A year or so back, generative AI was being touted by some as the most revolutionary development since the splitting of the atom. And, ironically, large vendors are turning to the technology of splitting atoms to power AI, as seen with Microsoft’s intentions to restart the Three Mile Island reactor.
Lately, however, generative AI has been seen more as in a bubble that’s on its way to deflating.
“While talk of a bubble has simmered beneath the surface while the money faucet continues to flow, we observe a recent inflection point,” wrote David Gray Widder, Postdoctoral Fellow at the Digital Life Initiative at Cornell Tech, and Mar Hicks, associate professor of data science at the University of Virginia, in an essay published by Harvard University’s Ash Center. “Interlocutors are beginning to sound the alarm that AI is overvalued. The perception that AI is a bubble, rather than a gold rush, is making its way into wider discourse with increasing frequency and strength. The more industry bosses protest that it’s not a bubble, the more people have begun to look twice.”
The AI hype cycle is simply following the patterns of past technology bubbles, Widder and Hicks argued. “Efforts to make AI indispensable on a large scale, culturally, technologically, and economically, have not lived up to their promises. In a sense, this is not surprising, as generative AI does not so much represent the wave of the future as it does the ebb and flow of waves past.”
Is this a fair assessment? While industry leaders concur that AI — particularly generative AI — is in a hype cycle, it is nonetheless already is delivering on many of its promises. “The hype is over the top, but the reality is that 85 percent of the G2000 are experimenting with gen AI solutions and are beginning to adopt AI at scale,” Steven Hall, partner and president of global technology research and advisory firm ISG, told me. “There are thousands of use cases and pilots in progress.”
While “the AI hype is real,” many organizations around the world are reaping gen AI rewards – driving productivity gains, delivering new customer and employee experiences, powering the development of new digital products and services, and creating meaningful business value.” said Matt Candy, global managing partner at IBM Consulting.
Beyond the bubble, it’s unclear how AI will reshape the world. “Could there be an AI-bubble? Yes,” said Gabriel Werner, field chief technology officer at Blue Yonder. “Is there any doubt that AI will have a lasting and profound impact on everyone? No. Do we all already know how that’s going to play out? No.”
The perception of what constitutes AI hype is at issue, Candy added. “it’s less about whether AI can rise to expectations, and instead whether organizations can take an enterprise-wide approach to AI adoption, shifting from AI as supplementary, to scaling with an AI-first mindset, using an open-source, multi-model approach, grounded in humans, trust and governance.”
Werner sees AI “is a global game changer like the internet was in the late ‘90s. We didn’t really know what it would mean for our day-to-day lives at the outset. That’s where we are with AI.”
Tellingly, “AI generated over $10 billion of new revenue with global service integrators over the trailing 12 months,” Hall pointed out. “This was up over 60 percent quarter-over-quarter and kept most service integrators in positive territory amid a pullback in managed services. AI represents less than two percent of the current outsourcing market, but it is growing at a tremendous rate.”
For its part, generative AI has helped skyrocket AI ROI from 13% to 31% since 2022, and operating profit gains directly attributable to AI doubled to nearly 5% from 2022-2023, Candy pointed out. “Several of our clients are already experiencing significant productivity gains.”
To turn the promises of AI into reality, AI proponents need to overcome “include unclear business strategies, complex data challenges, risk and governance implications, skills shortages, as well as infrastructure and cost considerations,” said Candy. “Overcoming these challenges requires an interconnected effort across the whole organization.”
Industry leaders also offer the following steps:
Recraft ROI expectations. Rather than straightforward ROI, look for a return on AI, or “ROAI,” Candy said. Find out if employees and customers trust these investments – “measured by adoption, engagement rates and user satisfaction. Trust can also be extended to model accuracy, data transparency, fairness and accountability.”
While the ROI for AI as applied to tasks such as software development, defect reduction, and testing is showing tangible results, “ROI for revenue-generating activities is still in the early phases,” said Hall. “We are only several quarters past the launch of GPT 3.5, which democratized AI. During this short period of time, organizations have trained thousands of people on genAI, established guidelines for its ethical use and began launching pilots. With many pilots launched in the past six months, it is too early to see a qualitative ROI.”
Measure results. “As with any successful project, a measurable objective has to be defined at the beginning so teams can quantify the outcomes after its technical completion,” said Werner. “In predictive AI, you could measure the prediction quality. For generative AI, you could measure adoption rates, or, in the case of agents, you would look at the business KPIs you use today — when your staff is not augmented by AI — and see whether things improve.”
Re-examine data resources, data management, and data security. “Data and data governance are the biggest issues facing enterprises today” when it comes to making AI success a reality. “The use cases show promise, but large language models raise data security concerns with clients,” said Hall. “Companies are turning to solutions that integrate their data with LLMs in secure environments, such as OpenAI running in a dedicated Azure instance. This creates data and data integrity challenges to training LLMs on proper responses.”
Be open, transparent, and adaptable. Start with a clear mandate and a growth mindset that reimagines how work gets done,” said Candy. “They should center their transformation around people and skills, ensuring no one is left behind. They are model agnostic – embracing an open approach and adopting different AI models for specific use cases. They have transparency into the data used to train the models and the ability to govern and manage these LLMs across the enterprise. And they prioritize AI governance and ethics above all else – beginning at the level of concept and continuing throughout the lifecycle of the AI solution to ensure applications are trustworthy and compliant.”
