(This column originally appeared in Forbes)
AI is a wonderful thing, right?
Well it certainly is for the big technology companies making and supporting it. The stock prices of Microsoft, Google, Amazon, Nvidia and other companies that are in the middle of the AI boom have skyrocketed over the past few years. Billions are being plowed into data centers, infrastructure and hardware to support the new technology’s hungry needs. Indirect companies from U.S. Steel to HP to local utilities and countless obscure startups are benefiting from the construction, energy consumption and compute demands created by this industry.
Hallucinations, Inaccuracies, Misinformation: The Data
And yet, the public is not convinced. For most, AI is fun and can even be productive. But AI applications — with their hallucinations, inaccuracies and misinformation — remain unreliable.
A recent KPMG survey of more than 48,000 people globally found that although about 66 percent of them say they use AI regularly, only 46 percent felt willing to trust AI systems. Another recent survey of more than 1,100 people found that about 82 percent are somewhat skeptical of AI “overviews” in search: 61 percent “sometimes trust” the results, 21 percent never trust and only about 8.5 percent always trust the answers provided. Another survey from Gartner backed up these findings, with about 53 percent of consumers saying they don’t trust the results of AI-powered searches or summaries. Why?
According to these surveys, many users keep seeing significant mistakes: 42 percent report that they experience inaccurate or misleading content, about 36 percent report missing important context and as much as 31 percent reporting bias in overview results.
Even the people writing software are dubious. A poll of developers found that although 84 percent of them plan to use AI coding tools, only about a third trust their outputs which is actually down from earlier years. According to the report, their main frustrations are around “almost right” results, which ends up costing extra debugging time.
How can any product that works so poorly be hyped so much?
Hallucinations, Inaccuracies, Misinformation: Big Tech Is The Problem
Blame big tech. Since OpenAI released ChatGPT three years ago, software companies, in an effort to keep up, have been rolling out their AI offerings that have not been ready for prime time. And yet these companies keep pushing their customers to buy their AI applications and use their AI agents to increase productivity, when in many cases the exact opposite is happening.
According to a report from NBC News AI “slop” (blurry logos, nonsensical text, generic or unpolished writing/code) is forcing companies to hire freelancers, artists, writers and developers to correct or finish what AI got wrong. These companies are finding that many fixes involve more effort than expected with many finding it just easier for the human worker to start from scratch rather than patching what AI produced.
“In April 2024, it seemed like agentic AI was going to be the next big thing,” writes Steven Newman, an AI expert. “The ensuing 16 months have brought enormous progress on many fronts, but very little progress on real-world agency.”
Hallucinations, Inaccuracies, Misinformation: It’s Not All Their Fault
It’s easy to blame big tech companies for the failure of their software to actually work. And they deserve a lot of the blame. But so do the companies — particularly large companies — who are throwing hundreds of millions of dollars at this stuff without properly thinking it through.
The New York Times recently reported that of the 80 percent of companies using generative AI, just as many say they’ve seen no significant bottom-line impact, with as many as 42 percent of companies reporting they they abandoned most of their AI pilot projects by the end of 2024, up sharply from 17% a year earlier. And yet, the Times reports, businesses continue to increase their investments “aggressively,” with generative AI spending expected to nearly double this year. It’s like their gluttons for punishment.
A another report from MIT also says that it’s not necessarily big tech’s fault, but rather how customers are deploying AI. It found that about 95 percent of AI pilot programs fail to deliver measurable profit-and-loss impact and only about five percent of these pilots are achieving rapid revenue acceleration. Those are the ones “focusing tightly on specific problems and executing well.”
The report found that many unsuccessful pilots didn’t fail because the AI models were bad, but because the tools were poorly integrated with existing workflows, didn’t adapt to company needs, or lacked learning-capable systems. The MIT researchers suggest back-office automation, finance and procurement are among the areas where the return-on-investment is better, but many firms instead are investing more heavily into functions like sales and marketing which are harder to scale.
“The 95 percent failure rate for enterprise AI solutions represents the clearest manifestation of the GenAI Divide,” writes Zvi Mowshowitz, a former hedge fund manager and long-time AI commentator. “Organizations stuck on the wrong side continue investing in static tools that can’t adapt to their workflows, while those crossing the divide focus on learning-capable systems.”
He’s right. I use AI for simple applications all the time (for example, helping me to summarize research and content from various sources that I’m using to help write this article). Many business people I know are happily using it to transcribe conversations, write emails, draft company policies and perform rudimentary analysis. For these kinds of activities, generative AI platforms can be very useful. But is it worth all the hype? All the high valuations?
Hallucinations, Inaccuracies, Misinformation: The AI Bubble
Chevaugn Powell, author of “The Trillion-Dollar AI Bubble Nobody Sees Coming” thinks all of this AI hype is creating a bubble not dissimilar to the dot-com collapse in 2001.
“According to research firm Gartner, spending on generative AI will reach $644 billion this year alone,” he writes. “While, last year those same hyper-scalers generated only $45 billion in actual AI-related revenue.”
He also points out the vulnerability of the entire industry to new innovations.
“A single Chinese startup (DeepSeek) proved that the emperor has no clothes in Silicon Valley’s AI kingdom,” he said. “When DeepSeek announced it had built an AI model rivaling ChatGPT for under $6 million, tech markets didn’t just wobble — they absolutely cratered.”
Still, it’s big tech that benefits. The pundits, experts and academics warn us that we “must” be investing in AI or risk falling behind, losing out or even going out of business. But it’s clear that the majority of business people still don’t think it makes sense to invest in something that clearly doesn’t work very well. Maybe there is a bubble.
Most of my clients — small and mid-sized business owners — are watching this technology warily. They see the potential. But they certainly trust it right now. And thanks to the disappointments, their trust levels for those big tech companies is also low. None I know are relying on AI to help run or manage their core business processes. And you can’t blame them. Big tech is hyping products that don’t work very well. And their credibility is getting hurt because of it.
