The building blocks of AI adoption

When respondents were asked in their own words what would most accelerate AI impact, the themes were strikingly consistent: Fix the data, train the people, modernize the systems. But these are not isolated fixes. Organizations are attempting to solve them ad hoc, when in reality they must be addressed as part of a coordinated shift in how AI is governed, deployed and scaled across the enterprise.

The survey data confirms this. Over one-third (34%) of respondents identified data quality as the top inhibitor to AI deployment, followed by security and privacy concerns (30%), legacy systems integration (28%), and talent or skills gaps (28%). 

These are the same requisite operational fundamentals that shape every major technology adoption cycle, and they are solvable. Once addressed, they can dramatically accelerate an organization’s ability to move from pilot to production. 

The importance of good AI governance

If data quality and security are the top barriers to AI deployment, governance is the logical first response. Most organizations recognize that: About two-thirds (67%) of respondents said they establish AI governance controls before implementing pilots or production.

But that means about one-third are forging ahead without governance in place, a segment that includes 16% who said they consider governance only after issues arise and 3% who have no governance at all. For organizations that cite data quality and security as their biggest obstacles, the absence of upfront governance isn't a minor gap.

"It's critical for organizations to establish clear AI governance and accountability early," says John Huyette, AI risk leader for RSM US. "Everyone needs to know who owns AI strategy, what data can be used, and how employees are expected to interact with AI systems. Even simple guardrails can prevent problems."

Measuring the value of AI

The disconnect extends to how organizations evaluate AI's performance. Nearly all respondents (99%) said their organization measures return on investment for AI. The top measurements reveal a telling pattern: process efficiency (40%), productivity and time savings (38%), and decision quality or speed (36%).

Outcomes such as revenue growth, competitive differentiation and innovation are crucial for moving beyond operational improvement, but organizations are measuring AI for what it's already doing well, not for what it must do next. This creates a structural risk for organizations in which investment continues to gravitate toward incremental efficiency gains, and the harder work of scaling AI across core operations gets deferred.

Generative AI is a perfect example. It is the most widely adopted application in the survey, used by 73% of respondents, followed by language AI (64%) and prediction AI (61%).

Generative AI is popular because it's ideal for straightforward tasks and often shows immediate, visible results. It is also the kind of adoption that can look like progress without requiring the foundational investments in data, integration and talent that are necessary for scale. These are investments that respondents themselves said are needed.

AI adoption is advancing faster than AI ambition

The pattern extends beyond measurement. When asked which business outcomes they most want from AI, respondents gravitated toward familiar territory: increasing employee productivity (44%), revenue growth (42%) and improved decision quality or forecast accuracy (42%).

Each priority is centered on doing current work better. None of them, on their own, represent a fundamentally different way of operating. Organizations are not only measuring AI conservatively, but they’re also setting their sights conservatively too. Without that shift, AI will remain a layer of productivity tooling—not a driver of competitive advantage.

"We aren't seeing many organizations create advanced, customized solutions or agentic systems that actively require an orchestration layer yet," says Sonya King, management consulting director at RSM Canada. "The majority of organizations are investing in safe bets or the low-hanging fruit of dependable AI that enhances productivity."

The data on measurement and ambition tells the same story from different angles. How organizations evaluate AI and what they ask it to do are both calibrated to the present rather than to what's possible.

AI workforce readiness

There is an encouraging signal in the data around internal expertise. The vast majority of respondents (88%) agreed their organization has the right in-house AI experts to implement solutions effectively, with 45% completely agreeing and 43% somewhat agreeing.

At the same time, more than half (52%) of respondents said their organization is currently engaging external consultants or advisors, and another 37% plan to do so within the next year. Additionally, 56% are establishing strategic partnerships with AI providers.

This highlights an important distinction: Having capable internal teams is not the same as having the specialized expertise required to execute advanced AI initiatives at scale. Middle market organizations may feel confident in their leadership and staff to use AI for workflow improvement. Yet when it comes to strategy formulation, governance frameworks, model implementation and change management, many still look to outside sources.

"Few professionals have the requisite understanding of AI to use it for innovative solutions," says Brad Collins, a principal and leader of tax digital services at RSM US. "Those with basic AI knowledge can still create value for their organization. But the average person moves only so fast, and the technology moves much faster."

AI introduces cross-functional challenges that go well beyond technical deployment: risk management, compliance, operating model redesign and adoption strategy. Outside expertise can accelerate progress while reducing the cost of missteps, particularly for organizations trying to move quickly and responsibly.