When the New York Times ran the headline, “Companies Are Pouring Billions Into A.I. It Has Yet to Pay Off,” it captured a growing sentiment: major investments in artificial intelligence have yet to move the needle on earnings.
Shortly after, a widely cited MIT study reported that 95% of enterprise generative AI projects fail to deliver measurable P&L impact within six months, triggering a sell-off in AI-linked stocks and raising fears of a bubble. At first glance, these headlines suggest AI has overpromised and underdelivered. But the real issue is not technology. It is timing. We are evaluating a long-horizon transformation with short-horizon expectations.
Like earlier general-purpose technologies (GPTs), such as electrification, railroads, and the internet, AI follows a “J-curve.” The early years demand large investments in data, infrastructure, and talent while producing modest returns. Value only compounds once the systems are fully integrated.
A broad body of academic work has documented this pattern. A widely cited example is Brynjolfsson, Rock, and Syverson (2021), who describe the Productivity J-Curve. They show that GPTs often record little or even negative measured productivity growth at first, because the most important inputs are what economists call “intangible investments,” such as redesigning processes, retraining workers, and building new organizational capabilities.1 These change-process costs depress short-term productivity but are essential foundations for the longer-term payoff.
Research also shows that when productivity gains from general-purpose technologies do arrive, they lag the initial wave of investment by years. Brynjolfsson, Hitt, and Yang (2002) found that IT-driven productivity improvements were eventually capitalized into higher stock market valuations, but only once firms had invested in complementary systems and processes. Basu and Fernald (2008) similarly showed that industries investing in information & communications technology (ICT) saw stronger productivity and profitability emerge only after a delayed period, once intangible investments had taken root. In other words, the financial payoff from GPTs tends to appear on a multi-year horizon, not in the first six months. AI is likely to follow the same arc: short-term dips while investments are made, followed by durable gains that ultimately show up in productivity, earnings, and valuation.
If judged by six-month ROI windows, AI looks like a failure. If judged by its long-term potential, it is an essential capability still under construction.
The MIT study assessed more than 300 enterprise deployments, 150 executive interviews, and 350 employee surveys. Its definition of success was narrow: projects that scaled beyond pilot and produced measurable P&L gains within about six months. Only 5% of pilots met that bar.
Crucially, the roadblocks were less about AI model quality and more about workflow misfit, brittle tools, and a lack of learning or adaptation. Many projects focused on visible functions like sales and marketing, where outcomes are easiest to measure, but overlooked back-office automation, which often can provide stronger long-term ROI. The study’s sample also ranged from widely piloted consumer tools like ChatGPT and Copilot to enterprise-grade systems built in-house or purchased from vendors. Many of the “failures” were narrow pilots, often concentrated in sales and marketing use cases, that never scaled across the organization and therefore lacked leadership attention or sufficient capital allocation.
Recent studies echo similar themes.2 They highlight that AI projects often fail for reasons common to other large-scale technology efforts:
The study also found:
In short, the study does not prove that AI lacks value. It shows that value takes longer to materialize than current yardsticks allow.
Short-term pressures across the investment system can distort decisions. In AI, this often means chasing incremental, visible wins over building strategic capabilities. The result is underinvestment in the foundations that will generate durable value.
This pattern is consistent with what we track more broadly through FCLT Compass: when horizons are compressed, capital allocation tilts away from long-term growth and innovation. Similar tendencies can be seen in incentive structure, in many CEO pay packages, or in the investment mandates between owners and managers, where short-term metrics can directly undermine multi-year commitments. AI is another example of this mismatch.
Three metrics can help shift the conversation from short-term disappointment to long-term value:
Examples from practice show how organizations that take the long view are already seeing results. Translating long-term principles into action requires a clear vision, structural commitment, and patience for payoff. : The bank set its sights early on becoming a technology leader and then invested steadily in the operating model and talent to back that ambition; their trajectory illustrates how a long-horizon approach can transform both organizational culture and financial outcomes.
The DBS example highlights a broader lesson: embedding AI requires clear leadership alignment, a vision anchored in long-term goals, and the discipline to scale what works.3These principles apply across industries, and they underscore why patience and persistence are so critical when evaluating AI’s impact.
The story is not that AI does not work. It is that we are timing it wrong. The MIT study highlights a mismatch between how quickly markets expect results and how long it takes transformative technologies to mature. For boards, executives, and investors alike, the challenge is to reset the clock—aligning with the horizon on which AI truly creates value.
AI is a vivid reminder of why the long-term perspective matters.
Intangibles here means the behind-the-scenes investments in skills, systems, and process redesign: costs that don’t show up as capex but are critical to enabling long-term gains.
For further reading: BCG (2024) AI and the Next Wave of Transformation, Citi (2024) Generative AI in Investment Management: The Pursuit of a Competitive Edge, KPMG (2024) The Key to AI Adoption in the Asset Management Sector, Mercer (2024) AI Integration in Investment Management, and QuantumBlack/McKinsey (2025), The State of AI: How Organizations Are Rewiring to Capture Value.
For further reading: Practical roadmaps, such as McKinsey’s Rewired: The McKinsey Guide to Outcompeting in the Age of Digital and AI, provide complementary lessons and cases. Further, McKinsey’s complementary article Rewired to Outcompete describes how DBS Bank and others embedded AI by aligning the C-suite around focused domains and linking improvements to both short- and long-term financial goals.
Innovation, Incentive Alignment | Article
24 April 2024 - Artificial intelligence has already changed the way data is compiled and processed on a mass scale – and in particular in the investment industry, reshaping how professionals make decisions, manage portfolios, and analyze market data. A prominent challenge facing the industry today is the design of executive pay – specifically, structuring it in ways that incentivize strong performance over many years rather than just over a few fiscal quarters.
Article
22 April 2025 - From strategic supply chain realignment to national investments in AI and digital infrastructure, several emerging markets (EM) are laying the groundwork for building structural resilience. While not evenly distributed, these trends are already reshaping the long-term opportunity set in select economies.
In the News
11 April 2025 - With the ability to uncover hard-to-find information and enable more frequent trading in traditionally illiquid asset classes, new technologies like artificial intelligence and tokenisation could be the biggest disruption most private markets investors will see in their lifetime.