Companies Can’t Scale AI Without First Fixing Their Data, Study Finds
As AI ambitions soar, the report urges companies to revisit their data architectures, governance, and integration model.
News
- Meta Leans on AI to Revive Marketplace’s Appeal Among Young Adults
- Baidu Stock Drops as Its New AI Model Fails to Win Over Investors
- MBZUAI’s New Model Builds Worlds That Remember Your Moves
- Microsoft Launches 'World’s First AI Superfactory'
- OpenAI Pushes Back on NYT’s Demand for 20 Million ChatGPT Conversations
- SoftBank Sells Out of Nvidia as It Bets Big on AI
A new Salesforce report finds that while enterprises are eager to lead the next wave of AI innovation, many are still strengthening the foundation that makes it possible — their data.
The State of Data and Analytics 2025 report reveals a widening disconnect between business leaders’ AI aspirations and the data infrastructures meant to sustain them — a mismatch that could derail efforts to build “agentic enterprises,” where humans and intelligent AI agents work in tandem.
According to the report, 76% of business leaders feel growing pressure to deliver measurable business value from data. Yet their biggest barrier remains unchanged: incomplete, outdated, or poor-quality data. Nearly two-thirds (63%) of data and analytics leaders admit their organizations struggle to use data to drive business priorities, despite an equal proportion of executives proudly describing their firms as “data-driven.” Less than half of business leaders say they can reliably generate timely insights, and almost half confess that poor business context often leads to incorrect conclusions.
The arrival of agentic AI, which are systems capable of autonomous decision-making, is magnifying these weaknesses. While 67% of data and analytics leaders feel pressure to implement AI rapidly, 42% lack confidence in its accuracy. The reason, Salesforce states, lies in data: over a quarter (26%) of organizational data is deemed unreliable. Inaccurate data is leading the companies to splurge, with 89% of organizations running AI systems reporting misleading or erroneous outputs, and more than half admitting to wasting resources training models on faulty data.
“Trusted, unified, and contextual data is the key that unlocks everything,” said Michael Andrew, Chief Data Officer at Salesforce. “This is the moment to shore up data foundations to confidently scale AI to its full potential.”
The report highlights that data architecture, not algorithms, is the real determinant of AI. Enterprises now use an average of 897 applications, but only 29% are connected — leaving nearly one-fifth (19%) of company data siloed or inaccessible. Alarmingly, 70% of data and analytics leaders believe their most valuable insights lie hidden within that inaccessible fraction.
To counter this, 56% of organizations are adopting zero copy data integration that allows data to be accessed across systems without moving or duplicating it. Salesforce found that firms using zero copy technology are 25% more likely to deliver superior customer experiences and 34% more likely to generate better outputs with AI initiatives. Meanwhile, natural language interfaces, or “agentic analytics,” are gaining traction as a way to make data more accessible to non-technical users.
Yet even as technical solutions emerge, governance remains a weak link. Only 43% of data leaders report having formal governance frameworks, and 88% agree that the rise of AI demands entirely new approaches to data management and security.
Salesforce CEO Marc Benioff, speaking at Dreamforce, called agentic AI “the next revolution,” but cautioned that “to get the most value and context from AI models, you’ve got to get your data right.”
The report’s central message is that AI will not fix bad data. Instead, it will only amplify its flaws. As enterprises rush to deploy AI at scale, those that fail to repair their data foundations risk creating a mirage of misinformation.
