Recent findings from a MIT study that that found an estimated 95 percent of corporate generative AI pilots are failing to scale is spooking C-suites everywhere. Boards are asking harder questions, and this striking statistic is also causing CFOs to scrutinise the ROI on AI investments more closely than ever.
At the same time, CIOs are under pressure to explain why projects that look promising in the labs stumble in the real world. But what if the problem isn’t the AI itself, but the very infrastructure upon which it is built?
While there are multiple contributing factors to this high failure rate, a critical and often-overlooked cause is inadequate network infrastructure. Today’s AI implementations have moved beyond futuristic experiments and are now front-line operational tools with demanding workloads that cannot perform to their full potential on a foundation of outdated infrastructure.
This "AI-Network gap" is a primary and foundational reason promising AI pilots fail to deliver the business value promised when deployed at an enterprise level. Through our conversations with customers and our own market observations, it becomes clear that the AI-Network gap is not a niche technical issue, but a systemic problem that can hold back innovation.
When we speak with technology leaders in the region, a recurring theme is the limitation of their current networks, which prevents them from not just executing, but successfully implementing large-scale AI projects.
The hidden cost of network limitations
The true test for any AI project is not the controlled environment of a pilot; it is the transition to a scaled, enterprise-wide solution. This is where network deficiencies, previously hidden, are brought to the surface, transforming the network from a passive utility, and into central inhibitor of business performance.
In our experience, the top network-related challenges hindering AI success are a lack of flexible network scalability, and poor network performance, including latency issues. These factors are also typical of legacy networks, which are simply not built to manage the intensive, real-world data demands of AI applications and workloads at scale.
These limitations lead to poor application performance and project delays. This in turn leads to clear and direct fiscal impact thanks to unsolved network issues impacting productivity and the ability to run day-to-day business.
Ultimately, all these factors come together to prevent AI from meeting critical ROI benchmarks, and in many cases, cause the project to be cancelled altogether.
Why a network-first approach for sustainable AI matters
If we are serious about leveraging AI’s potential, we must adopt a “network-first” approach. In fact, our research shows that business leaders in APAC organisations are increasingly recognising that networking and connectivity investments as top priorities.
For this approach to be truly effective, CIOs must adopt a platform-first mindset. This is not about incremental upgrades, but rather about building a "unified digital fabric" where data, connectivity, and automation work seamlessly. Modern solutions like SD-WAN and SASE are notable examples of technologies that enable this platform-based approach, providing the required agility and resilience.
Technology is only part of the answer. A major challenge today in the region is the well-documented and ongoing shortage of specialised IT and networking talent. To bridge this gap, organisations must strategically partner with managed service providers (MSPs). This should not be considered a cost-cutting measure, rather, a strategic investment in enabling access to expertise and accelerating implementation, while allowing internal teams to focus on core business objectives.
The CIO’s new mandate: From realist to orchestrator
The AI conversation has elevated the role of the CIO. We have seen firsthand how tech leaders are gaining more visibility at the board level due to their unique position in driving AI strategy and implementation by bridging potential knowledge gaps.
With this increased influence comes a new and critical responsibility. The CIO is no longer just the realist in the room, tempering hype with technical limitations. They must now be the orchestrator of change, proactively reshaping systems to make AI sustainable.
- Eric Wong, President, APAC, Expereo
A key part of this new mandate is to champion investments in the "invisible" layers of connectivity and architecture. It is the CIO’s job to ensure that innovation is not just impressive in a pilot phase, but also scalable and sustainable in the real world. The high failure rate of AI pilots is a clear symptom of neglecting foundational infrastructure, and more importantly, the financial and strategic risks are too great to ignore.
The goal is not to delay transformation; it is to enable it to scale. The organisations that win the AI race will be those who recognise the need to lay the groundwork first, ensuring their digital infrastructure is robust enough to deliver value at scale for years to come.
The future of AI is not in the algorithm; it is in the network.
Eric Wong is President, APAC, Expereo.
