Home Frontier Tech AI & ML 1 in 5 AI projects likely to fail sans smart data infrastructure

1 in 5 AI projects likely to fail sans smart data infrastructure

Up to 20% of AI initiatives fail without intelligent data infrastructure, according to a white paper from IDC.

Sponsored by NetApp, the report surveyed 1,220 global decision makers involved in IT operations, data science, data engineering, and software development for artificial intelligence initiatives. A quarter of these respondents were from the Asia-Pacific region.

In conducting this analysis IDC has developed an AI maturity model where organisations fall into one of four maturity levels based on their current approach to AI — AI Emergents, AI Pioneers, AI Leaders, and AI Masters.

IDC found that AI Masters optimise their data infrastructure for transformational AI initiatives by facilitating easy access to corporate datasets with minimal preparation and designing a unified, hybrid, multicloud environment that supports various data types and access methods.

AI Masters have more ambitious AI goals and yet experience data-related failures including infrastructure-based data access limitations (21%), compliance limitations (16%), and insufficient data (17%).

AI Emergents note similar challenges but also experience budget constraints (20% Emergents vs 9% AI Masters), insufficient data for model training (26% vs 17%) and business restrictions on data access (28% vs 20%).

According to the findings, organisations need an intelligent data infrastructure in order to scale AI initiatives responsibly. Where a company falls on the AI maturity scale is determined by the level of infrastructure they have in place that will not only drive the long-term success of AI projects, but also of their associated business outcomes. 

Those organisations that are just beginning or have recently begun their AI journey typically have disparate data architectures or plans for a more unified architecture, while AI Leaders and AI Masters are likely already executing on a unified vision. As a result, organisations with the most AI experience are failing less.

“This IDC white paper further solidifies that companies need intelligent data infrastructure to scale AI responsibly and boost the rate of AI initiative success,” said Jonsi Stefansson, SVP and CTO at NetApp. 

“With intelligent data infrastructure in place, companies have the flexibility to access any data, anywhere with integrated data management to ensure data security, protection, and governance and adaptive operations that can optimise performance, cost and sustainability,” said Stefansson.

Also, IDC found that 48% of AI Masters report they have instant availability of their structured data and 43% of their unstructured data, while AI Emergents have only 26% and 20% respectively.

AI Masters (65%) and AI Emergents (35%) reported their current data architectures can seamlessly integrate their organization’s private data with AI Cloud services.

According to the research, AI Masters know that their data architecture and infrastructure for transformational AI initiatives must offer ease of access to corporate data sets without any—or with only minor—preparation or preprocessing.

Further, IDC found that the inability for AI Emergents to progress is often due to a lack of standardised governance policies and procedures; only 8% of AI Emergents have completed and standardised these across all AI projects, compared to 38% of AI Masters.

While 51% of AI Masters reported they have standardised policies in place that are rigorously enforced by an independent group in their organisation, only 3% of AI Emergents claim this.

The study found that effective data governance and security are crucial indicators of organisational maturity in AI initiatives. Managing data responsibly and securely remains a key issue for enterprises, because AI stakeholders often try to shortcut security processes to accelerate development.

In addition, 43% of AI Masters have clearly defined metrics for assessing resource efficiency when developing AI models that were completed and standardised across all AI projects compared to 9% of AI Emergents.

More than three-fifths (63%) of all respondents reported the need for major improvements or a complete overhaul to ensure their storage is optimised for AI and only 14% indicated they needed no improvements.