2024 is a year of reckoning for AI

The AI marketing hype, arguably kicked off by OpenAI’s ChatGPT, has reached a fever pitch: investors and executives have stratospheric expectations for the technology. But the higher the expectations, the easier it is to disappoint. The stage is set for 2024 to be a year of reckoning for AI, as business leaders home in on what AI can actually do right now. 

The Gartner hype cycle suggests that after the peak of inflated expectations comes the trough of disillusionment. Take the most recent earnings from Microsoft and Alphabet, Google’s parent company. Alphabet reported a strong quarter: revenue of $86.2 billion, an increase of more than 13 percent from the year before. Its shares fell in response by more than 7 percent the next day. Microsoft, similarly, reported its net income rose by 33 percent — and its shares, too, fell, though only 2 percent. 

Investors weren’t expecting good. They were expecting perfect. The tech sector has been on a run, with the so-called Magnificent Seven — Nvidia, Meta, Amazon, Tesla, Apple, Microsoft, and Alphabet — pulling wide ahead of the market last year. Investments in AI haven’t paid off as handsomely as the market had been hoping.

The business community’s expectations are the natural result of the breathless AI hype of the last two years, with Wall Street analysts promising that AI is nothing short of “a generational transformation” that’s “analogized to the advent of electricity.”

“We’ve gotten very frothy very quickly,” says Mark Shmulik, an analyst at Bernstein. Some investors and analysts have analogized AI to the early internet, which puts a lot of weight on the shoulders of AI companies. “But we’ve already lived through a few of these cycles, with crypto and the metaverse, that we’ve discarded quickly.”

There doesn’t seem to be much doubt that there’s an appetite among the business world for AI tools. “The interest across pretty much every industry is universal,” says Bret Greenstein of PwC. “It’s early, though, early stages of adoption for everybody.” 

Our Business Leaders expect returns on their AI investments in 2024, with 61 percent of the CEOs surveyed expecting that generative AI will boost the quality of their products and services, according to PwC’s annual survey. The PwC report warns of a “moment of truth” for trust in AI, where mistakes can have big blast radiuses; this quarter’s earnings suggest that not delivering the promised productivity gains might be the most incendiary of all.

The first place close observers like Greenstein believe adoption may occur widely is IT — especially around software development, customer service, and back-office tasks, such as processing inbound forms. That makes Microsoft’s Copilot, the “AI companion,” an indicator for the overall field. One soft measure of Microsoft’s dominance is that, much as “googling” quickly became synonymous with web search, we are talking about Copilot and not offerings from other giants such as Google, Salesforce, or Oracle, says Adam Niewinski, co-founder and general partner at OTB Ventures.

“Microsoft is ahead of Google and basically anyone else,” Niewinski says. Their partnership with OpenAI — about which, more in a minute — also lets them source development in ways that may be more important than just simple investment return, he told me. “Everyone knows ChatGPT and OpenAI and everyone knows Copilot. But it took me a couple seconds to remind myself that Google has Bard, because no one is actually using this name.” (Bard has since been renamed Gemini, because nothing says reliable like the astrological sign most likely to “escort you off the premises of your own life.”)

Microsoft’s leadership sets it up for a positive feedback loop, where more users mean more training, which means better data, which means more users. So watching how Copilot is adopted could be a proxy for how AI is being implemented, says Brendan Burke of PitchBook. “That’s an easy-to-use application that developers are testing as to whether it can meaningfully help them, or is a distraction,” Burke says.

Whether Copilot is actually worth its price remains an open question. Certainly, AI didn’t help Microsoft in search — for all the hype around incorporating AI into its Bing search engine, its market share grew by less than half a percent. Microsoft seems to have been fumbling for a good use for Copilot, too; the brand started at GitHub and moved into sales products and then into Microsoft Office apps. Even Microsoft’s Super Bowl ad is vague — if there’s a killer use case, the company hasn’t yet found it.

AI is expensive. Take OpenAI, for instance; in December 2023, its annualized run rate was $2 billion. Because that’s a figure that takes the previous month’s revenue and then multiplies it by 12, we know that means that OpenAI made roughly $167 million that month. It is nonetheless operating at a loss and will likely need to raise “tens of billions more” to keep going, the Financial Times reported. Sam Altman, OpenAI’s CEO, has been seeking trillions of dollars in investment to entirely reshape the chip industry. Meanwhile, ChatGPT’s growth has ground to a halt.

During the era of zero interest rates, big tech could pour money endlessly into its pet projects — CEO Mark Zuckerberg’s little adventure in the metaverse burned through at least $46.5 billion since 2019, Fortune reported last October. Maybe if we were still in that era, a company like Google could just pour money into AI. “I don’t think Google can light money on fire to their heart’s content on these initiatives,” Shmulik says. “We are going through a period where investors increasingly care about profitability.” That doesn’t strike me as great news for Gemini, which, so far, appears to be a slightly better version of Google Assistant and falls short of ChatGPT, according to Tom’s Guide. 

Even Google is cutting costs (12,000 jobs last year and more in 2024), and the bar for funding will likely be higher in 2024. The challenge for AI startups now is creating sustainable business models as well as bringing AI to areas that haven’t yet been disrupted, Burke told me. And the high valuations, relative to revenue, assigned to these companies suggest that VCs expect them to become tech giants in the long run.

VCs have tightened their purse strings, and even the AI sector has been hit, according to a note from PitchBook’s Burke. Excluding deals made by Microsoft and Amazon, which aren’t quite traditional VC investments, only $7.9 billion had been committed by VCs to the space by last November— which meant about the same level of spending as in 2021.

There may be fewer exits for VCs investing in the AI sector, too. As rumblings about tech antitrust have gotten louder, there have been concerns about whether acquisitions could close — Adobe’s attempt at buying Figma didn’t, after all. “Many already believe these megacap companies have way too much power, and AI is clearly a touchy point right now,” says Angelo Zino, a technology analyst at CFRA Research. “That’s why you haven’t seen outright acquisition.”

Instead, 2023’s spike in AI fundraising was large-cap companies dumping money into the space. “We might not necessarily see the scale of a deal between Microsoft and OpenAI or Amazon and Anthropic in this coming year, just given those companies establishing their leadership in the space,” Burke says.

AI companies do have a compelling pitch to anyone wanting to use their services, though. For example, Bill McDermott, the CEO of ServiceNow, has said that he can make engineers 50 percent more productive. That’s a tremendous saving on expenses, says Scott Kessler, the tech sector lead at Third Bridge. Theoretically, AI should be in high demand.

But actual adoption and real profit may not materialize just yet. So far, it’s not clear that Microsoft’s Copilot is useful enough to justify its cost, according to The Wall Street Journal. Uptake has been slower than expected; Microsoft had to drop its initial requirement that companies sign up for at least 300 subscriptions. There also seems to be a drop-off in use a month after an initial surge. But don’t worry, Microsoft has a plan to address that: it’s going to make more alerts. 

Right now, most AI companies are generating money through premium pricing for better services, says PwC’s Greenstein. But that could potentially evolve into outcome-based pricing, he says. “The idea of paying on the outcome — you know, ‘I saved your money, I’ll get some, you get some, everybody wins’ — it’s a really interesting model in AI, because there are a lot of things [where] you can assure an outcome,” he says. “And I think those will be very compelling commercialization models.”

That might help stave off disappointment. Because if AI tools still require human review, they may not be faster or better than just having a person do it in the first place. Greenstein enthused to me on the phone about using AI to summarize information, but he and I have had very different experiences with that. Maybe AI works well to summarize forms, assuming it makes no errors.

But I’ve tried using AI to summarize podcasts I don’t especially want to listen to, then listened to them to compare how well the AI worked. I find that while the AI summaries are technically correct, neither Gemini nor ChatGPT have enough context (or social skills?) to pick up on subtle digs against rivals, subtext, insinuation, or other important parts of human communication. Anything that required background knowledge was beyond the AI’s capabilities. Using AI actually added time to my task — the time it took to ask for an AI summary and read it.

Even OpenAI is trying to backpedal on the hype. In December, OpenAI chief operating officer Brad Lightcap told CNBC that he keeps having to explain to people that AI can’t dramatically cut costs or bring back growth for struggling companies. Morgan Stanley’s AI chatbot is being bypassed by wealth managers because people want to talk with other people, The Information reported. News operations attempting to replace journalists with AI-written articles have faced backlash as those articles have been wrong, offensive, or useless.

So 2024 seems like the place where the rubber hits the road. If there are real use cases for large language models, ones that save businesses money, perhaps AI will be on the path to sustainability. But if these tools come into widespread use and lead to bad publicity, lawsuits, and congressional hearings, with minimal productivity gains, the trough of disillusionment may be coming — and it might be very deep indeed.