Can You Afford to be a Fast Follower in AI?

With the viral popularity of ChatGPT, AI is now recognized as the next great disruptor. A technology that will rival the Internet and the Smart Phone in the way it will change our lives. It was not surprising then, for me to find that it was one of the hot topics and this year’s National Association of Board Directors (NACD) Summit in DC last week. AI offers huge opportunities and threats for all companies and individuals and is a topic that corporate directors need to understand.

?At one of the panel sessions the question of adoption cadence was raised. One of the panelists was asked if it made sense to be a fast follower. The answer from the experts seemed to be a resounding “No.”? And who could blame them? The opportunities of AI seem to be self-evident. Every company, every function, could see game changing benefits. According to the Artificial Intelligence Index Report 2023” from Stanford University , “Organizations that have adopted AI report realizing meaningful cost decreases and revenue increases.”

?However, I felt pretty uncomfortable with this conclusion. As we start upwards on the first stage of what is already a stratospheric Hype Cycle for AI, we need to be careful to ensure that we are not caught up in unrealizable expectations. Despite its 50-year history, the large language models (LLMs) that are pushing the boundaries today are still bleeding edge. More work needs to be done to productize these innovations to reduce both the cost and the risks associated with adoption.

?AI is just a new technology and the same approach that we use when thinking about any other build-buy decisions apply here too.? The model I use to think about this may be familiar to you. I’m not sure where it’s from originally, but I have used it for years. I call it the Technology Adoption Bullseye.

?NETGEAR, of course, offers some very innovative technologies, our groundbreaking new Orbi? 970 WiFi 7 Mesh System for example, delivering unparalleled performance and speed,, or our Nighthawk? M6 Pro 5G Mobile Hotspot that provides unbeatable, secure connectivity on the go. But our core offerings today would not trivially benefit from AI. On the other hand, there are many straightforward opportunities for AI adoption within the business, from customer support, through supply chain optimization and even software development. So how do we decide where to start?

?It's important to look for low-hanging fruit. Rather than trying to build our own AI models from scratch, we are partnering with large technology providers who have demonstrated successes and existing models they can reuse and fine-tune. And we are looking for packaged solutions, like Microsoft’s co-pilot, a kind of auto-complete on steroids for software developers, that provide a low cost and low risk way to harness the power of AI.

?If you do have areas where it makes sense to be developing your own models from scratch, you should start by asking yourself if you have enough data. AI models require lots of data. The more data, the better the model. The data also need to be clean, sometimes 50-80% of the work to create a model can involve data cleaning. A key benefit of partnering or buying asolution is that you may get to leverage existing models and data sources owned by the partner.

?At the end of the day, it’s important to take a practical look at your business and try to decide where AI can help you most. If you are in a business where your core customer offerings can be accelerated with AI, then you may need to invest at the bleeding edge. On the other hand, if your opportunities are more around internal business functions then even though there may be huge benefits, it may well make sense to partner with experts who have existing proven models or to buy off the shelf technologies. And remember that just because something can be done does not mean it should be done. Always evaluate the cost-benefits carefully.

Disclaimer: The views expressed in this post are my own

Martin Westhead is CTO of Software at NETGEAR and holds a PhD in AI from the University of Edinburgh.

?