What does ChatGPT, the “fastest-growing consumer application in history,” mean for the future of work? More broadly, will generative AI soon graduate from the being the latest consumer entertainment to become a significant business application and a new basis for competitive advantage? And are enterprises ready for AI, any type of AI?
Digital transformation concept.
Forrester just published a report on generative AI that tells its enterprise clients not to ignore or downplay its impact.Enterprises should start right now to experiment with generative AI, recommends Forrester, focusing on existing processes that can be enhanced by technologies “that leverage massive corpuses of data, including large language models, to generate new content (e.g., text, video, images, audio, code).”
It would be “a costly mistake,” says Forrester, to ignore the potential of generative AI to enable production of content at scale, to accelerate the speed and precision of data science practices and app development, to produce synthetic data for training AI and machine learning models, and to provide new defense opportunities for security professionals. In short, generative AI presents an opportunity to augment and even automate existing work processes in IT, marketing, customer service and other business functions.
ChatGPT was made publicly available on November 30, 2022, and given the attention that PR stunt has received, it is safe to label the era before that date as B.G. (Before Generative AI). We are now living in the new and exciting and frightening G.A (generative AI) era, where executive FOMO may lead to embarrassing public failures (as in Google, that started paying attention to G.A. already in 2017, losing $100 billion in market value in one day).
Are enterprises ready for the new era, for the pressures to do something about generative AI, even just careful experimentation, as Forrester recommends?
We can get a sense of the state of AI in the enterprise by looking at recent surveys of business and IT executives, reporting about their current experiences with AI. The surveys—by Deloitte, cnvrg.io, Run:ai, and LXT—have been conducted over the six months just before the arrival of the G.A. era so they reflect what the respondents knew about “generic AI,” not necessarily generative AI.
Perceptions of AI are certainly positive in the business world. 94% (Deloitte) say that AI is critical to success over the next five years and 89% (cnvrg.io) are seeing the benefits of their AI solutions. At 48% of organizations, “AI is in production, or already part of the business DNA” (LXT). 91% of companies are planning to increase their GPU capacity or other AI infrastructure by an average of 23% in the next 12 months per the Run:ai survey, which concludes that “despite the uncertain economic climate, companies are still investing in AI due to the potential and value they see in it.”
According to the Deloitte survey, 79% say they have fully deployed three or more AI applications, up from 62% the year before, with top applications being cloud pricing optimization, voice assistants, chatbots and conversational AI, predictive maintenance, and uptime/reliability optimization. LXT found that Natural Language Processing (NLP) and speech/voice recognition solutions are the most highly deployed AI applications, followed by predictive analytics and conversational AI.
But challenges abound. Just 37% (Run:ai) of AI models make it into production and 46% (LXT) of all AI projects fail to reach their goals. Deloitte found 29% increase from the year before in the number of respondents self-identifying as “underachievers,” and the top challenges associated with scaling were managing AI-related risk (50%), lack of executive commitment (50%), lack of maintenance and post launch support (50%). 57% (cnvrg.io) reported low AI maturity with less than 4 models running in production and only 28% (Run:ai) reported having timely and sufficient access to computer power upon demand.
Challenges abound with deploying AI in general but when it comes to generative AI, businesses face a “labyrinth of problems,” according to Forrester: Generating coherent nonsense; recreating biases; vulnerability to new security challenges and attacks; trust, reliability, copyright and intellectual property issues. “Any fair discussion of the value of adopting generative AI,” says Forrester, “must acknowledge its considerable costs. Training and re-training models takes time and money, and the GPUs required to run these workloads remain expensive.”
So what’s a business executive to do? What is the right response to the pressures of “missing out on the new new thing could be a very costly mistake”?
As is always the case with the latest and greatest enterprise technologies, tools and techniques, the answer to “what’s to be done?” boils down to one word: Learn. Study what your peers have been doing in recent years with generic AI. A good starting point is the just-published All-in On AI: How Smart Companies Win Big with Artificial Intelligence. Tom Davenport and Nitin Mittal profile the companies (outside of Silicon Valley) that are “making big and intelligent bets that this technology will lead to major business improvements, and they already have evidence that these bets are paying off.”
Another type of learning is carefully examining the landscape of what’s on offer (Sequoia Capital counts 109 generative AI startups and CB insights lists 250 in 45 categories). Just like the hundreds, maybe thousands of startup that added “AI” to their profile over the last decade, a safe bet is that by the end of this year many more will claim “generative AI” as their bread and butter. What’s important is their proven expertise in what matters to your company and your customers.
The most relevant startup for you may not even claim the mantle of “generative AI” but has been demonstrating in recent years its benefits and what it can do for your business. An example is Anyword, a startup that predicts the audience your content (e.g., advertising copy) will resonate with and how well it will perform. It provides a predictive performance score based on its analysis of millions of copy pieces in a way that connects the conversion rate, profile of the audience, and the style and content of the message. It has been doing it for publishers since 2013 and, since 2021, for any marketer.
Most important, keep in mind that there’s no magic involved, and that the men and women behind the curtain have been steadily advancing the state of “machine intelligence” ever since the very first computers were called “giant brains” seventy-five years ago. “AI” is just another step in the evolution of modern computing and the continuation of by now familiar data-driven computer applications, i.e. machine learning and predictive analytics. “Generative AI” is just another step in the evolution of modern AI, i.e., deep learning or statistical analysis of very large volumes of data.