Generative AI adoption feels like a victory for most business leaders. Your teams are using it. Your departments are experimenting with it. You might even see marginal time savings in specific workflows. The reality is far less encouraging.
A new report from Kaufman Rossin reveals a massive gap between experimentation and execution. While 94% of mid-market companies are using generative AI, 83% are stuck in the trial phase. Only 2% have operationalized AI at scale [1].
Your team is busy testing tools. Your competitors are building systems.
The Human Behavior Behind the Gap
The data problem is not a technology problem. It is a behavior problem.
Harvard Business Review’s third annual “AI in the Wild” study analyzed nearly 13,000 real-world AI use cases drawn from Reddit, LinkedIn, TikTok, and YouTube. The number one use of generative AI is not coding. It is not financial analysis. It is not process automation. It is therapy and emotional support, which grew from 5% to 11% of all catalogued use cases in a single year [2]. “Fun and nonsense” ranks third.
Autonomous agentic operations, the kind of AI that actually transforms business operations, only just entered the top ten at sixth place.
The HBR study is direct about what this means for business: most AI activity in organizations produces marginal benefits, not paradigm shifts. People use AI to speed up existing processes. Useful. But far from the promised transformation. Cases with explicit, measurable ROI remain rare in the dataset [2].
This is the gap your competitors are not talking about. Everyone is using AI. Almost nobody is using it to build a competitive advantage.
The Cost of Decentralized Adoption
The Kaufman Rossin report names the structural reason why. Individual employees and isolated departments are making independent decisions about which tools to deploy [1]. This decentralized approach creates silos, overwhelms executives, and makes enterprise-wide strategy impossible.
The HBR study puts a name to the behavioral consequence: thinkslop. It describes the lazy, approximate thinking that emerges when people delegate not just the execution of a task to AI, but the thinking that should precede it. People open a chatbot before they have figured out what they want to achieve. They accept the first output without checking it. The result is a company full of individuals generating AI-assisted content with no shared standard, no governance, and no measurable outcome [2].
The HBR study also documents what happens next: shadow usage proliferates. Employees use AI without telling their managers because company policies are vague or restrictive. One case in the study describes a developer who automated 50% of his own workload independently after management rejected his formal proposal. He kept it entirely to himself. Top-down initiatives struggle. Bottom-up ones stay invisible and never scale [2].
This aligns directly with what the National Center for the Middle Market identifies as the defining trait of the fastest-growing companies. Centralized Innovators maintain centralized control over decision-making to drive alignment and invest heavily in technology [3]. They do not let individual departments guess at strategy. They build a unified engine.
As we explored in Stop Protecting. Start Growing: What the Data Says Separates the Companies Winning Right Now, the companies that grow fastest are not the ones with the most tools. They are the ones with the most disciplined systems.
The Three Barriers to Scale
Moving from experimentation to enterprise-wide operations requires overcoming three specific barriers. Most companies fail here.
The first is the AI skills gap. Access to qualified talent who can build and integrate AI systems is limited. The second is cybersecurity. Risk management considerations are slowing deployment across the board. The third is legacy systems integration. Connecting modern AI tools with outdated infrastructure is complex and expensive [1].
These are not technology problems you solve by buying better software. They are organizational problems you solve by making deliberate decisions about governance, infrastructure, and people.
Weidenhammer, a technology consulting firm, saw a 30% time savings on routine tasks and 25% faster development cycles by starting with tools employees already used and identifying high-impact use cases [4]. They built a culture of experimentation that fed into a centralized strategy. The key word is centralized. They did not let every employee run their own experiment and hope the results would aggregate into something useful.
We have seen the same dynamic play out in accounting. As we covered in Your Firm Is Short Two CPAs. Hiring Won’t Fix It., the firms that win are not patching the problem with general-purpose AI tools. They are rebuilding how the work gets done with purpose-built infrastructure.
Stop Measuring Adoption. Start Measuring Outcomes.
Return on investment remains the universal challenge. Time savings are easy to identify. Quantifying the financial return on AI investments continues to challenge nearly all organizations [1].
The HBR study offers a clear reason why. When the most common uses are emotional support, brainstorming, and process acceleration, the output is diffuse. You cannot draw a straight line from “my team uses ChatGPT” to a revenue number. But when AI is deployed against a specific business outcome, with defined workflows and governance, the results become measurable. The study cites AI-personalized email campaigns generating a 20 to 30% lift in conversion rates as one of the few cases with explicit ROI in the dataset [2].
The difference is not the technology. The difference is the intention behind the deployment.
Most mid-market companies plan to increase AI spending, viewing it as essential to future competitiveness [1]. That investment is accelerating even as measurement frameworks remain underdeveloped. This is the critical inflection point. Leaders recognize the potential. They lack the framework to execute.
You must transition from a Dabbler or Tester to an Operator. That means moving beyond uncoordinated exploration and structured pilots. It requires building underlying infrastructure and operationalizing AI across the enterprise with measurable outcomes.
Build the Infrastructure Before the Wave Hits
The shift toward agentic AI applications is already happening. These autonomous, task-driven implementations will replace simple knowledge work acceleration. If your foundation is fragmented now, agentic AI will break it completely.
We see the same dynamic in manufacturing. As companies look for ways to realize efficiency gains, technology and automation are the clear path forward, despite the headwinds of integration [5]. The manufacturers that win are not the ones who bought the most equipment. They are the ones who rebuilt their operations around it.
Marc Cenedella, CEO of Ladders, put the individual stakes plainly: “The people who struggle won’t be the ones AI displaces. They’ll be the ones who spent all these years putting off learning it.” [6] The same logic applies at the organizational level. The companies that lose are the ones that refused to build the infrastructure. The companies that win are using AI to do work that was never feasible before.
Stop treating AI as a novelty. Start treating it as infrastructure. Define your governance model. Centralize your tool selection. Connect your data. Measure outcomes, not activity.
The 2% who are scaling AI are not smarter than you. They started building the foundation before everyone else was ready to admit they needed one.
Demand Gen Solutions helps B2B firms transform their growth strategy through revenue systems, human performance training, and strategic alignment. Let’s merge our business goals and build your growth engine today.
Frequently Asked Questions
Why are so few companies scaling AI successfully? While 94% of mid-market companies use generative AI, only 2% have operationalized it at scale. The primary barriers are an AI skills gap, cybersecurity concerns, and the technical challenge of integrating AI with legacy systems. Harvard Business Review research adds a behavioral layer: most AI usage in organizations produces marginal benefits because it is decentralized, ungoverned, and disconnected from specific business outcomes.
What is “thinkslop” and why does it matter for business leaders? Thinkslop is a term introduced in the HBR 2026 AI in the Wild study. It describes the lazy, approximate thinking that emerges when people delegate not just the execution of a task to AI, but the thinking that should precede it. For businesses, it manifests as teams generating AI-assisted output with no shared standard or measurable outcome. It is one of the core reasons AI adoption produces activity but not results.
How can companies move from AI experimentation to operations? Companies must transition from uncoordinated exploration to centralized strategy. This requires defining a clear governance model, building underlying infrastructure, focusing on specific business outcomes, and investing in employee training to support the change. The companies seeing measurable ROI are deploying AI against defined workflows, not distributing licenses and hoping for results.
