Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

3 reasons why bad AI is so prevalent

Seek authentic, innovative AI solutions instead.

[Photo: Getty Images]

Profile photo of Chris Erhardt

Business & Management Consultant · Chris Erhardt

Artificial intelligence has gone from a technical concept to one of the most abused words in modern business. Every startup markets itself as “AI-powered,” every legacy tool claims to provide “AI-driven insights,” and every clunky customer-service bot is suddenly an “AI assistant.”

The result is predictable. People are confused, skeptical, and increasingly convinced that the technology is overhyped.

When working with governments and businesses, I explain how modern AI can streamline regulatory compliance, improve documentation, or reduce operational workload. The most common responses I hear are, “AI doesn’t work,” or “AI can’t do that yet.”

For a while, I assumed a resistance to innovation. But over time, I realized these leaders are not rejecting real AI. They’re rejecting the disappointing tools mislabeled as AI.

Don’t call it AI

A recent experience made this observation clearer than ever. After reading about Panda Express introducing AI ordering at its drive-throughs, I decided to give the new system a try. What I encountered was not intelligence.

If I didn’t say a menu item exactly as written, the system froze. It couldn’t handle natural language, clarify substitutions, or interpret anything outside its narrow script. The system followed pre-programmed upgrade prompts and stalled when a request didn’t match its expected pattern. It wasn’t AI, but a search function with a voice on top, a rigid menu tree dressed up in buzzwords.

Another example comes from the City of Dallas. For a period of time, the city website’s “AI chatbot” attempted to help residents navigate city services. In reality, the chatbot rarely loaded. Most days, the widget simply displayed an error message. Eventually, the city removed it from the site.

Municipal leaders could use genuine AI systems to streamline operations, but instead recall the non-functional, time-wasting chatbot. Business owners could automate documentation or customer support, but remember the stiff, client-frustrating “AI assistant.”

Then, when a customer interacts with a disappointing product, they assume AI itself is the problem. If your first interaction is with a broken government chatbot or a fast-food ordering system unable to understand a basic request, AI skepticism is logical.

The failure wasn’t due to AI’s inability to answer municipal questions. Today’s language models can digest thousands of pages of ordinances, policies, and procedures and respond with clear explanations. The failure was labeling something as AI that was incapable of performing at the expected level.

These situations point to a larger trend. Companies slap the AI label onto anything automated, no matter how basic. Scripted chatbots, rigid rules engines, and decision trees are being marketed alongside systems that can interpret text, understand context, and generate original outputs.

3 reasons why fake AI is so prevalent

Several forces contribute to this cycle of AI hype and disappointment.

1. Marketing pressure: Calling a product AI-powered attracts investors, media attention, and credibility. Teams use the term to sound modern rather than to describe real capabilities.

2. Fear: Organizations know they need to adopt AI. But instead of investing in the necessary data foundations, training, and systems integration, organizations add a superficial AI feature on top of legacy infrastructure. It checks the innovation box but fails to deliver meaningful results.

3. Misunderstanding: Teams that’ve worked with basic automation often mislabel their tools as AI simply because they don’t know the difference. Then, people believe AI is overhyped because their first encounter was a mislabeled automation tool. A distrustful gap widens between what real AI can do and what many people experience under the AI label.

You judge the technology based on the version you’ve experienced—although mislabeled—not the version that actually exists.

The promise of authentic AI

The real problem is that most products today aren’t using meaningful AI. Genuine AI already works in ways many people have not experienced. AI can:

  • Summarize long regulations in seconds.
  • Analyze contracts, draft reports, and generate complex documentation.
  • Surface insights from large datasets.
  • Power customer-facing chat systems that understand context and adapt responses based on user intent.

When used well, AI can eliminate repetitive work, improve service quality, and free up time for higher-value tasks.

The truth is simple. Properly used, AI refers to systems that can interpret information, learn from patterns, and generate useful outputs with context and flexibility. When organizations deploy real AI with clear goals and good governance, the results can be transformational.

The strongest case for AI comes from quiet, consistent results. Processes that took days now take minutes. Workflows that required multiple people now require only one. Information buried in dense documents is now instantly accessible.

Leaders who genuinely want to benefit from AI must stop treating it as a buzzword and start treating it as infrastructure.

5 questions to improve your AI approach

Organizations and leaders must start asking better questions before placing the letters “AI” in a product description:

  1. What models are being used?
  2. What data are they trained on?
  3. How is accuracy measured?
  4. How does the system learn or adapt?
  5. Can we explain how it works to users and employees?

It also means being honest with customers. When leaders misuse the AI label to describe simple scripts, they damage adoption for everyone. If a tool is rules-based, label it as such.

If the AI capabilities are limited or experimental, say so. People do not expect perfection but do expect clarity and transparency. Exaggeration builds disappointment and distrust.

If organizations want people to trust AI, they must stop pretending to deploy it and actually start using it. Invest in real capabilities rather than buzzword features, prototyping rather than overpromising, and educating users rather than confusing them.

Only then will more people experience what modern AI can truly deliver. Only then will the word “AI” regain its meaning and its value.

Chris Erhardt is a business and management consultant.