Fivetran report finds enterprises racing toward AI without the data to support it

Fivetran has released new research showing that fewer than half (49%) of enterprise technology leaders believe their current data architecture can handle the demands of AI. At the same time, 89% say they plan to use proprietary data to train large language models (LLMs) this year. The disconnect highlights how quickly companies are pushing forward with AI, even as they acknowledge their data systems aren’t ready.

The report, 2025 and Beyond: How a Strong Data Foundation Fuels Enterprise Success and AI-Driven Innovation, is based on a survey of 500 US-based C-suite technology leaders at companies with more than $500 million in annual revenue. It found that while enthusiasm for AI is high, many organisations are still struggling with basic issues like data integration, scalability, and compliance. Sixty-eight percent of respondents said they rely on 50 or more data sources to support decision-making, and more than a third cited integration complexity as a major hurdle. Others pointed to scalability limitations (34%) and security and compliance risks (33%) as top concerns.

“If you don’t centralise your data, you can’t see what’s really happening in your business,” said George Fraser, CEO of Fivetran. “Most companies have the intent, but execution is where things fall short. Moving diverse data quickly and reliably requires the right strategy and the right technology. That’s why we’re seeing forward-looking teams invest in platforms that scale securely, manage complexity, and make AI possible.”

Many companies are taking action. Half of the executives surveyed said their organisations plan to invest $500,000 or more in data integration over the next year. Their focus areas include reducing manual pipeline maintenance, improving real-time access to data, and ensuring data quality and governance. Still, challenges remain. Forty-five percent reported a lack of automation or self-service capabilities, 44% said legacy systems and implementation costs were holding them back, and 41% pointed to talent gaps on their data teams.

Technology leaders also recognise that advancing data and AI capabilities means taking on greater security and compliance responsibilities. Sixty-four percent of CIOs said they’ve had to delay innovation efforts specifically to address compliance concerns, putting key projects on hold while they work to get their data in order.

The report also shows how the role of the technology leader is shifting, with 48 percent of respondents expecting to take on more responsibility for data privacy and compliance, and 45 percent anticipating a larger role in company-wide data strategy. Some organisations are already seeing results. At Sedgwick, a centralised and well-governed data foundation has enabled the team to move quickly on real-time AI applications.

“Now that our data is centralised and governed with Fivetran, we’re building AI tools that surface real-time insights for our claims examiners,” said Adam Fisher, Chief Data Officer at Sedgwick. “Adjusters can see what similar claims have cost in the past and how likely they are to be litigated, right as they’re reviewing a case. This kind of visibility wasn’t possible before, and it’s where AI is starting to make a real difference.”

Key findings include:

• 89% of tech leaders plan to use proprietary data to train LMMs this year, but only 49% believe their architecture can support AI workloads.

• 68% say they rely on 50 or more data sources to support decision-making.

• 64% of CIOs have delayed innovation efforts due to compliance concerns.

• 50% plan to invest $500,000 or more in data integration in the next year.

• Nearly half of respondents expect to take on more responsibility for privacy, compliance, and company-wide data strategy.

As more companies move from AI pilots to real-world applications, strong data infrastructure is becoming a priority. Without reliable, centralised access to data, even the most advanced tools fall short. This report shows that while investment is increasing, many organisations still have work to do. Building the right foundation now will be key to making AI efforts successful in the months and years ahead.