AI is exposing the limitations of legacy IT and traditional operating models as CIOs face growing pressure to deploy new capabilities safely, economically and at enterprise scale, says Willie Stegmann, EVP, Composable IT and Ecosystems, TM Forum.
AI has vast potential to dramatically improve customer experience, operational efficiency and broader ecosystem growth opportunities. In contrast, most CSPs are constrained by legacy IT, fragmented architectures, poor data quality and most importantly business and operating models unfit for AI Native execution and value realisation.
The primary challenge is not access to AI technology, it is the failure to deploy AI safely, economically and at enterprise scale. AI exposes operating model and associated capability constraints that were manageable when transformation could be sequenced but become blockers when everything moves at once.
The telecom industry has been fast to exploit the power of Autonomous Networks, however AI bolted onto legacy IT systems generates fragility and not autonomy. This is why many CIOs are now gravitating towards composable IT not as an architectural ideal, but as a practical response to continuous pressure to exploit the potential of AI.
Composability shifts the focus from planning end-state transformations to building execution capability: the ability to reassemble systems, processes and teams quickly as priorities change. We have proven that composability is a critical building block in the AI era. Open, composable architectures allow AI agents to interact with components through managed, open APIs.
The breakdown of phased transformation
Digital and cloud transformation programmes were meant to make this moment easier, but many stopped short of the simplification CIOs were promised. In practice, core platforms were often migrated or wrapped rather than retired, while new systems were layered onto old ones.
Complexity shifted but rarely disappeared. AI exposes the consequences of those compromises quickly because models depend on clean integration, shared data and consistent orchestration across systems.
Fragmented stacks make AI adoption harder, slower and more expensive than anticipated, undermining the economics leadership increasingly expects AI to deliver. The issue is not that modernisation work failed to start, but that it rarely removed enough friction to support what AI now demands.
The operating model constraint
This is not a technology gap. It is a leadership and operating-model challenge. AI does not arrive as a single system that matures independently. It spreads across customer interactions, operational processes and internal engineering workflows, with multiple models coexisting and influencing outcomes downstream.
As that landscape expands, unresolved architectural and organisational constraints become visible in real time through rising costs, delivery friction and slower-than-expected scaling.
In earlier transformation cycles, CIOs could make trade-offs over time. Cost optimisation might lag while new capability was built and legacy reduction could follow innovation. AI collapses those timelines.
CIOs are now expected to modernise legacy systems, maintain, extract savings and deploy new AI capabilities simultaneously. Traditional optimisation approaches, designed for incremental efficiency gains, struggle under these conditions, while multi-year modernisation road maps often feel disconnected from immediate pressure to demonstrate impact.
Execution as a capability
This pressure is why the most effective CIOs are reframing the problem. Rather than treating AI as an additional initiative, they are shifting focus from programmes to execution capability.
AI is applied internally to automate engineering tasks, streamline operations and identify inefficiencies across complex estates. Savings generated through these improvements are reinvested directly into further modernisation and AI deployment, creating a reinforcing cycle in which AI increasingly funds its own expansion. This is particularly important, given flat IT capex allocation prevalent in the industry.
Modernisation in this model is driven less by ideal end states and more by incremental value realisation: what reduces friction, improves velocity and enables the next capability to be deployed faster and at lower cost.
When this approach works, the conversation with senior leadership changes. AI is no longer framed as an investment chasing future returns, but as a practical mechanism for simplifying IT, improving productivity and strengthening cost control at the same time.
Complexity is not eliminated overnight, but the trajectory shifts as AI becomes embedded in how IT operates rather than layered alongside it. Change stops being episodic and becomes continuous.
Reinvention as a way of operating
What is abundantly clear is a heightened sense of urgency that manifests in significant pressure on CIOs and IT leadership teams to respond and deliver results – fast. Given the reality, described above, of legacy drag, the call to action is clear – reinvent IT for the AI Era – with real intent.
This leadership-led reinvention must solidify IT AI foundations, incrementally modernise legacy IT – to become more open, composable and responsive – and critically leverage AI to advance agentic customer experiences and unlock broader growth opportunities.
At this point, many CIOs are no longer talking about reinvention as a destination or a programme. It is emerging instead as a way of operating, a recognition that the environment they are working in no longer allows for extended periods of stability between waves of change.
AI accelerates this shift not because it introduces more initiatives, but because it removes the buffers that once existed between strategy, execution and outcomes.
What stands out across CIO conversations is how this changes decision making. Planning cycles shorten, architectural choices are made with reversibility in mind and progress is evaluated less by roadmap completion than by reduced friction, faster learning and ongoing value unlock.
Reinvention, in this sense, is not about pursuing perpetual change for its own sake. It is about building the capacity to adapt continuously without destabilising the organisation.
This mindset also reframes risk. Rather than treating reinvention as a disruptive event to be tightly contained, CIOs increasingly see rigidity itself as the greater threat. In an environment where expectations shift faster than systems can be fully modernised, adaptability becomes a control mechanism, not a liability.
Continuous reinvention as leadership reality
AI is not exposing a lack of ideas, talent or intent among CIOs. It is exposing whether organisations can still operate when the assumptions that once structured change no longer apply. Sequencing, buffering and long periods of stability are increasingly difficult to sustain as expectations accelerate and timelines compress.
For CIOs, this shifts the role from managing transitions between initiatives to leading through continuous overlap. Reinvention becomes unavoidable not because of ideology, but because the environment no longer allows change to be neatly staged.
Decisions are made with incomplete information, systems evolve while still in use and progress is judged by reduced friction rather than finished programmes.
This is where the opportunity lies. CIOs who treat reinvention as a necessary capability rather than a one-time event are better placed to remain in control as AI reshapes how technology delivers value. By focusing on execution, adaptability and learning while in motion, they can turn sustained pressure into forward momentum.
