Snowflake, a AI Data Cloud company, has released a global research report titled “Radical ROI of Generative AI” in collaboration with Enterprise Strategy Group.

The report surveyed 1,900 business and IT leaders from nine countries, all of whom are currently using artificial intelligence for at least one use case.

According to the findings, 92% of respondents reported that their AI investments are already paying off, while 98% plan to allocate more funds to AI in 2025.

As AI adoption grows across industries, establishing a strong data foundation has become essential for successful implementation, although many organisations are still working to make their data AI-ready.

Baris Gultekin, Head of AI at Snowflake, noted,

“I’ve spent almost two decades of my career developing AI, and we’ve finally reached the tipping point where AI is creating real, tangible value for enterprises across the globe.”

He added that more than 4,000 customers use Snowflake for AI and machine learning each week, often seeing significant improvements in efficiency and productivity, as well as better access to data-driven insights across entire organisations.

The research highlights that early investments in AI are proving effective for most enterprises, with 93% of respondents stating that their AI initiatives have been very or mostly successful.

Two-thirds of those surveyed are already measuring the return on investment from generative AI projects, reporting an average return of US$1.41 for every US$1 spent, largely through cost savings and revenue growth.

The report also identified regional differences in AI adoption and outcomes. In Australia and New Zealand, organisations reported a 44% return on AI investments.

Companies in this region were more likely than the global average to prioritise customer satisfaction as a key goal, while internal projects received less attention.

Canadian organisations reported a 43% return, although many are still in the early stages of AI adoption, with a higher proportion only pursuing initial use cases.

French organisations reported a 31% return, with fewer companies making use of techniques like retrieval-augmented generation to train or enhance large language models, suggesting slower progress in AI maturity.

German businesses saw a 34% return on their AI investments.

However, many of these organisations reported infrastructure challenges, particularly in terms of meeting the storage and computing demands required for AI projects.

In Japan, businesses reported a 30% return. Unlike other countries, Japanese companies were less likely to focus their AI efforts on customer service or financial performance and were more inclined to apply AI for cost reduction.

South Korean organisations reported a 41% return and stood out for their mature use of AI, with the highest reported use of open source models and strong adoption of retrieval-augmented generation techniques.

In the United Kingdom, businesses reported a 42% return, with many emphasising AI’s value for operational efficiency and innovation.

In the US, companies also reported a 43% return, and respondents were more likely than those in other countries to describe their AI initiatives as “very successful” in achieving business goals.

Despite the reported success, organisations are facing increasing pressure when it comes to identifying which AI use cases to pursue.

A majority of early adopters acknowledged that there are more potential projects than resources to support them. 54% of respondents admitted that choosing the right use cases based on cost, business impact, and feasibility is challenging.

A similar proportion expressed concern that selecting the wrong projects could harm their organisation’s position in the market, while 59% also worried that advocating for the wrong initiatives could put their own jobs at risk.

The research further revealed that many organisations are working to integrate proprietary data to enhance AI performance.

About 80% of respondents reported fine-tuning models with their own data. However, significant challenges remain in preparing data for AI use.

A majority of early adopters reported difficulties in integrating data from different sources, enforcing governance, monitoring data quality, and preparing data for AI applications.

Many also cited the difficulty of scaling storage and computing capacity to meet AI demands.

Featured image credit: edited from freepik