Over the past decade, I’ve seen user behaviors and emerging technology adoption conform pretty closely to the Gartner® Hype Cycle™. Innovation triggers the rise to the Peak of Inflated Expectations. That’s followed by the slide into the Trough of Disillusionment and the eventual gentle climb up the Slope of Enlightenment to the Plateau of Productivity, where tech becomes mainstream and delivers value.
Although some people look down on the “hype” and inevitable impatience that surrounds new ideas and innovation, it’s always seemed like a natural part of the process and pretty predictable to me. Excitement around the “what if” and the inclination to dream about the possibilities will inevitably move faster than even the most well-planned tech solution go-to-market strategy.
However, generative AI seems to be drawing a curve all its own, looking more like a cardiac sinus rhythm than a usual Hype Cycle.
Expectations have risen sharply since ChatGPT launched a year ago, with 70 percent of organizations exploring how this technology could benefit them, and 45 percent increasing their investment in AI. Gartner analysts now see generative AI at the Peak of Inflated Expectations.
Buckle Up for the Plunge
So, now teetering at the top of the curve, the infatuation with generative AI is about to end. Of course, there’s no denying the potential this technology has. Generative AI, as the name suggests, generates text, code, images, video – or whatever other types of content your team can imagine. And it does it in a fraction of the time it would take a human to produce it. Additionally, large language models (LLMs) can give a generative AI platform the ability to understand users when they make requests or ask questions in natural language. As a result, accessing information and increasing productivity is easier, even for users who aren’t tech-savvy.
However, challenges with generative AI are revealing themselves as well. Some of the hurdles you’ll have to cross for this technology to deliver healthy ROI and, in some cases, more good than harm, include:
-
-
- Data governance
-
You’ll have to balance making information available to the platform to generate valuable content while retaining control over which users can see various types of data. You need to find a way to control access to personally identifiable information, intellectual property (IP), and other sensitive data, particularly if your organization or your clients must comply with data protection regulations.
-
-
- Legal issues
-
Deloitte points out that training data could inadvertently include data protected by copyright or other IP rights. Plaintiffs have already filed lawsuits regarding attribution for the information a platform uses. Conversely, there is some question over whether the output of a generative AI system could be copyrighted, which could create some risk for companies.
-
-
- Domain- or use-case specificity
-
Another consideration is what the platform is trained to do. An AI developer I’ve spoken to has, on several occasions, stressed that AI solutions work best when they’re trained for a specific purpose. Platforms aimed at horizontal vs. vertical markets take longer to implement. Furthermore, the answers they provide initially may not be precise. So, although the tool may learn with use, users may lose their enthusiasm if underwhelmed with the content it generates out of the box, and adoption may suffer.
-
-
- A viable business case
-
Nothing will lead to disillusionment faster than a CFO who points out that the numbers don’t add up. While generative AI could perform a wide range of tasks, you need to ask whether it should. A knowledgeable team member must review the content that a platform generates to ensure it meets standards, for example, for accuracy, relevance, compliance, or brand. In many cases, using generative AI makes sense, for instance, in research using vast quantities of data or summarizing long reference documents. But if time savings are overshadowed by quality control – after an expensive implementation—it may not be the right use case.
-
-
- Expertise
-
A McKinsey survey found that there aren’t enough skilled IT professionals to meet the demand for AI projects. To reach the Plateau of Productivity, the industry must find ways to develop AI talent and build expertise.
-
-
- Bias
-
A generative AI tool may do a great job of allowing users to interact conversationally. It may even appear to have a personality. But it only “knows” what it has been trained to know. If training data is biased, for example, because it isn’t balanced across demographics, global regions, or other factors, its outputs will be skewed in a similar way.
The Solution to Generative AI Challenges: Actual Intelligence
For the past year, the industry has focused on the pros of generative AI, building momentum to reach the Peak of Inflated Expectations. With great enthusiasm, people have shared their vision for what this technology can accomplish and all the good it can do.
But when the hype dies down, it will be time to create marketable solutions that solve specific business problems and for innovators to find ways to overcome challenges related to generative AI. It will take your insight and creativity – your intelligence – to bring solutions to market that decrease risks and maximize value.
The industry is waiting to see just how smart you are.
