Seeing the never-ending headlines about artificial intelligence, it’s easy to conclude that AI is about to transform society overnight or, just as easily, that it is totally overhyped. We have a different reading than either of those extremes: Adoption across the corporate landscape is coming—not at the blistering pace that some have expected, but steadily and inexorably.
At this point, we acknowledge, most big businesses are trundling along in the AI slow lane. We can see this by using a large language model to scour corporations’ federal financial filings and then score each firm over a 10-year period on the text describing its deployment of AI. Through the end of last year, our analysis shows, less than a quarter of S&P 500 companies had AI either deeply integrated into their business processes or were using AI in the production of goods and delivery of services.
In the AI fast lane, not surprisingly, is the technology sector, which accounts for two-thirds of extensive AI integration and use. By contrast, fewer than two dozen non-technology S&P 500 firms—including Moderna, Mastercard, Bank of New York Mellon, and GE Healthcare—have achieved full AI deployment. (We define full deployment in our model as AI being “a core component of the firm’s strategy and financial performance, deeply embedded across business functions and operations.”)
And while technology firms are heavily weighted in stock market indices, it is non-technology firms that make up 90% of the U.S. economy. It is among these enterprises where a tremendous amount of value, as yet untapped, will be created for both businesses and workers.
Several factors will drive AI uptake across corporate America, including continual improvement of the technology. For AI to meaningfully affect the way work is done, it must be able to perform a task with access to the right information and be cost-effective. In many cases, AI models remain too expensive to run at the level of precision required. For now, the tasks the technology can theoretically automate remains much bigger than those it actually automates.
Nonetheless, advances are occurring. Imagine that you were asked to create a 10- to 12-slide presentation for a quarterly customer review on a nine-month SaaS integration project. Before AI, such a task might take three to four hours to complete. Our research shows that two years ago, large language models could knock out this sort of project instantly at about a 50% success rate. A year later, LLMs could do something like this at a 65% success rate. If these trends persist, LLMs will complete most text-related tasks with good-enough quality at success rates of 80% to 95% by 2029.
More than half of the S&P 500 have AI pilot projects underway. Some will succeed; many will fail. Lessons will be learned. Success will be copied. This portends significant gains in productivity and efficiency—and, in turn, higher revenue and profits—that will simply be too good for most businesses to pass up and that will push them far beyond, say, using ChatGPT to write a job description.
Consider, for instance, a supermarket chain with 500 stores, each handling 50,000 stock-keeping units, or SKUs—the unique, alphanumeric code retailers use to identify, track, and manage their inventory. That’s 25 million SKUs that must be forecast every two weeks, a highly dynamic process that’s not always easy to get right, even when using a sophisticated IT system.
Deploying AI to more accurately project demand for different products, and then sharing that information with vendors throughout the supply chain, promises to better meet consumers’ needs. And that will be a real difference-maker in a low-margin business like groceries, where it would mean less wasteful spending on items customers don’t really want.
Similarly, project management at a giant manufacturing or construction firm tends to be very labor-intensive. Across these industries, AI will help pinpoint whether things are lining up on time and on budget—and if there’s a miss, where and why.
Beyond boosting business processes, AI will be just as critical, if not more so, for creating new products and services. It will hasten the development of breakthrough drugs at pharmaceutical companies, lead to the engineering of drought-resistant seeds at agricultural concerns, and allow automakers to come up with more innovative vehicle designs, to cite just a few examples.
None of this, of course, is guaranteed. Companies must reorganize the way work gets done while fostering a culture that encourages employees to embrace new roles augmented by AI. We are not Pollyannaish in this regard. AI will undoubtedly lead to layoffs, in some cases on a large scale. But we are optimistic overall, confident that most businesses will wind up moving beyond the simple substitution of technology for workers.
By building on the complementary efforts of workers and technology, business leaders will devise new products and services, introduce entirely new business processes, create better tools, and enable new goals. All of that will translate into new hiring.
We believe this is especially likely as more executives come to understand the advantages of partial automation, in which AI models do some of the work while humans perform the remainder.
Partial automation deserves a central role because of the way the cost of AI increases. Achieving “good” accuracy on a task may be relatively inexpensive, but pushing from “good” to “near-perfect” accuracy can be significantly more expensive for both AI builders and AI users (in terms of token spend).
When the marginal cost of further accuracy exceeds the marginal labor saving, firms do well to stop short of full automation. For many companies, then, partial automation will not merely be a transitional state on the path to full automation. It will often be the best arrangement, period.
Meanwhile, the fact that most companies are moving at something less than breakneck speed on AI shouldn’t be surprising. Our research shows that those considering AI adoption often face a variety of bottlenecks. Some have difficulty customizing solutions or accessing necessary data. Others are struggling to figure out which business processes stand to gain the most from AI.
More generally, many corporate leaders are still assessing the risks of adoption. As time passes without a major mishap, they will become less cautious, the pace will pick up, and the financial rewards will reinforce further AI deployment. This is because AI has a “J-curve” effect on the bottom line, with those in the later stages of adoption enjoying more profitability than those in the earlier stages.
Whether AI adoption will be truly widespread by 2030 or 2032 or even a little later is anybody’s guess. But whatever the case, you can count on this: A wide swath of corporate America will have moved from the AI slow lane to the fast lane soon enough.