The current landscape
AI, with its capabilities for data processing, prediction, and automation, is generating undeniable excitement. Industry leaders are captivated by its potential to revolutionize sales and procurement processes by optimizing demand forecasting, automating routine tasks, enhancing customer service, and providing deep insights for more informed decision-making, to name a few applications.
A healthy skepticism
Despite the enthusiasm for AI, my position remains one of skepticism, especially regarding its short-term impact on our industry.
You have probably noticed that AI is excellent at language processing and, in particular, natural language processing (language two humans might use to interact, for example). Language processing is the foundation of modern computing: computers are, in fact, built on languages from the hardware up. While these languages aren’t spoken (though there are some great jokes in those languages that I’ll spare you), they still have all of the core characteristics of a spoken language: parts of speech, verbs, nouns, sentence structure, etc. These elements are called grammar, and all computation is built on top of grammar (special thanks to Noam Chomsky for his contributions here).
It is, then, not a surprise that combining data processing with natural language processing could produce a solid AI-like linguistic experience. But make no mistake: this is still very much a finite computer program. It is limited to only what it can be programmed to do, which is to (1) assemble data, and (2) understand or create language.
Because so much of human interaction is founded in language, the combination of these capabilities makes AI feel very real and very alive. But this AI lives in a very controlled environment. While it has access to datasets that allow it to accumulate information and assemble linguistic responses very quickly, outside of those pre-defined functions, it has the same problems as any other computational system: integrations. Ask it to unlock your front door, start your car, or even the simplest of digital tasks like sending an email on your behalf, and you’ll begin to notice very quickly how isolated and limited AI really is, even with robust and excellent integration interfaces in place.
Challenges and considerations
Having spent nearly three decades working on system integrations, I expect that AI (as a whole) will remain rooted at this spot, with nominal advances, for years to come, with data and systems integrations as the major impediment to growth. For background, think about the myriad systems you interact with on a daily basis: alarms, games, and messaging on your phone; Banking, bill-pay, travel arrangements, and shopping on your phone or computer; even entertainment on your computer, console, or TV.
At work, you interact with a quote system, a website, an ERP/MRP, a time-tracking system, a ticketing system, and hundreds of other systems that help your company run. For AI to be able to interact with these things on your behalf, it must be properly integrated in the form of an API: a digital connection between systems, and that means a software or data integration initiative. Does that sound familiar? I bet it does–it’s the root of all business computing problems, siloed systems, and the flow of data between them. Unfortunately, the data stores used by those systems (i.e., databases, EDI, XML, etc.) are not so neatly defined or nearly as finite as natural language, and integrating them will be exponentially more difficult, more specific, and more expensive than the resources expended to build AI to this point.
Beyond the hype: practical recommendations
All of that said, AI is a tool like any other. It’s not a solution to all problems, but it’s a fantastic resource that allows companies to do more with less. Given all of these factors, my advice to our customers, partners, and industry decision-makers in general is grounded in pragmatism:
Take a deep look: Don’t take my word for it, and don’t take anyone else’s word for it. Go try to use it for something real and tangible. Exercise it beyond an exchange of words, and see what it can do. My guess is that you’ll be impressed in some areas and disappointed in others, but ultimately you will get a feel for its capabilities and how it might be applied in your company.
Focus on tangible benefits: Keep the emphasis on concrete, measurable outcomes. If you are able to find a spot where generative AI can be of value (i.e. assembling marketing language or providing an excellent human-language interface to specific customer-facing functions), then approach these problems specifically. At this time, a whole-company AI approach is not necessary (or even affordable). Find small, discrete use cases and regularly evaluate impact and ROI to ensure they are delivering value.
Consider intellectual property and trade secrets: It may seem like the usage of public AI is free, but it surely is not. You are paying for your usage by means of the data you provide to the AI. Information shared via an AI prompt should be considered by your team to be fully exposed to the public, and information considered confidential by your organization should not be joined or provided (unless that AI has been properly secured and privately deployed with those terms of service in place, and if that sounds like a big project, that’s because it is.)
The lure of AI and generative AI is compelling, and its long-term impact on the electronics industry supply chain may well be substantial. However, as we navigate this journey, let’s proceed with thoughtful consideration, realistic expectations, and a commitment to strategic, informed integration. The future is promising, but the path demands careful, conscious decision-making and a drive toward the creation of value within the supply chain.
