Why This CEO Believes AI Failures Will Dominate The Headlines In 2025

Forbes

03-04-2025

While AI offers value to businesses and individuals globally, many industry experts continue to ... More question its true capabilities. (Photo by Manaure Quintero / AFP) (Photo by MANAURE QUINTERO/AFP via Getty Images)
'Majority of AI researchers say the tech industry is pouring billions into a dead end.' That was the big, bold title of a report by Futurism. The report, which quoted the AAAI's survey of 475 AI researchers, aligns with growing concerns that much of what's been described as the 'AI boom' is overdriven hype.
OpenAI CEO Sam Altman has long argued that simply scaling large language models will lead us to so-called artificial general intelligence — human-level intelligence where AI has cognitive capabilities. But other industry leaders, like Meta chief AI scientist Yann LeCun, remain skeptical. 'We're just not going to get to human-level AI by just scaling LLMs,' LeCun told Alex Krantrowitz on the Big Technology podcast. Lecun has also previously argued that current AI isn't smarter than a cat.
Gary Marcus, American psychologist, cognitive scientist and a leading voice in the AI industry, has written copiously about how there's no real moat for most of the AI models — including ChatGPT, Grok, Claude and others — that have been bandied around the industry.
In a now-widely shared thread of his predictions for where AI will be by the end of 2025 on X, Marcus said that 'we will not see artificial general intelligence this year, despite claims by Elon Musk to the contrary ' and that 'profits from AI models will continue to be modest or nonexistent,' although he noted that 'chip-making companies will continue to do well.'
Despite these debates, few would argue that AI offers no value. Even LeCun and Marcus have repeatedly acknowledged the real-world benefits of some of today's AI systems. But even at that, Jerry Haywood, CEO of Boost.ai, believes that this year's AI story will be defined not by breakthroughs, but by breakdowns.
In an interview, Haywood warned that AI failures will dominate the headlines in 2025. He said that's because too many companies have rushed headlong into an AI future with vendors who overpromise and underdeliver — and the cracks are starting to show.
When most people think about AI going wrong, they think of chatbots hallucinating — confidently spewing wrong or misleading answers. But Haywood sees something broader at play.
'There's a difference between failed implementations and use cases that never even go live,' he said. 'We hear about it when a live AI agent makes a critical mistake, but what about the thousands of enterprises that are left with nothing after vendors fail to deliver even a working solution?'
Take the 2024 case involving Air Canada's virtual assistant — dubbed 'the lying AI chatbot' by multiple industry commentators. The AI tool wrongly promised a bereavement discount to a customer. When the case ended up in court, a judge ruled that the airline was liable for what the chatbot had said, exposing just how risky poor guardrails can be.
As Haywood told me, reputational damage is only part of the equation. In regulated industries, hallucinations can open companies up to fines and investigations. 'AI failures in customer interactions don't just cause frustration,' he said. 'They can instantly erode years of built-up trust and loyalty.'
For Haywood, today's AI boom is very similar to the early days of smartphones, when the word 'smart' was slapped onto any device with a screen, regardless of its actual capabilities. The term 'enterprise-ready AI' seems to be heading down the same path.
'In 2025, we're likely to see many organizations burst,' Haywood noted. 'Those who were sold dream use cases now face a harsh reality: their chosen vendors are struggling to turn promise into performance,' he added.​
One major factor that would fuel AI failures this year, Haywood explained, is failed implementations, or the costly gap between expectations and execution. Many companies that invested in AI are discovering that their tools require extensive customization, long implementation timelines, and still fall short on delivering meaningful ROI. The result of such endeavors is a wasted budget, frustrated teams and mounting pressure from boards to show results. That's why, according to Haywood, selecting the right vendor is crucial, as a poor choice can lead to severe consequences.​
So, how do you cut through the hype to get real value from AI? Haywood's answer is that businesses should focus on areas where AI can drive the most impact, such as scaling support, personalizing interactions, or streamlining operations. 'Don't start with the technology,' he said. 'Start with the problem.'
AI has endless applications, but without a clear business case, it can quickly become a costly distraction for even the most robust organizations. Instead, enterprises should clearly map out opportunities where AI can make the most meaningful difference. Early vendor conversations are critical and businesses need to inquire about proven results in similar use cases, the time required to deliver value and the experiences of others in their industry.​
'AI should feel like a natural enhancement to your existing operations,' he said. 'When AI aligns with your needs from the start, it doesn't feel forced; it just works.'
For all the excitement surrounding AI's possibilities, business leaders need to stay grounded. Innovation for its own sake — or under pressure from the boardroom to 'do something with AI' — is always a risky path.
Beyond the allure of AI, business leaders should be more concerned about how exactly it can transform their businesses rather than a futile pursuit of the next shiny toy in town.
What executives need to know before using AI is if the tools they want to deploy really align with their organizational goals. They also need to conduct thorough due diligence in vendor selection and maintain a clear focus on practical applications of the tools they spend money on.
As we move through 2025, the headlines may indeed be dominated by tales of AI failures. But these narratives can serve as valuable lessons, guiding businesses toward more thoughtful and effective AI strategies that prioritize substance over hype.​