The companies pulling ahead in artificial intelligence aren't necessarily the ones spending the most or deploying the fastest. According to new research from Korn Ferry, Deloitte, KPMG, and Accenture, the real differentiator is something far less glamorous: the foundational infrastructure underneath AI initiatives, including how organizations measure results, govern their systems, and prepare their workforce.

Why Are Most Companies Stuck Between Experimentation and Real Value?

Three in four organizations expect artificial intelligence to reshape their business model. Yet only 6% describe their transformation efforts as truly transformative. This gap between ambition and execution reveals a hard truth: rolling out AI tools to more people is not the same as transforming how a business actually operates.

Deloitte's research on agentic AI (autonomous software agents that can make decisions and take actions with minimal human intervention) found that while 43% of organizations are expanding AI agent deployments across functions, only 15% have reached scaled, orchestrated multi-agent deployments. Workforce readiness sits at just 20%, and only 16% of businesses said their current processes were prepared for agentic adoption. Yet 74% of leaders expect half of business processes to be redesigned around AI agents by 2030.

The disconnect is stark. Companies are moving fast on adoption but much slower on the harder work of rebuilding how they operate with humans and AI agents working together as part of their labor force.

What Separates Winners From Everyone Else?

When researchers compared organizations that rate themselves above average in AI adoption against those below average, four critical differences emerged. The front-runners have built stronger foundations underneath their AI work:

• Measurement Frameworks: Only 1 in 5 organizations has a clearly defined ROI (return on investment) framework that tracks both adoption and business impact, yet those that do are roughly eight times more likely to invest more than 10% of operating expenses behind AI.
• Shared Platforms and Data Systems: Organizations rating themselves above average are 4.9 times more likely to have or be developing a centralized machine learning operations platform, the engineering setup that lets AI run reliably at scale.
• AI-Driven Business Models: Leading organizations are 2.5 times more likely to have intellectual property or monetization tied to AI capabilities, and 5.7 times more likely to have organization-level agentic AI inside core business processes.
• Governance and Accountability: About one in three organizations have put none of seven core governance practices in place, from formal AI policy to ethics review boards. Organizations above their industry average have 3.1 governance practices in place, compared with 1.4 for those below.

The difference isn't how many tools they adopt. It's what they build to make the work reliable, repeatable, and measurable.

"We like to say humans in the lead, not in the loop," explained James Crowley, a co-author of Accenture's report on agentic AI. The distinction is deliberate because a human "in the loop" merely reviews what the agent did, whereas a human "in the lead" means accountability for the work rests with the human, not the agent.

James Crowley, Co-author, Accenture-Wharton Study

How to Build AI Infrastructure That Actually Delivers Results

• Start With Measurement: Before increasing AI investment, build an ROI framework that tracks both adoption metrics and actual business outcomes. Six in 10 organizations plan to increase AI investments next year, but only about 1 in 5 have a clearly defined way to measure whether those investments are paying off.
• Invest in Platforms and Data First: Most organizations launch AI initiatives one at a time, ending up with siloed pilots and limited data reuse. Leading organizations build shared foundations: one common platform to run initiatives consistently, connected data so teams aren't working from mismatched versions of the business, and ways to plug new projects into existing workflows.
• Frame AI as Growth Opportunity for People: Only 16% of leaders frame AI around their people's growth, but those who do see dramatically different results. Among leaders who emphasize growth heavily, 61% say their workforce is ready to use AI, compared with just 11% among those who put little or no emphasis on growth.
• Establish Clear Governance Before Scaling: One in three organizations report having no formal governance structures for AI transformation. Clear guardrails let people use AI freely because they know where the lines are and don't have to stop and ask permission at every turn.
• Define Accountability and Decision Rights Early: Accenture's research found that 50% of working hours across the US economy are being reshaped by roughly 60 digital and physical AI agents, yet AI agents are spreading across enterprise systems faster than formal governance strategies can keep up. Organizations need explicit P&L targets, human-led operating models, and clear decision rights defined before agents go live, not after.

Why Workforce Readiness Matters More Than You Think

Eighty-five percent of leaders say their people feel safe experimenting with AI and raising concerns without penalty. But only 1 in 10 leaders call their workforce extremely ready to adopt AI-enhanced processes, and just 4 in 10 say they're very or extremely ready. Most organizations are scaling AI faster than their people can absorb it.

Feeling safe to try a tool isn't the same as knowing how to use it well. Employees can learn a platform's mechanics, but the judgment to trust an output, challenge it, or set it aside develops over time. This gap between psychological safety and actual capability is one reason why adoption has tripled in the past year, yet measurable ROI remains limited.

KPMG's Global AI Pulse survey of 2,145 C-suite and business leaders across 20 countries found that 76% of businesses now see real business value from AI, up 12% in a single quarter. However, the barriers to demonstrating ROI are also accelerating, with difficulty scaling use cases and skill gaps roughly doubling quarter after quarter as the top obstacles.

What Does This Mean for Different Industries?

AI adoption looks different by industry because each sector has its own data, compliance demands, legacy systems, and competitive pressures. Financial services leads in some areas because banks operate in highly regulated environments with complex processes and significant cost burdens tied to compliance and risk management, making them ideal candidates for AI-driven transformation. Manufacturing, by contrast, has abundant operational data but structurally constrained systems, legacy infrastructure decades old, and fragmented data across disparate systems, making AI deployment more challenging despite clear opportunities.

Retail leads most industries in AI adoption, though maturity varies widely. The most advanced retailers are scaling AI against defined growth and margin objectives, while others remain limited to isolated pilots constrained by gaps in data quality, performance measurement, or governance.

Private equity firms managing over $40 billion in assets are establishing dedicated AI teams that operate at both the fund level and across portfolio companies, focusing on accelerating due diligence and improving deal execution. Mid-sized funds are taking a more targeted approach by hiring individual AI specialists, while smaller funds often lack dedicated resources and increasingly seek external guidance.

The Bottom Line: Governance and Accountability Are No Longer Optional

For many leaders, governance sounds like more process, more approvals, more reasons to slow down, exactly what no one wants when the pressure is to move fast. But that's not the full picture. Clear guardrails let people use AI freely because they know where the lines are.

The front-runners treat governance differently, not as a constraint on AI but as the thing that lets them scale it with confidence. When decision-makers can see how AI is being used and trust that it's controlled, funding the next stage of investment becomes easier.

The research from Deloitte, Accenture, and KPMG tells a single, coherent story: AI adoption is not the hard part. Most companies have agents live; adoption has increased threefold in the past 12 months. The hard part is everything adoption exposes: governance frameworks for autonomous AI agents, the gap between deployment and demonstrable ROI, the gap between piloting and true broad usage, and the leadership discipline required to convert efficiency into growth without losing accountability.

Over the next 12 to 24 months, this distinction will separate the organizations that create real value from those still experimenting. The winners won't be the ones with the most AI tools. They'll be the ones with the strongest foundations underneath.