I’m now seeing comments about the “AI bubble” and its “bursting” even beyond the tech media. We certainly had a major tech dump in stocks last week, so it’s fair to ask whether the problem is, as Axios said, that the AI bubble had burst. I guess you know by now that I’m going to say that it’s complicated. I’m going to say that we have a problem, a problem bubbles have largely created, that goes beyond stock market prices.
Tech today lives in a climate that favors, even loves, bubbles. Stocks going up is always good news. Companies are rewarded for saying things that make their stock rise, and for saying things that limit its falling. Ad sponsorship of nearly all tech news (and most other news, too) means that we tend to see or hear what those who place ads think will help them. And let’s face it, we all click on stuff that’s interesting, and the more clicks something can generate, the better coverage it will get.
Back in the 1980s, I was doing semiannual market surveys as we transitioned networking from paid subscription to ad sponsored. At the time this started, we had about 14 thousand organized points of network equipment procurement in the US, and the reader base for primary network publications was about the same. A decade later, I found we had increased the reader base by a factor of ten, and had gained less than 500 new points of purchase. If you added up the budgets that people said they were responsible for on those reader service cards, the total exceeded the total capex budget, and approached the US GDP. The point? Readership was now dominated by network amateurs. What do you think online tech news has done?
AI has been around for decades. In the 1990s, I did consulting in the field, in fact, and at that point I don’t recall much interest in the topic, even in tech publications. You could read about “knowledge engineers” and “subject matter experts”, and anyone who wasn’t pretty well-grounded in the topic had no idea what any of that meant. Today, we’ve solved that problem. We have stories about how AI is going to steal your job or make mankind extinct, and millions read them. Consumability generates clicks, not insight. Search engine optimization (SEO) targets the masses, and the masses don’t good technology project decisions. Sadly, most of what non-technical C-suite people read falls into this mass-click category.
Generative AI is consumable, at two levels. First, we can read about how it’s advancing, enhancing, and eventually coming for us. Apocalypse Now, or maybe Apocalypse Z. Second, we can play with it online for free, and get subscriptions for AI-as-a-Service. The second of these lets AI bypass the usual capital ROI controls a CFO would apply, so AI can sort of sneak into businesses. Business case, ROI, who needs it? As a manager or executive, or maybe just a senior person, I can sign for up to a hundred bucks of AI service a month, and everyone is doing it so my decision will probably never be questioned.
We’ve created a hype and bubble industry that used to be tech.
Or, well, we’ve created a hype and bubble industry on top of tech. What’s most likely to be published is what search-engine optimization determines is most likely to be read, which favors mass readers over those whose information needs drive actual tech progress. We still need all that tech stuff, the stuff that the 14 thousand professionals read about, but how they get the things they need, how they learn about the aspects of AI that aren’t sensational enough to get noticed, is all pushed into the background. It’s the tortoise in the race, with the hype hare out in front by a mile.
And that, my friends, is where we are right now with AI. The sad truth is that AI is a revolution, something that could add billions to IT spending because it could generate more billions in benefits. How? You’ll probably never know until it’s already happened.
Last week’s stock market wasn’t really AI’s fault, but AI played a role. The stock market isn’t what you think either. You probably picture it as professional investors looking for value, but the great majority of tech stock trades aren’t made by investors at all, but by traders, and the majority of those are the hedge funds you hear about. Investors can make money when their stock goes up, but traders can make money when their stock goes down, too. What they can’t do much with is stocks going nowhere. When something like AI comes along, they love the upside it generates. When AI uncertainty raises its head, they love the downside too, and trading professionals have tools that actually let them break a market. When things start looking dicey, they trigger those tools and the market dumps. A couple days later, the market goes back. Every up, every down, wins. These swings are the easiest to see, and to cause, in market areas where hype has raised expectations unrealistically, like AI.
The real value of AI, a value that companies like IBM saw from the first and that I’ve blogged on for a year or more, is that it can generate functional elements that can be introduced into business processes just like we already introduce software elements. These functional elements are what enterprises have been telling me are “AI agents”. As this piece points out, though, the concept of the agent has already been contaminated. We’ve made it into, paraphrasing the article, generative AI with an instruction manual. It’s really a software component to enhance a business process.
That’s the reason why AI isn’t all hype. What’s hype is our conception of its value, a conception created by the way we learn about technology advances. The real value of AI will take time to develop, perhaps more time because focusing on the wrong thing invariably leads to delays in recognizing the right one. Read through the article, though, and you’ll find comments from Anthropic and Salesforce that seem to straddle hype the reality, at least. A year ago, we didn’t even see that much. When reality starts to dominate, we’ll see another AI wave, and while it probably won’t be as exciting as the current one, and may not even create the same Wall Street boom, it will surely deliver new winners.
How will enterprises get to the real AI value proposition? The answer lies in two factors. First, there are some vendors who enterprises believe are delivering strategic AI advice—IBM and Oracle are the leaders in terms of enterprise comments. These vendors, where they have influence, can help management frame AI projects optimally. Second, some enterprises are creating special AI teams, combining technical people with AI skills, technical people who understand current business process and workflows, and decision-makers. It’s a bit early to call these teams a success, but the early signs are positive. Their greatest benefit may be their ability to translate AI technology into business-consumable form, something that getting AI projects approved and making them successful surely demands.
