When Not to Use AI: A Pragmatic Guide for Leaders Who Actually Think

Published: April 3, 2026
Reading time: 5 minutes
Category: AI Without the BS

When Not to Use AI

Every conference keynote tells you to "embrace AI." Every consultant pushes another use case. Every vendor promises transformation.

Nobody talks about when to put the AI down and use your brain.

That silence is costing executives billions in misapplied technology, eroded trust, and decisions that required judgment but got algorithms instead.

Here's the contrarian truth: knowing when NOT to use AI is the real executive skill of 2026.

The "AI Everywhere" Fallacy

The current narrative treats AI like a universal solvent — just add it to any business problem and watch it dissolve. But as McKinsey's Carolyn Dewar recently wrote in Fortune: "Efficiency without empathy is not progress. Innovation without judgment is not leadership."

Recent research backs this up. A study found that 62% of leaders consider human judgment and creativity the most vital skills over the next decade — not because they're anti-AI, but because they understand what AI can't do.

Let's be specific.

Five Situations Where AI Is the Wrong Tool

1. When Empathy Is the Deliverable

Terminating an employee. Delivering bad news to a major client. Navigating a difficult board conversation. These moments aren't about efficiency — they're about human connection.

An AI can draft talking points. It cannot look someone in the eye, acknowledge their pain, and preserve their dignity. The executive who outsources this to AI-generated scripts will be remembered — but not the way they hoped.

The test: If the recipient would feel diminished knowing AI was involved, don't use AI.

2. When the Decision Is Novel (Not Routine)

AI excels at pattern recognition — finding what's similar to what it's seen before. It fails spectacularly when the situation is genuinely unprecedented.

The New York Times recently reported on a company choosing between selling two business units: one high-growth but volatile, one slow but mission-aligned. AI analysis recommended selling the slow-growth unit — mathematically logical. Human judgment recommended the opposite — strategically correct.

The test: Has this exact decision type been made thousands of times with documented outcomes? If not, AI is pattern-matching without patterns.

3. When Accountability Cannot Be Diffused

Someone has to be responsible for high-stakes decisions. AI can inform, but it cannot own consequences.

When regulators ask "who decided this?", "the algorithm" isn't an acceptable answer. The EEOC, FTC, and state agencies have made clear: existing employment, credit, and consumer protection laws apply to AI-mediated decisions. The liability stays with humans whether they like it or not.

The test: If this decision goes wrong, can you point to a human who owns the outcome? If the answer is "the model," you've already failed.

4. When Context Changes Faster Than Models

AI models are trained on historical data. In rapidly evolving situations — market disruptions, PR crises, regulatory shifts — that historical lens becomes a liability.

As one enterprise AI analysis noted: "AI does not fail because the technology is weak. AI fails because organisations deploy it without the right governance" for handling edge cases and changing business conditions.

The test: Is the situation stable enough that patterns from six months ago still apply? If not, human judgment is your real-time advantage.

5. When Trust Is Being Built (Not Scaled)

AI is excellent at scaling trust that already exists. It's terrible at building trust in the first place.

Your first meeting with a potential partner. A negotiation where relationships matter more than terms. The moment a team needs to believe in a new direction. These require presence, improvisation, and genuine human connection.

The test: Is this interaction about efficiency or about relationship? Scale with AI; build with humans.

The Judgment Layer Framework

Here's a practical approach: before any decision, ask yourself three questions:

What is the failure mode? If AI gets this wrong, is it embarrassing (recoverable) or catastrophic (irreversible)?
Who owns the outcome? If the answer isn't clear, AI is probably inappropriate.
What does "right" even mean? If success is measurable and objective (reduce processing time by 30%), AI fits. If success is subjective and relational (build board confidence), it doesn't.

The Counter-Intuitive Truth

The executives who will win in 2026 aren't the ones who use the most AI. They're the ones who know when to use it — and when to trust their own judgment instead.

As Forbes recently put it: "GenAI interprets rather than decides. As final decision-makers, employees and leaders will shift from fearing AI to thinking about how this tool can help us work better."

That's the key insight: AI helps you work better. It doesn't replace the work that matters most.

Your Monday Action

This week, identify one decision you've been considering offloading to AI. Run it through the three questions above. If it fails any of them, keep it human.

The goal isn't to use less AI. It's to use it better — which sometimes means not at all.

Tommy Kenny is a business attorney, fractional executive, and executive coach who helps leaders navigate digital transformation without losing what makes them effective. Subscribe to Digital Executive Insight for weekly frameworks that actually work.

Related Posts:

Sources

Dewar, C. (2026). "AI Won't Decide the Future — Leaders Will." Fortune, February 1.
Research.com. (2026). "AI, Automation, and the Future of Organizational Leadership."
The New York Times. (2026). "Why A.I. Can't Make Thoughtful Decisions." January 25.
VisionGroup. (2026). "5 Enterprise AI Challenges in 2026."
Forbes. (2026). "5 Workplace Predictions: Why Flexibility, AI, and Empathy Will Decide Who Wins."

When Not to Use AI: A Contrarian Guide