It seems paradoxical: AI adoption has never been higher and research reports keep notching double-digit productivity gains. But ROI is lagging.
For many leaders, it feels like waiting for a punchline that never hits. The results have fallen short of the promise. So, what’s actually holding organizations back, the technology or everything around it?
We recently surveyed 1,400 business and IT leaders to capture where organizations are in their AI adoption journeys. We looked at what’s standing in the way of greater ROI, and what successful AI adopters are doing differently. If you’re a business leader, here’s what you need to know to succeed with AI.
Adoption is soaring, but results are wanting
If your last few months mirror mine, your leadership meetings have been full of AI talk with questions like: How do we operationalize AI? Where do we use it? Can we trust it? And what’s next?
We believe AI will reshape modern business. You likely believe that too, and our research shows most organizations feel the same:
- 97% of organizations are piloting AI programs
- 89% plan to maintain or increase AI spend in 2026
- About two-thirds of leaders are using AI more than they did a year ago
Despite this momentum, fewer than one in four organizations (23%) have scaled their AI pilots into production. Less than half (41%) say AI is having a meaningful impact on their organization’s ability to achieve its goals. And when it comes to agentic AI, 47% say early experimentation hasn’t produced measurable impact.
There’s a gulf between adoption and payoff, with many organizations perpetually stuck in the pilot phase.
The two biggest barriers to ROI: integration and data readiness
Adoption is almost universal, so we know that’s not the problem. According to our research, two constraints show up time and again: integration into core systems and workflows, and data readiness.
Integration
If AI lives outside your core processes and daily workflows, it doesn’t have a chance to improve business outcomes. In our survey, the #1 barrier preventing AI pilots from scaling into production was the inability to integrate pilots into systems and workflows (52%).
For many organizations, this ties back to the state of the data stack itself. The majority (88%) report dealing with legacy tech in their data stack, like inflexible, on-premises data centers. Only 12% reported having a modern data stack. AI can do incredible things at scale, but it needs the right foundation.
Integration is also a people challenge. You have to put AI into the hands of the people who know the business best, the employees talking to customers every day, the finance teams who understand budget-to-actuals, and the marketers who have a real-time pulse on what’s resonating. AI can’t produce results in a vacuum. It has to transform how people are doing their work.
Data readiness
Here’s the trickier problem. The majority of business and IT leaders don’t trust their data. Only 18% of leaders said they’re making decisions with full confidence in their data’s accuracy and validity.
That lack of trust has a ripple effect. When leaders don’t trust the underlying data, they’re hesitant to trust AI outputs that depend on it, especially for high-stakes decisions.
Leaders are far more comfortable using AI for routine tasks than strategic ones. While 48% trust AI to automate repetitive tasks, only 28% trust AI to support decision-making, and just 27% trust it to facilitate forecasting or planning.
Business leaders cite familiar concerns as to why they don’t trust AI: data privacy, security, and compliance risks; inaccuracy and bias in AI outputs; and a reluctance to allow AI to make decisions without human oversight.
So, it’s a series of cascading trust issues that start with your data and ladder up into AI systems. If you can’t trust the inputs, you won’t use the outputs. And if you won’t integrate AI into core processes and put it into the hands of employees because of that lack of trust, it will never produce the ROI you need it to. It will never get the chance.
What AI scalers do differently
A useful lens here comes from the analytics world. In 2007, analytics scholar Thomas H. Davenport and co-author Jeanne G. Harris introduced a five-stage analytics maturity model in “Competing on Analytics.” The core idea was that the organizations “winning” weren’t the ones with the most data. It was the ones who could best operationalize their data across the business, the ones with the highest analytics maturity.
This current AI moment feels similar.
Adopting enterprise LLMs is a big step in the right direction, but competitive advantage won’t ever come from having more AI tools. Just like with analytics success, it will always come from your organization’s ability to operationalize AI at scale, implementing these tools where meaningful work is done and putting them into the hands of the people who know your business best.
This requires high levels of trust throughout your AI ecosystem, which requires strong governance around your data and your AI systems.
So, what do the organizations that successfully operationalize AI do differently than their peers?
Our research found 5 distinctive factors that separate the 23% of businesses that have scaled AI past pilots and into production:
- Advanced data maturity
- Strong data governance
- Data transparency
- High-quality data
- Integration of AI into core workflows
These “AI scalers” have high-quality data that their organizations trust. They have strong governance and processes in place to ensure data and AI are being used safely and correctly. They know which data is being used by which AI systems, where the humans in the loop come in, and how and where the outputs are used.
That’s the delta, the difference between organizations that aren’t seeing ROI and those that are: trusted data, clear governance, and integration into core workflows. A modern data stack simplifies it all.
Agentic AI: a look ahead
One final note as we step into 2026. Nearly half of the leaders we surveyed said agentic AI isn’t making a measurable impact today (47%), with 26% dismissing it as mostly hype. However, when asked what would help agentic AI reach its full potential, the answers were virtually identical to the barriers I mentioned earlier: high-quality, accessible, well-governed data (49%) and seamless integration across systems (45%).
Agentic AI is still in its early days. Whatever happens this year, the legwork to create an operational and scalable AI strategy is still the same: Build trustworthy data foundations, put governance around how data and AI are used, and integrate AI into daily workflows. Position yourself for success by putting AI in the hands of employees who know your business best, whether you’re trying to leverage LLMs today or the newest AI agents tomorrow. This is what will finally turn your pilots into ROI.