We might enjoy talking about the future of the cloud, or the network, or the data center, but all three of those things are subordinate to the future of “empowerment”, the use of technology to improve business decisions and productivity. The biggest questions on tech, then, are really questions about applications, which means the software that drives business benefits and consumes tech resources. What are users thinking about that? Let’s see what they’ve said to me over the last two months.
The most important question about applications is the extent to which users expect to enhance them versus simply maintaining them. On this question we have both bad and good news. The bad news is that on the average funds for new applications or major enhancements to existing ones are very limited, accounting for only 13% of the application budget and far less than that of the IT budget. The good news is that five of every six users think that in 2024 this number will be significantly higher—18% in fact. That’s an increase of just shy of 40%.
The reasons for this sharp change are first that companies believe strongly that they need improvements in worker productivity and executive decision-making, and that they believe that fundamental economic and political factors that have impacted their businesses in 2023 will be easing through the second half and be largely gone by 1Q24. Almost all companies hold both these views, but some are still expressing a need to wait and see what develops. This group says that they’ll make their decision in the last quarter of this year, and it will then frame their budget planning for the following year.
I noted above that companies are looking at productivity and decision-making as their targets, and that’s the order of importance they’re basing their targeting on, but the sequence of effort has shifted in the last two month toward the decision-making piece. A big reason for this is the new emphasis on artificial intelligence in application and IT planning. A big question is whether the interest increase is real or just a reaction to AI hype.
One of the three AI tools that actually gets strong user validation is aimed at business intelligence and analysis, and if this tool and its applications were the stated target of AI interest, I’d be happy to say that there was a real trend developing. The problem is that while the use of AI in decision support is in fact the target, the tool that’s currently most-often used isn’t the specific target. Instead, it’s ChatGPT. The number of users who said they employed ChatGPT (or any form of generative AI) was at the level of statistical insignificance, and even the number who claimed to have actively reviewed generative AI for their analysis was less than a quarter of users, but approaching three-quarters of users said they thought generative AI would be an element in their business analysis investment in 2024. To me, this means that it’s really generative AI hype at play here.
One obvious problem here is that our collective experience with generative AI comes from what’s essentially a public-data-model application of the technology, and any business insights would almost surely have to come from an application of generative AI to a private business database. It’s not clear from public comments, or from my user contacts, just how many have tried to do this and how successful it might be. To me, based on admittedly limited experience with AI development, a narrow scope to generative AI would create something not all that different from current “inference engine” AI technology, including that package that gets good user reviews.
That could also be a factor in the prioritization of AI at decision support rather than at productivity. It’s pretty easy to see how generative AI could be used to analyze business data, but more complicated to work out how it would make specific sets of workers/jobs more productive. Still, there’s a lot of interest in AI for productivity, but the only place where that interest seems to have a focus is in the automation of operations tasks relating to software deployment and network support. In those spaces, the tools that have the most positive current user views are the tools that are also getting the nod most often for increased investment next year.
The problem with a broader targeting of productivity through AI is really the problem with targeting productivity through any means. We’ve spent decades, for example, addressing the problem of getting empowering data to workers. The data existed, but the pathways that had to carry it were missing or limited. This was one big factor behind the golden age of network investment. We’re past that now; the problem is that we don’t have the data, the application framework. A decent number of CIOs told me that the couldn’t let an RFI/RFP for productivity tools because they had no idea what to ask for other than that broad and simple category identification. If they went out with a general request, which only a few bothered to do, they got responses that weren’t helpful at all, or even insightful. I’ve blogged about this in the past in the context of a “metaverse of things” or “digital twinning”. We don’t have a complete toolkit, and we don’t even have widespread understanding of what might go into one.
I think that the uncertainty on how to approach future productivity enhancements, and the confusion over how AI could improve business decisions through analytics, is both inhibiting and perhaps driving the growth in application spending users speculate they’ll accept next year. On the “driver” side, the fact that specific tools aren’t yet presenting themselves means users are free to speculate on value using only hype as their guide. On the “inhibit” side, almost every user said that if they had a clear set of tools for both the decision support and productivity support missions, they would “increase their budget in 2024” beyond what they speculated would be an almost 40% rise. More companies than not said their application spending would double or more.
Meanwhile, operations-driven AI is quietly gaining ground too. Some companies include security in this mission, while others separate it out. If the two are separated, spending on operations AI next year would grow by almost 70% if user estimates are to be believed, and if they’re combined spending on operations AI would grow 2.2 times over 2023 levels. This suggests that security AI presents a potential growth rate of about a 50%, which lags operations AI. I think that may be because companies are having a hard time planning any major security changes because they fear it could open holes for at least the period of transition.
Both new features for security and operations AI are being adopted almost exclusively in “vendor lakes”, meaning that the new tools are adopted where they’re offered by the vendor who dominates the current equipment and platform software purchases, whether we’re talking servers and data centers or networks. One general operations tool gets a user vote of confidence, but it’s not as popular as vendor tools. But even vendor tools are usually criticized by users. Only one network AI ops tool, for example, gets a strong endorsement by its own users.
The vendor specificity of AI, at the moment, is a barrier to adoption. If there were a number of multi-vendor options available, competition among them would drive feature advances, but vendors realize that the inertia of the infrastructure they’re operationalizing would make it difficult for a user to adopt a competitive tool. To do so would in most cases also making a major change in infrastructure, and that’s not in the cards for 2023 or 2024 according to users.
There is an important overall lesson here. Benefits drive investment in technology, and incremental investment (meaning spending for something new rather than to sustain the existing tools) requires new benefits. That’s the challenge for the tech space, particularly since it’s now true that new benefits can be captured only through new applications, and those applications will then have to drive changes in cloud commitments, hosting and data center spending, and network spending on services and devices. Pushing for changes lower than applications on the value chain will likely be frustrating and unsuccessful, so platform and network vendors need to be thinking about how to accelerate the advance of applications, including the realistic application of AI, if they want their own revenues to rise.
Please take note that this blog is the first of my blogs to be posted exclusively on Andover Intel. If you are subscribed to my first blog site, blog.cimicorp.com, please resubscribe to the Andover Intel feed or you won’t get an RSS/email on future blogs. Click on the green “Follow” icon on any post and you can immediately get an RSS feed by clicking “Subscribe”, or click on “Show Original” in the upper right to enter an email address.
