Stargate – A $500 billion joint AI venture between Oracle, OpenAI and Softbank. The aim is to construct data centers and infrastructure needed to power AI development. Unrelated to this announcement, many conversations and posts exist on gen AI, AI agents, and agentic workflows. Many SaaS organizations are launching agentic workflows within their product suite. There are many predictions on the role of agents, with some suggesting that agents will soon represent individuals in discussions, leading to scenarios such as – Have your agent talk to my agent, and we can discuss after our agents converge. AI Assistants are on every app and every platform. There are numerous models for consumption on LLMs, many open-source, and many new ones for us to consider every few weeks/days. There is so much excitement and AI in the news, on notable magazine covers and advertisements from large Tech providers.
For most organizations I engage with, however, adoption is not widespread. One sees productivity “microbursts” primarily driven by gains from specific repeatable tasks. But 10 hours per employee per month is not a game-changer for anyone. This is not to say that specific organizational roles are not impacted significantly. Meta CEO Zuckerberg told podcaster Joe Rogan about how the company was looking to replace “midlevel engineers” with AI. Organizations are also realizing that this is expensive. Are the returns commensurate with the investments across engagements? Finally, hallucinations remain a real risk. Even Apple recently rolled back its AI news aggregator.
The lack of widespread adoption at this stage raises the question: Is the adoption of generative AI in our everyday professional and personal lives inevitable? Or is this another technology that a few roles within an organization will adopt because there is a meaningful impact, but this is unlikely to be widespread? The answer is that we expect large-scale effects from this technology. Throughout history, combining different technologies has been crucial to significant advancements, and generative AI also follows this pattern.
Technological convergence refers to integrating two or more unrelated technologies. This convergence can accelerate these unrelated technologies or sometimes come together to form new ones. Some notable examples of this phenomenon include mobile phones and the internet. Mobile phones were initially designed for voice communication, but with the introduction of smartphones, they have evolved into multifunctional devices. One can call, text, browse the internet, take pictures, transact business, and more. This convergence has driven the adoption of mobile phones and has accelerated the development and use of Internet services. Another example, on a smaller scale, is smartwatches, such as the Apple Watch. You can use the smartwatch to keep track of time, track your activity, track key health parameters, communicate, and engage with various apps on a wearable device.
What are the components that support convergence for generative AI?
Cloud—The data movement to the cloud has been ongoing and evolving, strengthened by technological advancements and powerful computing capabilities. This has been happening for a while, and it is opportune that we find ourselves in a situation where most organizations are in the mature stages of their journey.
Data—We are capturing more data, which is also more readily accessible. We can consume data in many forms, structured and unstructured. In addition, the fact that one can use generative AI to engage with all this data in its current forms is an essential motivator for using this technology.
Digital literacy—Organizations have increasingly good digital literacy. Further, these foundation models have democratized AI, from data scientists sitting in an ivory tower designing algorithms that the rest of the organization does not understand to putting the power of these models in the hands of everyone. Engaging with ChatGPT does not require any special knowledge of how AI works.
Multimodal—What started as large language models (LLMs) that understand, generate, and manipulate natural language has now transformed into processing images, audio, and video. So, they are technically large multimodal models that are more natural in how we make decisions.
Based on my conversations with leaders across organizations, here is my educated guess on the impact. Generative AI will fully automate 20% of daily tasks, allowing more time for creativity by eliminating the mundane. After all, we will adopt anything convenient for us. It will further enhance efficiency and productivity in another 60% of tasks. The impact here could be wide-ranging – 25%-50%. Meanwhile, the remaining 20% of tasks will evolve and still need human oversight, blending technology with a human touch. All this to say that the question of broad adoption may be premature. We must understand that it is a journey and begin with where adoption will drive the greatest value. Writing code, for example. And as we identify other sources of value, we will be sure to design effective and economically viable solutions that work.
While challenges remain, the outlook for the future of generative AI is promising. We are poised to change how we create, communicate, and engage. However, we must balance this innovation with a responsible approach that will allow us to harness the potential of generative AI, creating efficient and fair tools for all users.
