Is generative AI worth the money? Technology leaders like Microsoft Corp. Chief Technology Officer Kevin Scott says costs will come down and capabilities improve, but as Wall Street heads toward correction territory — Nvidia Corp. and Microsoft, two stocks that have ridden the AI wave, are down more than 15% and 8% respectively since July 10 — enterprises are grappling with a deeper problem: How do they put AI to use and measure the return on that investment?
Confusion around the answer has led to a growing malaise in recent months, sparked by bearish reports from Goldman Sachs Group Inc. and Sequoia Capital questioning whether “genAI” will make as much money as the market seems to think. AI capital expenditure will reach between $600 billion and $1 trillion in the coming years, the two reports estimate, and spending on information technology will rise 8% this year, according to a prediction from Gartner Inc.
If investors are taking a breather, that’s a good thing. As my colleague John Authers recently argued about big tech stocks, corrections can be healthy when a market position has been extreme. The AI-driven market boom and a startling jump in capital spending at companies such as Alphabet Inc. and Tesla Inc.1 has happened far too quickly, benefiting those who didn’t always deserve the rally. A senior AI executive at a large tech company recently told me point blank that artificial intelligence, despite being his bread-and-butter business, had become overblown in the market. He’d lived through several hype cycles before, he added, and this one was no different.
Today at least, there are healthy differences to something like the dotcom boom and bust. Businesses are emphasizing tangible outcomes more than they were back in the early 2000s, when the focus was more speculative and based on market potential. Metrics for success are focused on efficiency gains and cost savings instead of grabbing eyeballs. And businesses are using more sophisticated risk models and arguably more rigorous ROI calculations because they’ve learned hard lessons from the dotcom era.
The nagging problem for generative AI is that tech firms have set expectations too high by marketing the technology as a magical, general-purpose solution that can enhance lots of business processes. It often can’t right away. Some companies are also rapidly rolling out gizmos to their workforces without showing how to use them effectively. That’s always a mistake. Give a power tool to someone with training and they can build a treehouse; give it to an amateur and they’ll make a mess.
But even while the process is often fitful, generative AI is taking hold in several markets. In the video games industry, major studios like Activision Blizzard Inc. are using it for concept art and asset generation, leading in some cases to widespread layoffs, according to an investigation by Wired magazine published this week.
JPMorgan Chase & Co. Chief Executive Officer Jamie Dimon recently told Bloomberg News that the bank was putting AI into all of its processes — “trading, hedging, research, every app, every database” — sometimes to replace humans. Payments company Klarna Bank AB said earlier this year that its customer service chatbot did the work of 700 customer service reps, meaning it could reduce the number of contractors it regularly hired, saving it $40 million a year.2 TravelPerk, a startup for booking corporate travel, says its margins have grown to 70% from less than 40% since it started using generative AI last year.
Analyzing AI’s return on investment is difficult. It involves taking abstract values like efficiency and productivity and turning them into numbers. Take CNH Industrial NV, a farming and construction equipment manufacturer. It’s been using an AI-powered chatbot to guide its service technicians on repairs, and to help its software developers write code, according to a profile in CIO.com. It’s easier to measure success on the first project (did customer satisfaction scores go up?), but not so much for the second (did coders log how much time they saved?).
And don’t forget that AI is also a controversial technology. Klarna’s CEO Sebastian Siemiatkowski got some blowback from users of X when he posted details about how the company’s AI assistant “performs the equivalent job of 700 full time agents.” Some suggested it was callous to celebrate the displacement of so many human workers. If companies find that generative AI is helping their bottom line, they might not always want to talk about it.
Businesses and investors risk going too far in dismissing generative AI as all hype, especially if that means they’ll neglect some of its more insidious traits around the perpetuation of bias, copyright infringement and the erosion of human agency. The current skepticism is healthy and much needed — but in at least in some markets, including customer service, marketing, gaming and other creative industries, it’s also clear that generative AI is here to stay.
1. Alphabet invested $13.2 billion on AI and computing power in the most recent quarter, compared with the $12.2 billion expected by analysts.Tesla spent $600 million on AI infrastructure as operating expenses climbed to $3 billion, a 39% jump from a year ago.
2. Klarna also halved its marketing team and now gets generative AI to do 80% of its copywriting, its CEO, Sebastian Siemiatkowski, said this month on the Big Technology podcast. Generative AI’s tendency to spout mistakes was an issue, but comparing transcripts of Klarna’s chatbots and human agents suggested the AI tools were either on par or better than the humans for errors.
A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out some of our webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Parmy Olson
