GenAI paradox: Companies investing billions on AI without seeing results
Nearly four decades ago, during the height of the personal computer boom, a phenomenon referred to as the “productivity paradox” emerged. It highlighted the paradox that, in light of substantial corporate expenditures on innovative technologies, there remains limited evidence to suggest a parallel increase in worker productivity. Currently, a similar paradox is emerging, this time in the realm of generative artificial intelligence. Recent research from McKinsey & Company indicates that nearly eight in 10 companies have adopted generative AI; however, a similar proportion has noted “no significant bottom-line impact.”
AI technology has been advancing rapidly with chatbots such as ChatGPT, driven by intense competition among major tech firms and affluent start-ups, leading to the anticipation that various sectors, including back-office accounting and customer service, will undergo significant transformation. However, the returns for enterprises beyond the technology sector are trailing, hindered by challenges such as the frustrating propensity of chatbots to fabricate information. Consequently, firms will need to persist in allocating substantial capital to remain competitive; however, it may take years for the technology to yield a comprehensive economic benefit, as enterprises slowly determine the most effective strategies.
Refer to it as “the generative AI paradox,” a term coined by McKinsey in its research report. According to IDC, a technology research firm, investments in generative AI by businesses are projected to rise by 94 percent this year, reaching a total of $61.9 billion. According to a survey conducted by S&P Global, a data and analytics firm, the proportion of companies discontinuing the majority of their AI pilot projects increased significantly to 42 percent by the end of 2024, compared to just 17 percent the year prior. This survey involved over 1,000 technology and business managers. Projects encountered failures not solely due to technical challenges, but frequently as a result of “human factors” such as resistance from employees and customers or deficiencies in skills, as noted by Alexander Johnston, a senior analyst at S&P Global.
Gartner, a research and advisory firm known for its analysis of technological “hype cycles,” forecasts that AI is entering a phase referred to as “the trough of disillusionment.” According to John-David Lovelock, the chief forecaster at Gartner, the nadir is anticipated in the coming year, prior to the technology ultimately establishing itself as a validated instrument for productivity enhancement. Historically, the trajectory of past technologies such as personal computers and the internet has exhibited a distinct pattern: an initial phase of exuberance, succeeded by the arduous process of mastering the technology, ultimately culminating in a profound transformation of industries and labor dynamics.
The beneficiaries to date have been the providers of AI technology and consultancy services. Prominent players in the AI sector encompass Microsoft, Amazon, and Google, all of which provide AI software solutions, whereas Nvidia stands out as the dominant force in the AI chip market. Executives at those companies have asserted that AI is transforming their workforces, reducing the necessity for certain entry-level coding positions and enhancing the efficiency of other employees. Many predict that AI will eventually replace entire swaths of human employees, a perspective that is being widely embraced and echoed in the corporate mainstream. During the Aspen Ideas Festival in June, Jim Farley, the chief executive of Ford Motor, stated, “Artificial intelligence is going to replace literally half of all white-collar workers in the US.”
The occurrence and timing of such revolutionary change are contingent upon the empirical testing environments provided by numerous businesses. “The raw technological horsepower is terrific, but it’s not going to determine how quickly AI transforms the economy,” stated Andrew McAfee, a principal research scientist and co-director of the Massachusetts Institute of Technology’s Initiative on the Digital Economy. Nonetheless, certain enterprises are discovering methods to integrate AI, albeit in many instances, the technology remains significantly distant from supplanting human labor. USAA serves as a case study for the dual nature of AI, showcasing both its potential benefits and inherent limitations in the context of providing insurance and banking services to military members and their families.
Following the completion of various pilot initiatives, some of which were subsequently terminated, the company launched an AI assistant aimed at enhancing the accuracy of responses provided by its 16,000 customer service employees to particular inquiries. USAA is monitoring its investments in artificial intelligence; however, it has yet to quantify the financial returns, if any, associated with the call center software. However, the feedback from its employees, according to the company, has been predominantly favorable. Despite the availability of software applications designed to address customer inquiries online, the call centers still manage an impressive volume of approximately 200,000 calls each day. “Those are moments that matter,” stated Ramnik Bajaj, the chief data analytics and AI officer of the company. “They desire a human voice at the other end of the phone.”
This parallels an AI application created over a year ago for fieldworkers at Johnson Controls, a significant provider of building equipment, software, and services. The firm has integrated its operational and service manuals for its machinery into an AI system designed to produce a summary of issues, recommend repairs, and relay this information directly to the technician’s tablet. In testing, the app has reduced the duration of a repair call by 10 to 15 minutes from an hour or more — a notable efficiency improvement, yet not a complete overhaul of workplace dynamics by itself. Currently, fewer than 2,000 of the company’s 25,000 field service workers have access to the AI helper; however, plans for an expansion are underway. “It’s still pretty early days, but the idea is that over time everyone will use it,” stated Vijay Sankaran, the chief digital and information officer at Johnson Controls. The long-term vision posits that firms will leverage AI to enhance various systems, encompassing sales, procurement, manufacturing, customer service, and finance, he stated. “That’s the game changer,” stated Sankaran, who anticipates that this transition will require a minimum of five years.
Two years prior, JPMorgan Chase, the largest banking institution in the United States, restricted access to ChatGPT on its systems due to concerns regarding potential security vulnerabilities. A limited number of data scientists and engineers were granted the opportunity to engage in experimentation with AI. Currently, approximately 200,000 employees at the bank utilize a general-purpose AI assistant — fundamentally a business chatbot — via their work computers for functions such as data retrieval, addressing business inquiries, and report generation. The assistant, specifically designed for JPMorgan, utilizes ChatGPT and various AI tools, all while maintaining data security for sensitive bank and customer information. Approximately 50% of the workforce utilizes it consistently, indicating a reduction of up to four hours per week in fundamental office activities, according to the company’s report. The bank’s wealth advisers are utilizing a more specialized AI assistant that leverages bank, market, and customer data to furnish affluent clients with investment research and guidance. The bank asserts that it gathers information and assists advisers in formulating investment recommendations at nearly double the speed compared to previous methods, thereby enhancing sales.
Lori Beer, serving as the global chief information officer at JPMorgan, manages a technology workforce comprising 60,000 individuals across the globe. Has she ceased operations on AI initiatives? She indicated that the total likely amounts to hundreds. However, numerous shelved prototypes, as she indicated, generated concepts and code that were integrated into other ongoing projects. “We’re absolutely shutting things down,” Ms. Beer stated. “We’re not afraid to shut things down. We do not consider it to be a negative development. I believe it is a prudent decision.” McAfee, the research scientist from M.I.T., concurred. “It’s not surprising that early AI efforts are falling short,” stated McAfee, a founder of Workhelix, an AI-consulting firm. “Innovation is a process of failing fairly regularly.”
