Generative AI headlines are outpacing enterprise adoption

If you’re a parent, surely, you’ve experienced the feeling that your child grew even after a short trip. Well, if that child were generative AI, you’d think, judging by the headlines, that the kid grew from three years old to twenty after a day trip to Austin. With every new headline, CIOs wonder: am I moving the organization in the right direction? Am I falling behind?

Despite the buzz that generative AI has stirred in media and boardrooms, we believe the headlines are outpacing adoption in the enterprise. Here are five themes – garnered from dozens of conversations with enterprise CIOs in recent weeks – that we believe characterize the current state of generative AI conversations and activities in the typical enterprise. Likely, you’re doing better than you think. 

Caution is king.

For all of generative AI’s allure, large enterprises are taking their time, many outright banning tools like ChatGPT over concerns of accuracy, data protection, and the risk of regulatory backlash. Even among the companies permitting the tools, many are publishing stringent usage guidelines, and are proactively working with technology partners to accelerate access to enterprise-grade solutions with more robust security.  

Business fundamentals still apply.

While the backlogs of AI use cases generated in hackathons are long, the business cases for those use cases are not always compelling. Many CIOs suggest resisting the urge to invest in technology for technology’s sake. The recent gold rush is creating FOMO in board rooms and loosening the purse strings at a time when project dollars come at a premium. Be ready to ask tough questions to avoid short-sighted decisions. And remember that today’s price for enterprise licenses of, for example, OpenAI are “astronomical.” Many CIOs see this as an obstacle to scaling generative AI use cases in the short run. 

Generative AI literacy is nascent.

Many CIOs have become the de facto generative AI professor and spent ample time developing 101 materials and conducting roadshows to build awareness, explain how generative AI differs from machine learning, and discuss the inherent risks. Some organizations have welcomed professors from renowned universities to educate their leadership teams. Others have established generative AI Centers of Excellence composed of business and IT resources with the mandate to drive a common knowledge base across the organization.

Building the right mindset is key.

Key to managing expectations and garnering buy-in is assuring that generative AI, for the most part, will augment and enhance—not replace—human capabilities. In a recent article, Rajeev Ronanki, CEO of Lyric and author of bestseller You and AI, attributes the failure of a 2013 joint venture between MD Anderson and IBM Watson Health to the wrong mindset. According to Ronanki, they aspired to replace human doctors with machines in diagnosing cancer. Ultimately, they set the bar too high. Ronanki posits that the venture might have succeeded had the goal been to position AI as an “automated second opinion,” one that could, for example, support X-ray analysis and thereby help detect cancer sooner. Given how skeptical many professionals are of generative AI’s accuracy, its use may be limited to a case of “Humans AND AI” for the foreseeable future. CIOs should position it accordingly within their own organizations, acknowledging its infancy, emphasizing its potential for productivity improvements, and heading off concerns that AI will mean rapid layoffs.

Experimentation with a use case driven approach.

Exploring the art of the possible and identifying top use cases is the name of the game. For those companies that have chosen to experiment, they’re doing so largely in a controlled sandbox environment, with an emphasis on learning and understanding how to pair the technology with human capabilities to optimize for value and risk. At least for now, they seem focused on use cases that improve productivity, with compelling opportunities in the areas of sales & marketing, code generation, and document generation. 

Looking forward.

Beyond the data centers (and of course the lawyers), it’s still unclear where in the generative AI ecosystem the most value will be captured. Many hypothesize that customizing Large Language Models (LLM) for your enterprise will be futile, and that real value will come from the unique data your organization can “sprinkle” into LLMs, which will become more and more commoditized. Meanwhile, CISOs are rethinking their security posture, anticipating that bad actors will use generative AI to launch more effective phishing campaigns, among other schemes to compromise security.

Nearly all of us have heard the adage that the early bird gets the worm. Fewer have heard, or given appropriate weight to, the appended version that the second mouse gets the cheese. As exciting as generative AI is, its adoption by the business community might in time prove to be a second-mouse scenario. Surely, generative AI will exceed our imagination in time, but first, mistakes will be made. And as powerful a tool as it is, some of those mistakes may be catastrophic for the organizations that make them.

Besides, for any organization that employs AI, the technology will be useful, still, in proportion to the cleanliness and completeness of the information available to it. Perhaps, then, this is a time for meditation, a time to focus on tidying and positioning your organization for the day the clearest and most valuable use of AI presents itself to you. By that measure, you will indeed have done better than you thought.