More than nine in every 10 (92%) business and IT leaders in eight markets across the globe say their AI investments are already paying for themselves.
Also, 98% of them plan to invest more on AI in 2025. As AI adoption accelerates across global enterprises, a robust data foundation has emerged as the cornerstone of successful implementation, yet respondents are still grappling with how to make their data AI-ready.
This is based on a survey conducted by Enterprise Strategy Group and fielded between Nov. 21, 2024, and Jan. 10, 2025.
There were 3,324 respondents from the United States (45%), Canada (8%), the United Kingdom (8%), France (8%), Germany (8%), Australia and New Zealand (ANZ) (8%), Japan (8%), and South Korea (8%).
Respondents belonged to firms that are already using generative AI to augment and execute business processes in production.
They also belong to firms that are using generative AI tech beyond consumer-grade “off-the-shelf” solutions — such as subscriptions to ChatGPT — to commercial and open-source models that can be trained and tuned with proprietary data to improve accuracy and relevance.
Findings show that early AI investments are proving to be successful for the majority of enterprises, with 93% indicating that their AI initiatives have been very or mostly successful.
Two-thirds of respondents are already starting to quantify their generative AI ROI today. For every dollar spent, they are seeing $1.41 in returns (or 41% ROI) through cost savings and increased revenue.
However, there are global nuances around where organisations are focusing their AI efforts that directly correlate to each country’s AI maturity, and their results in terms of driving ROI across regions.
ANZ respondents have seen a 44% return on their AI investments.
Compared to the global average, organisations in ANZ were more likely to cite enhancing customer satisfaction as a key goal for their AI initiatives (53% versus 43%), and less likely to prioritise internal-facing projects (47% versus 55%).
“Local organisations are funding gen AI at a rate above the global average which bodes well for the development and growth of AI in our region,” said Theo Hourmouzis, senior regional VP of ANZ and ASEAN at Snowflake.
“While there is a clear appetite and drive to be ahead of the AI curve, there are hurdles that local business and IT leaders are at pains to overcome – the two biggest challenges being talent and data,” said Hourmouzis.
Respondents in ANZ more often reported certain challenges in their generative AI initiatives than the global average.
For one, there are many competing priorities. Respondents in ANZ more often struggled to identify the right use cases to pursue (71% versus 54%).
There are also data obstacles. Compared to the global average, local organisations more often cited a lack of data diversity/range (56% versus 42%), time consuming data management tasks (62% versus 55%) and data preparation (59% versus 51%) as difficult areas.
ANZ organisations more often say it’s hard to break down data silos (76% versus 64%).
Further, there are unexpected costs. The majority (84%) of ANZ organisations say half or more of their gen AI use cases have cost more than expected to get into production, compared to the global average of 78%.
Snowflake found that, despite this widespread recognition of data’s importance, significant challenges persist in making this data AI-ready.
With the majority struggling to make use of their most valuable asset, organisations claim that the following are the biggest data hurdles for driving AI success.
First, breaking down data silos. Close to two-thirds (64%) of early adopters say integrating data across sources is challenging today.
Second, integrating governance guardrails. Three-fifths (59%) say enforcing data governance is difficult.
Third, measuring and monitoring data quality. Three-fifths (59%) say measuring and monitoring data quality is difficult.
Fourth, integrating data prep. Three-fifths (58%) say making data AI-ready is a challenge.
And fifth, efficiently scaling storage and compute. More than half (54%) say it’s difficult to meet storage capacity and computing power requirements.
