Customers Are Getting More AI Than They Can Put To Work, Survey Finds
The next wave of enterprise AI value may depend less on what vendors can build and more on how closely they build it with customers.gettyDemand for enterprise AI continues to climb and accelerate. In response, software providers are stepping up, shipping more features at a faster pace. This dynamic suggests a win-win scenario; one where customers are getting new AI-based capabilities and vendors are maintaining and strengthening their product relevancy. However, the results of a recent survey paint a less obvious picture. While software companies are delivering code at a frenetic clip, it turns out that for 51 percent of them, fewer than one in four customers are using what’s been built.The problem doesn’t appear to be demand. It’s increasingly a problem of readiness, alignment, and the ability to translate AI into measurable business value. This finding provides valuable and urgent lessons for enterprise leaders and software providers alike.Enterprises Want AI, But Not Just Any AIMax Risen is the president of M&A at Banyan Software, a company that owns and manages over 120 vertical market software (VMS) companies. These are businesses that provide highly specialized feature sets for specific industries such as software to help run a health clinic on a college campus. Max Risen, President, M&A at Banyan SoftwareBanyan SoftwareTo serve his portfolio of VMS operators, Risen developed a survey to understand how the marketplace was responding to their approach to AI feature development and deployment. He knew VMS businesses were not immune to the outsized momentum of AI but wanted to derive any unique insights when compared with what was being experienced by the well-known, large enterprise software companies. More than 260 VMS companies responded and the results were surprising. A core finding was that 51 percent of software providers reported that fewer than one in four customers were using the AI capabilities they had built.VMS operators were creating and deploying code quickly and customers were expressing their high interest in AI use but there was a notable misalignment.Building AI Faster Doesn’t Mean Customers Will Use ItOne characteristic of the emergent intelligence age is that software can be developed, tested, and deployed at a far higher velocity than in the past. That’s largely a good news story given a history of enterprise software that has moved slower than market demand. One consequence of faster software deployment is that customers get a lot more features quickly, but they may not be what end users want. Risen’s survey validates this.A larger volume of smart software might sound like a gift to the enterprise, but in the short term this is presenting more challenges than it may intuitively seem.Enterprise customers want AI, but they are still hesitant to adopt AI quickly for change management reasons that include concerns about integration with existing workflows, security and privacy risks, and the nondeterministic nature of many generative AI systems. Most importantly, while businesses want AI capability, many don’t yet know how best to think about using and getting value from it. After all, large-scale adoption in the enterprise, particularly with VMS solutions, is fairly recent. Using More AI Doesn’t Mean Creating More ValueBen Schein is Chief AI and Analytics Officer at Domo, a leading AI-based business intelligence and analytics provider. He has a front seat on where enterprises are with AI and how they’re using it. Ben Schein, Chief AI and Analytics Officer at DomoDOMOHe concurs with Risen’s survey findings that AI being delivered today is often exceeding the rate of adoption and value generation. In his words, capability is easy, deployment is hard, but outcomes are the hardest. Schein says AI demonstrates what is technically possible but translating that capability into measurable business outcomes is another matter altogether.According to Schein, too many companies are experiencing what is called tokenmaxxing: employees are consuming large volumes of AI tokens, giving leaders the impression that adoption is strong, without corresponding evidence that the activity is producing measurable business value.He recommends that leaders get a handle on what AI is currently being used and how. First, to know what people believe are useful tools and second, to fully understand the costs of AI. An inventory of AI tools may quickly reveal that costs are far exceeding meaningful productivity.Unsurprisingly given his role, Schein acknowledges that many enterprises need to get their data house in order if they’re going to get trusted results from AI. Data governance must become a priority so that the data feeding AI systems is accessible, high quality, appropriately governed, and secure.AI Vendors Need To Get Closer To Their CustomersFor many executives today, they’re thinking, for example, that AI should be deployed against an existing workflow to make it faster and easier. That will sometimes be true, but it’s a narrow way to think about the transformative properties of AI. Risen says education is required here since thinking this way is limiting. He says that enterprise leaders need to think about how AI can reinvent a process entirely, not just enhance it. Risen wants leaders to focus on optimal outcomes. That’s closer to the promise of AI in the enterprise.Schein offers one caution. He suggests that just because AI can do something doesn’t mean it should. Careful consideration must be given to whether or not to pursue a particular function for ethical reasons or because an easier, non-AI route may be better. It’s a healthy reminder.For VMS operators, Risen suggests that the provider and the customer need to work more closely together in determining new AI abilities. While VMS businesses have always tried to understand what users want through observations of behaviors across their customer base, this same approach may not necessarily always work in the age of AI. Risen encourages fast development and deployment, weeks not months, to see what’s working or not, and make decisions on what features to pursue quickly.By working much more closely with customers, VMS companies can also assist with change management, helping ensure that AI capabilities can integrate with existing workflows and other concerns such as security are addressed promptly.How Leaders Can Close The AI Adoption GapRisen provides the following recommendations for leaders trying to figure out their AI path forward. If there’s been hesitancy to get started with an AI roadmap, it’s now time to move forward. The lesson, however, isn't simply moving faster. It's to move sooner while measuring what happens, so that investment can quickly shift toward the capabilities producing results.Don't assume that because AI software is becoming easier to build, the enterprise should build it itself. There will certainly be cases where developing a proprietary solution makes sense, particularly when the required capability is highly specialized or unavailable in the market. But Risen argues that building should often be the exception rather than the default.Beyond the survey findings, Risen believes the CEO needs to be an AI champion. If the top leader isn't bought in, he argues that the company's AI strategy may struggle to thrive.Deeper collaboration between enterprises and software providers should result in AI capabilities that users actually want, adopt, and ultimately translate into business outcomes. The finding that three in four users aren’t using the features being shipped isn’t that enterprises don’t want AI. It’s that shipping AI and creating value from it are two very different things. To paraphrase Schein, companies shouldn't be celebrating usage. They should be celebrating results.
