing, and day-to-day operations departments.

Research shows that many CSPs’ automation initiatives were built on an implicit assumption that automating visible workflow steps would automatically translate into material value realization. Operators were successful with automating the visible portions of operational workflows – those steps that could be clearly defined and digitized – while leaving untouched the judgment, interpretation and contextual decision-making required to execute work in live network environments.

The cost is between the workflows

Traditional OSS environments were designed to manage state transitions: create an order, reserve a resource, activate a service, log a problem, execute a workflow, update inventory. While these actions are important, they do not capture the full reality of telecom operations.

Engineers expend the highest effort on work that has the highest risk around these abstractions such as manually validating OEM release guidance before a software update, reconciling conflicting inventory and topology views, translating local engineering rules into deployable configurations, assessing maintenance impact, preparing rollback conditions, and determining whether a change is safe enough to execute.

For many CSPs, the largest operational cost is not the transaction itself, but the labor wrapped around executing the transaction. Engineers interpret the documents the system cannot consume. Planners perform calculations the workflow does not natively understand. Operations teams bridge the gap between formal tooling and local operational conditions.

In effect, humans become the integration layer between structured systems and unstructured reality. This is exactly why CSPs have not been able to materially change the labor-intensive model underlying planning, engineering, testing and field operations despite success in improving process KPIs.

Capability exposure: the next frontier

The problem is not legacy OSS environments. The deeper issue is that these systems were architected for system-of-record integrity, process control, workflow control and deterministic transaction handling based on the design assumption of structured repeatable tasks.

OSS has traditionally been treated as a collection of coarse-grained functions delivered by monolithic in-house codebases or commercial-of-the-shelf (COTS) software suites. This architecture breaks down in an operational environment where tasks depend on fragmented documentation and the application of contextual judgment to address inconsistent data quality, or operational nuances that have never been formally modeled. In this reality, engineers become the hidden integration layer within the operations architecture.

TM Forum’s work on business architecture, especially the Business Architecture: Capability Driven Transformation guide (IG1277) and the Value Stream Framework (GB1029V), offers a more suitable lens for the next phase of CSP transformation. IG1277 makes an explicit distinction between value streams, capabilities and processes. It views value streams as delivering stakeholder outcomes through capabilities, while capabilities define what the enterprise must be able to do and are enabled by people, processes, information and systems, not process logic alone.

This is the architectural pivot telecom now needs to make. Instead of continuing to treat OSS primarily as a collection of applications and user interfaces, CSPs should treat it as a platform that exposes reusable, modular, implementation-agnostic operational capabilities that can be composed, governed and invoked across workflows.

Delivering the next generation of operational value requires a shift from process optimization to capability-centric execution, where OSS functions are exposed as reusable, governed components that can be dynamically orchestrated. This unlocks the value between process stages and enables a more intelligent, agentic operating model aligned with TM Forum’s Open Digital Architecture (ODA).

How agentic operations become real

This architectural pivot is also how CSPs should look at AI. The opportunity in autonomous operations lies in intelligent agents that can reason across context, invoke tools, execute bounded actions and return auditable outcomes within explicit guardrails. These intelligent agents can address the operational challenges OSS was never able to address including unstructured documentation, local operating knowledge, exception handling and contextual decision support. These agentic capabilities create a significant opportunity to eliminate the need for humans to act as a “cognitive glue” middleware layer.

For agentic operations to scale operationally, CSPs need an execution fabric between agents and operational systems. Emerging standards such as Model Context Protocol (MCP), and the TM Forum’s evolving work on Open APIs, MCP and the telecom-specific MCP-T concepts, point toward a more standardized way for agents to discover context, access tools and operate within policy boundaries. This standardization changes the operating model materially by making engineering work decomposable into machine-usable capabilities along with governed reasoning.

Five lessons for autonomous networks

As CSPs push toward autonomous operations five shifts set leaders apart:

1. Adoption of TM Forum best practices – value streams express stakeholder outcomes; the Business Process Framework (eTOM) structures the business processes; and capabilities represent the reusable abilities required to deliver both. Conflating these concepts weakens architecture, misguides investment decisions, and does little to reduce underlying costs.
2. Focused AI transformation – it’s necessary to expose fine-grained, policy-aware micro-capabilities that agents and orchestration layers can invoke, ensuring that operational logic does not remain trapped inside monolithic applications or human-only interfaces.
3. Curation of human knowledge – standard operating procedures, OEM manuals, local templates, maintenance logic and engineering decision patterns are all seen as operational assets.
4. Focus on compliance and trust – Operations are not just about fulfillment speed; they’re also about trust, compliance, control and accountability. The agentic operating model must respect this architecture, not bypass it. This means building policy, assurance and audits into runtime processes from the start.
5. Embedded oversight and control at every layer of the architecture – human in the loop approvals, observability, traceability, rollback discipline, role-based access and policy enforcement are necessary. TM Forum’s architecture direction is clear on the need for governance, standards, security and disciplined execution. Agentic operations without these controls is not transformation; It is simply a faster route to poor change.

The real prize lies beyond process automation

The telecom industry has an operations cognition gap. It has automated what was easy to formalize while leaving untouched most of the human engineering and operations work that sits between system states and process steps. While this has improved workflow efficiency, it has not proportionately changed the cost base or agility of engineering or operational teams.

The manual work between system states is where engineering time disappears, and where operational risk accumulates. It is also where the next major efficiency gain still sits. The next frontier in autonomous operations will belong to the CSPs that can shape their architecture using standards to expose modular, governed, reusable and intelligent micro-capabilities to absorb more of the operational work still dependent on human judgment.