Agentic AI, artificial intelligence and machine learning, next generation technology and secure development
EXL study finds data and business processes separate leaders from laggards
Jennifer Rawinski •
June 23, 2026
Many business leaders need a reality check when it comes to artificial intelligence.
See also: AI agent introduces new insider threat model
According to the third annual EXL U.S. Enterprise AI Survey, 76% of companies surveyed believe they are better than their competitors when it comes to AI, but when researchers dug into how deeply AI is embedded across business functions such as customer service, marketing, finance, and human resources, only 10% of companies were actually AI leaders.
The report divided respondents into three categories. “Reader” has fully developed AI capabilities with 6 to 8 features. “Followers” meaningfully implemented AI in three to five features, while “laggards” implemented AI in two or fewer features. The study surveyed 322 executives and senior decision makers in the banking, finance, insurance, retail, utilities, life sciences, and health insurance industries.
“Every company is leveraging AI in some way today,” said Anand Roghani, executive vice president and chief AI officer at EXL. “But when it comes to large-scale enterprise integrations, leaders are starting to differentiate themselves.”
The data shows that the impact of overestimating AI’s success goes beyond hurting self-esteem. This is also reflected in the financial statements.
The study found that leading companies reported that AI reduced costs by an estimated 26%, increased revenue by an estimated 27%, and improved profits by an estimated 22% in specific workflows where AI was applied. Laggards continued to emerge in all categories.
Rogani said what separates leaders from laggards is a willingness to stop treating AI as an add-on and start redesigning workflows around AI from the beginning. Leaders are deploying AI at scale and re-engineering their operating models to take advantage of it.
Logani said that to understand why many companies are overestimating the maturity of AI, we need to understand how the influx of AI into the enterprise technology stack differs from previous technology cycles.
For example, when companies started moving to the cloud, business leaders agreed to the investment but were not involved at a granular level. But when it comes to AI, boards and executives have a vested interest in where and how AI is deployed, and they know more about it than cloud, ERP systems, or data warehousing.
“Because they are connected, there is a lot of education flowing down to the executive level suite,” Rogani said. “So you’re relatively more informed than in previous waves.” But being well-informed about AI is not the same as deploying it at a meaningful scale.
Early success makes the problem worse. Organizations that have conducted pilots have real-world evidence that AI works. Rogani said companies compare themselves to companies they hear about in the media that are struggling and feel like they’re ahead of the curve.
Another problem, Rogani said, is the lack of benchmarks. “‘Good’ is only relative to where you think you are and where you think you’ll be,” he says. “If you’re meeting those expectations, you’ll feel like you’re ahead of the curve, but we haven’t fully established what best-in-class looks like yet.”
The gap between trust and maturity is especially evident in how companies manage their data. Even organizations with promising AI pilots can struggle to scale them if the underlying data is fragmented, inaccessible, or poorly managed. Seven out of 10 respondents said data is a challenge to using AI effectively.
24% of respondents cited data privacy and security as a barrier, and 31% said siled data across multiple sources was an issue. Additionally, 58% said they lack the skills to effectively use AI to leverage data, and 61% said they do not have consistent and quick access to data to support timely AI-enabled decision-making.
The gap between leaders and laggards in data management was wide. 44% of Leaders have access to data across the enterprise, compared to only 17% of Laggards. Meanwhile, 91% of Leaders said they use best practices or have cutting-edge data management practices, while only 61% of Laggards said the same.
For CIOs looking to strengthen their data strategy, Rogani said they shouldn’t start with a radical data integration challenge. “Don’t do a five-year data integration strategy. Those days are completely over,” he said. Instead, we recommend selecting high-impact use cases and working backwards from your desired outcomes to determine the data strategy, architecture, context layer, and semantic layer required for those use cases.
“Put yourself in the high-impact areas and work backwards,” he said. “You’ve chosen a strategy, you’ve chosen an architecture, you’ve chosen a context semantic layer. Then you can continue to build and layer on top of that architecture based on the high-impact cases you want to implement.”
Rethinking business processes is emblematic of how leaders have transformed their operating models. According to the survey, 44% of leaders said they have completely redesigned their company’s operating model to support AI. Among laggards, only 23% did the same.
Rogani said the distinction is important because most organizations making “significant changes” are redesigning processes, rather than just leaving workflows intact and moving tasks to AI.
“How would this workflow, this team, this decision change if AI were built in from the beginning?” Rogani said. “While large-scale adoption of this is still rare, people clearly recognize that re-imagining and transforming operating models are core pillars of AI.”
