CATEGORY: Developments
Companies are currently pouring billions into AI with the frantic energy of a gold rush, but the ROI is largely hallucinated. PwC’s latest Global CEO Survey reveals that CEOs have the lowest level of confidence in five years for their revenue outlooks.
It’s becoming clear that the problem isn’t whether AI is valuable; it’s that many organizations are deploying it without an explicit, coherent business model, which taints their ability to judge the potential value of their investments. Without a framework that clearly defines how value is created, how it is monetized, and who is accountable for outcomes, even the most advanced AI tools struggle to deliver meaningful returns. If leaders don’t address quality, risk, and monetization together, AI won’t be the “great unlock” – it will become a quiet but massive value destroyer.
AI’s Weak Foundation
The current AI landscape is built on a shaky tripod:
- Quality (The Hallucination Problem): AI doesn’t just get things wrong; it gets them wrong with confidence. In mission-critical fields like healthcare or law, “mostly right” is a failure state, not a success metric.
- Risk (The Liability Gap): We are seeing a massive shift in the legal landscape. I’ve long maintained that AI platforms are legally responsible for their outputs. If a system provides a medical diagnosis or legal advice that causes harm, the “it’s just a beta” excuse no longer holds water. Business models that fail to assign responsibility for AI-driven outcomes are structurally incomplete.
- Monetization (The Profit Void): The open secret in Silicon Valley is that most AI companies are losing money on every query. Burning VC cash to subsidize compute costs isn’t a business model; it’s a countdown to a cliff. We’ve seen this movie before – some winners and a lot of losers.
Taken together, these flaws explain why AI investment is outpacing AI impact. With CEO revenue confidence at its lowest point in five years, the window for expensive, aimless experimentation is closing. 2026 is shaping up to be the hard deadline where business models must finally align with reality.
AI Needs Structure to Create Value
Without a clear business model, AI remains an experiment. That structure is not technical; it’s organizational. It defines which decisions AI supports, how success is measured, and who owns the outcomes. A lot of this is attributed to the fact that teams treat AI like an add-on feature instead of a tool enabling larger structural changes. In so doing, leaders are rushing to replace humans to make up for costs, indicating a reactive approach.
While AI can accelerate processes and identify patterns, it cannot understand context, weigh risk, or be accountable for consequences. Those responsibilities sit squarely with human leaders. The real opportunity is not removing humans from the system but deliberately designing where human judgment is required for an AI-driven model to function effectively.
When leaders fail to acknowledge this gap and expect AI to drive success independently, strategies underdeliver. Models are being implemented, yet 95% of initiatives are failing because organizations haven’t determined which decisions AI is meant to inform, how they will measure whether it has improved outcomes, and who is accountable if it doesn’t. The result is slowed execution rather than acceleration. Impact remains limited, and the consequences are wasted capital, low adoption, and tarnished trust.
Where many organizations go wrong is that they’re taking incremental approaches to AI adoption and implementation, producing only short-term results. In making the mistake of layering AI onto existing systems and workflows, 78% of companies report they’re struggling to make it work. What would deliver results is taking a few steps back to design a business model that AI can reinforce rather than one that’s awkwardly layered onto.
From “Clicks” to “Clients”: Redesigning the Value Prop
The failure of AI monetization stems from a legacy mindset: rewarding activity over outcomes. Most models still prioritize “clicks,” “queries,” or “seats.” But in an AI-driven world, activity is cheap, and noise is infinite.
Customers don’t buy AI for volume; they buy it for confidence, correctness, and safety, attributes that only matter when outcomes are clearly defined. This is where business model clarity becomes decisive.
AI systems don’t just make mistakes; they do so with authority. When hallucinated outputs influence medical, legal, or financial decisions, the cost of inaccuracy becomes real and unavoidable. When revenue is tied to outcomes rather than usage, organizations are forced to define success precisely, embed human oversight where risk is highest, and take responsibility for results.
We’ve seen this play out first-hand at Pearl by shifting the focus entirely from activity to outcomes. We stopped delivering ‘clicks’ and started delivering clients. This distinction is the difference between a business model that burns cash and one that builds wealth. By being ‘on the hook’ for the final result, we forced our AI to be a tool for conversion, not just a generator of digital clutter.
The Judgment Layer: Why “Human-in-the-Loop” is Non-Negotiable
The rush to replace humans is the most expensive mistake a CEO can make. Human judgment will always have a place in organizations, regardless of how advanced AI becomes.
While AI excels at pattern recognition and scale, it lacks context, ethical reasoning, and accountability – the very attributes required in high-stakes decision-making. Human involvement should not be reactive or bolted on after failure. It must be designed into the business model from the outset.
In healthcare, legal, and financial services – and increasingly across all industries – AI should do everything except make the final judgment call. Every business, regardless of sector, navigates legal exposure, financial risk, and regulatory constraints. When AI influences those domains, human judgment is not optional. Every organization should establish a liability threshold: any AI-generated output involving a financial or legal risk exceeding a specific internal dollar amount must require mandatory human sign-off before execution.
That final mile (oversight, judgment, and accountability) must belong to a human. This isn’t a limitation of AI; it is a prerequisite for trust and scale. A well-designed business model treats AI as a super-agent: an engine for data, execution, and organization, while reserving judgment and responsibility for human leaders.
The 2026 Mandate
The competitive advantage of the next decade won’t belong to the company with the fastest model or the most impressive demo. It will belong to the leaders who redesign their business models to blend AI efficiency with human judgment and who make value creation, pricing, and accountability explicit.
The better question is no longer, “What can AI do for us?” It’s: “How should our business model change so AI is profitable, accountable, and human-led?” If organizations fail to build a clear judgment layer into their models, they aren’t innovating. They’re simply waiting for the hallucinations to show up on the P&L.
On Monday morning, ask your AI leads three questions: Are we paying for activity or outcomes? If the AI makes a six-figure error, who—specifically—is on the hook? And where, exactly, is the human ‘Judgment Layer’ embedded in our workflow?
Author:
Andy Kurtzig
Email:
authors@the-ceo-magazine.com
LinkedIn:
https://www.linkedin.com
Website:
https://www.pearl.com/
Andy Kurtzig is the CEO of Pearl, the super-agent platform powering the independent professional economy by delivering clients, not just clicks.
