A recent study by the National Bureau of Economic Research found that three out of four companies are now using artificial intelligence to improve productivity, but most companies report no positive effects. A small number of companies, such as Klarna and Duolingo, have famously reversed course by replacing customer service staff with bots, only to rehire them due to customer backlash. But for most senior executives, a second wave of big spending on AI is one of their top corporate priorities in the coming years.
Over the past decade, U.S. companies have invested a total of $471 billion in artificial intelligence. In 2026 alone, just five companies plan to spend an additional $700 billion: Amazon, Microsoft, Alphabet, Meta, and Oracle. This new annual investment level is expected to exceed annually for the foreseeable future, reaching $1.1 trillion annually by 2029. And what did this historic investment in the holy grail of technology do for American companies?
- 95% of AI investments see no return at all.
- 42% of companies will abandon their AI investments by the end of 2025, often leading to thousands of job losses.
- Google wiped out $100 billion in shareholder value within hours based on a single hallucinatory chatbot response.
Now that the most expensive learning exercise in history has exposed how poorly companies understand this new technology that they are deploying at scale, how can we spend the next trillion dollars more wisely? More importantly, how can business leaders leverage this incredible set of capabilities to create opportunities, drive growth, and improve the lives of their employees, shareholders, and customers?
The path to better begins with recognizing and avoiding the pitfalls of the first wave of AI, learning from its warning signs, and working backwards from the outcomes that really matter, rather than the bullish promises that led to some of the biggest AI failures in recent years.
The automation trap
This pattern is consistent across nearly every industry. Some companies recognize AI as a priority. Leadership green-lights the budget. The team's first instinct was to Automate what already exists. Bolt AI into your current workflow and measure how fast it runs. And in many cases, these companies measured cost savings and declared premature victory without fully understanding the impact of automation. This trap is not unique to AI and is, in fact, another case of history repeating itself.
All general purpose technologies have followed the same trajectory. electricity, computers, internet. Adoption may be rapid, but meaningful outcomes, if any, will be slower. Economists call this the “productivity J-curve.” Production decreases before it increases. Because real value only comes from redesigning processes around what new technology makes possible, rather than simply plugging new tools into old processes in hopes of making imperfect output faster and cheaper.
When companies prioritize speed over delivering products and services smarter or better, the results speak for themselves.
What happens if you make a mistake
Speeding without judgment almost always ends in failure. When AI quickly produces seemingly sophisticated work, people tend to create more, faster, and cheaper, rather than using the technology to generate better inputs that improve performance. Thoughtful strategy disappears because machines often make it seem unnecessary. Assumptions are not challenged.
Volkswagen invested $16 billion to establish Cariad. Cariad is a software division aimed at developing a single AI-powered operating system for all 12 of the company's car brands. The company employed 6,000 people but had no unified planning, decision-making authority, or structure beyond the siled processes that each employee inherited from their homegrown brands.
Instead of fixing VW's problems, Cariad made a $7.5 billion operating loss, delayed the launch of its flagship Audi and Porsche cars by several years, and cost Volkswagen's CEO his job. Volkswagen then paid Rivian an additional $5.8 billion for access to software intended to create something that Cariado never could.
What are the important differences? VW was an automaker that was trying to become a technology leader by pouring money and people into this problem without changing its management approach. Rivian happened to be a technology company that made cars, and it delivered what VW needed at a much lower cost and in a shorter amount of time. This pattern has been repeated countless times on smaller scales throughout corporate America. Consistently, the pressure to prove benefits versus sunk costs leads to decisions that are far more costly than the AI was intended to prevent.
What does it mean to do it right?
The few companies that are generating real profits from AI started with a fundamentally different question. Instead of asking, “What can we do faster and cheaper with fewer people?” they first asked, “What's wrong with the decisions we're making?”
That reframing changes everything. Companies that master AI in the second wave will not only use it to produce more, but also to understand better. They analyze customers to a depth never before possible and thoroughly test strategies before allocating a budget. Additionally, by distinguishing between correlation and AI-generated illusions and causation, every dollar spent can be tied to a specific outcome rather than a vanity metric on a dashboard.
This is the difference between using AI to save costs and using AI to create new value. Companies realizing 5x to 12x returns from AI aren't just producing more content or expanding their audience segments. They make fewer, smarter decisions and target resources to validated and scalable opportunities.
Incalculable risk: talent depletion
There is another consequence of an automation-first approach that all leaders should consider. If AI can handle entry-level analytical, creative, and operational tasks and companies stop hiring junior talent as a result, who will be senior leaders in five years? The path to good decisions cannot be automated. People develop it as they work, sometimes make mistakes, and learn to distinguish what actually has an impact from most of the noise that AI generates. Companies that cut junior employees to save money in the short term have a negative impact on their future leadership pipeline and demonstrate a lack of commitment to investing in talent and fueling the current wave of entrepreneurial competition.
The next question for every executive
The question all leaders should stop asking is: “Are we using AI?” you. So does everyone. The questions that separate the next chapter from the last are more difficult and more important. “With AI, what are we seeing that we couldn't see before, and what are we doing about it?” Five years from now, when the story of this second wave of AI is told, the companies that best answer that question will not simply survive the AI revolution. They will prove the business case for generations to come.
