Rakesh Malhotra February 23, 2023
The technology industry often hypes new innovations only to see them fall flat when in the hands of actual users. This can create hype “fatigue,” and rightly so. Since the iPhone in 2007, you’d be hard-pressed to identify the introduction of a genuinely transformative new technology. Even in the case of the iPhone, many observers at the time dismissed its chances of market success, while its proponents largely underestimated the wave of innovation it would unleash.
Thanks in large part to the release of ChatGPT, the world is now captivated by the potential of artificial intelligence (AI). While some may dismiss this as another technology hype cycle, it is much more likely that we are actually underestimating the profound impact this technology will have on life as we know it. It’s both fascinating and impossible to wrap your head around what happens to humankind in a world where technological and scientific advancement might be sped up by orders of magnitude over a short period of time. You probably didn’t come to this blog post for a philosophical view of AI, so I’ll focus on the impact we see for our business and public sector clients at Nuvalence.
At Nuvalence, we build mission-critical software for the world’s most ambitious organizations. This is often done in the context of “digital transformation” as we help our clients build new revenue streams and differentiate with software. What we are telling all of our clients today is quite simple: “If your digital transformation strategy does not deeply consider the impact of AI, it’s dead on arrival.” It would be like building a software strategy in 1998 and not considering the impact of the Internet. Traditionally, software companies would be the first to adopt new technologies (cloud, SaaS, mobile, etc.), while other businesses followed slowly behind. If you believe, as we do, that all companies need to become software companies to compete, then traditional enterprises cannot afford to wait. By the time the executive memo comes out, it will be too late.
To help get the conversation started, I’ll outline three principles that business and technology leaders need to internalize and incorporate as part of any digital transformation strategy.
1. AI is Not a Feature
It’s tempting to look at AI as an enhancement to an existing product or use case. Tactically speaking, this is probably true. For instance, at Nuvalence, we’re helping our clients use state-of-the-art AI technologies to detect fraud, run call centers, validate identity, or create chatbots.
However, in the early days of the Internet, it would have been a mistake to think of the Internet as a feature or extension of your existing business — for example, brick-and-mortar delivery, a distribution mechanism for your newspaper, or a way to buy DVDs online. The winning strategy was to completely re-imagine how your business model and products worked in a world where internet access and information delivery were nearly free and ubiquitous. The same will be true for AI. What does it mean for your business in a world where access to cognitive human labor is nearly free and ubiquitous?
2. AI Requires a Platform Approach
One thing that enables ChatGPT to do what it does so magically is that it has been trained by large volumes of data at internet scale. Intellectual property considerations aside, it’s clear that making it easy to access this training data is essential to its effectiveness.
Compare this to a traditional enterprise, where knowledge workers can’t easily locate the latest sales data from within their own line of business. This is obviously not because search is an unsolved problem. Breaking down data barriers between organizational silos is challenging, but AI success is highly dependent on it. From a technical perspective, a platform approach is the only way to solve this problem. Central to the effectiveness of Large Language Model (LLM) based AI systems (on top of which ChatGPT is built) is the size and quality of the dataset used to train it. Businesses and public sector organizations will want to fine-tune models to adapt to domain-specific use cases important to them. This might include health records, vehicle telemetry, or transaction history at the DMV. None of this works if the data within an organization is not easily accessible.
What we are telling all of our clients today is quite simple: “If your digital transformation strategy does not deeply consider the impact of AI, it’s dead on arrival.”
A platform approach prescribes that for all members in your LLM platform’s multi-sided ecosystem they:
(A) Each can extend the platform and/or the underlying LLM (for example, by injecting structured data sources as “knowledge” into the model or fine-tuning the model itself)
(B) Have access to shareable, discrete unit-level LLM outputs via APIs.
As a practical matter, adhering to these two prescriptions guarantees that silos can’t exist (with the assumption, of course, that the LLM turns highly unstructured bits of contributed data into useful knowledge). Why can we make such a guarantee? Because (A) requires members to contribute their data or expertise to align the LLM with the knowledge that the LLM needs in order to make the LLMs underlying knowledge architecture useful to themselves (and as a consequence, to others) and (B) makes that knowledge accessible to themselves, and as a consequence, democratizes access equally to all members.
3. Play Offense and Play it Quickly
When a new and disruptive technology emerges, organizations usually start by considering how it might negatively impact their current business model and work to defend it. This likely played some role in Google playing catchup in this space despite originating much of the technology, though it would be foolish to count them out. There are many other great examples of the innovator’s dilemma in recent history, so this is not new.
What’s new with AI is the speed with which change and adoption are happening and will continue to happen. Unless you are Google, a “fast follower” strategy is likely a losing one. ChatGPT reached 100 million users within its first two months of release. Notably, this was done without marketing, a wonky name, and some significant limitations. Compare that to TikTok and Instagram, which took nine months and 2.5 years, respectively, to reach this milestone. We must also remember that ChatGPT is only the first platform, and it’s unclear if or how big of a moat it really has. Much more innovation is coming (and coming quickly). Organizations that mobilize immediately, even to simply experiment, will be in a much better position to compete.
What Lies Ahead?
As with most new and transformational technologies, it is nearly impossible to predict where this wave of AI innovation will lead. This is no reason for organizations to postpone considering the possibilities and getting to work now.
In 2003, Ray Kurzweil wrote:
“We’re entering an age of acceleration. The models underlying society at every level, which are largely based on a linear model of change, are going to have to be redefined. Because of the explosive power of exponential growth, the 21st century will be equivalent to 20,000 years of progress at today’s rate of progress; organizations have to be able to redefine themselves at a faster and faster pace.”
Comparing the coming AI era with those that came before it will be tempting. Even I have done so in this post to help define precedent. However, the AI era will be, in retrospect, unprecedented. Fear of disruption can be a powerful motivator, but truly ambitious organizations will view this as a generational opportunity to redefine their industries.