Login
Register
Member Insights
Telecoms operators agree on the need for composable IT. The challenge lies in creating the authority and accountability needed to make it happen, argues Florian Hollender, of Statista Strategy.
Florian Hollender, Statista Strategy
Without a decision mandate, composable IT is just a slide in the strategy deck
The operators with the highest rate of cloud migration also have the weakest IT cost efficiency. That is not an anecdote but a correlation McKinsey measured across more than 20 operators worldwide. They moved into the cloud fast without optimizing the architecture underneath, which raises the bill instead of lowering it.
Almost no CIO in telecoms disagrees with the idea behind composable IT, yet for years implementation has barely moved. The gap was on open display this summer at DTW Ignite 2026, where a chief architect at a European operator admitted that his organization had not yet solved it. Before AI can do anything useful, he argued, the underlying data has to be exposed through composable, API-first interfaces; otherwise, the “automation tax” simply gets too high.
This is no niche concern for architects. McKinsey’s research shows that data platforms are common, yet fewer than half of operators reach the data quality that generative AI (GenAI) use cases demand.
Fragmentation is just as widespread, with 55% of survey respondents maintaining at least three separate technology stacks. And the average operator relies on six to seven vendors across operational and business support systems (OSS/BSS), where the most advanced operators need only three to four.
The fragmentation has grown over decades. A business unit builds its own reporting system, pulls data from other domains and reads it through its own lens, with each decision reasonable on its own. The result is an estate where new platforms are introduced without the old ones ever being retired.
Parallel operation is not a transitional state you can sit out. It is the normal state for as long as no one holds the cross-functional view and can enforce what gets rebuilt first.
As Jeanne Ross and colleagues at MIT CISR documented across two decades of research about enterprise architecture (culminating in Designed for Digital in 2019), the operational backbone that keeps a business running reliably is different from the digital platform used to compose new offerings — and firms that collapse one into the other, or try to shortcut the organizational stages between them, end up with neither. The payoff from composable IT arrives only when an accountability framework that spans domains sits above both.
To close the gap between insight and implementation, one authority has to enforce several things at once:
Two objections usually follow: Doesn’t AI solve this? And did we not learn that monolithic IT control fails?
Case studies of legacy systems modernized in days rather than years are piling up, and AI is a real booster, from process mapping to refactoring code. But AI-accelerated initiatives still need this frame.
And in looking back, the centralization waves of the 2000s produced the opposite of speed. The mandate is no license for a new head-office bureaucracy; it works only if the central authority decides on dependencies between systems, never on the content within a domain. Where that line blurs, composable IT tips back into the bureaucracy it was meant to replace.
The most fully enforced composable architecture on record came not from consensus but from instruction. Around 2002, Jeff Bezos ordered every Amazon team to expose data and functions only through service interfaces, no exceptions. AWS and much of its marketplace grew from that forced modularity.
Enforcement pays off for operators too. In Accenture’s CSP survey from 2024, operators with a below-average share of legacy IT grew revenue by 3.8% in 2023 against 2.9% for those with above-average share. Technology leaders also held IT operating costs to 0.8% growth on average over two years, against 3.4% for operators further behind.
Data infrastructure behaves like a bridge or a rail network: Neglect stays invisible for a long time, until friction rises sharply and repair becomes very expensive. Every new application placed on an uncoordinated architecture adds to a remediation bill that is coming anyway. AI helps us get ahead of the curve; it does not release us from anchoring central responsibility for composable IT in the organization.
