Sticker shock: Are enterprises growing disillusioned with AI?

Josh Hawley/Getty Images

Artificial intelligence (AI) has been remarkable -- at least on a smaller scale, as seen in personal assistants, robots, and mobile devices. However, the jury is still out on large enterprise projects. Executives and professionals may be waking up to the possibility that their hopes for AI may be more complicated than planned. AI technology is getting expensive, enterprises aren't prepared, and return on investment (ROI) Is still a big question mark.

Also: Organizations face mounting pressure to accelerate AI plans, despite lack of ROI 

That's the warning from David Linthicum, a highly-regarded analyst who literally wrote the book on enterprise integration, cloud, and many others. But now he isn't optimistic about the success of AI projects -- at least right now. He believes a "downturn" in enterprise AI buying is imminent, as companies realize the reality isn't meeting the hype, and descend into a trough of disillusionment. However, what will emerge after a year or two will be solid AI use cases and implementations, aligned closer to the business.

There are four reasons why enterprises are becoming disillusioned with AI, Linthicum explained:

Hitting a "data wall": The main issue enterprises are running up against is "not because the generative AI technology is bad, but because their data's bad," he explained. The challenge is "there's no easy fix for this, you're going to have to stop what you're doing, loop back, and fix your data. For many of these organizations, that particular problem hasn't been addressed for the last 20 or 30 years. [Moreover], It's a significant expense and risk, and someone has to go into the board of directors meeting and tell them we're going to spend $30 million to fix our data before we're able to get into gen AI. Those are tough conversations to have." Financial sticker shock: Building, implementing, and sustaining AI requires more resources than previous tech waves such as cloud or mobile. "These things are very expensive," he said. "They cost at least two to three times that of traditional environments, they need specialized processors like GPUs, they need a lot of resources, they need a lot of ecosystem-based components, they need the training data that's the data tuning, the model training, the model tuning, all the things that come with AI."A lack of strategic direction: "Enterprises need to get better at planning," Linthicum stated. "Not understanding the state of your data until you work on a gen AI project, [that's] not the way to do it. It's looking strategically at how your data needs to align with your utilization of this new technology."Lack of skills: AI success requires well-trained people -- "and I'm not talking about the certification training around learning one cloud provider's AI platform," Linthicum said. "I'm talking about understanding architecture, understanding data science, understanding AI ethics, understanding model tuning, understanding performance benchmarking, and understanding synthetic data." This is "very different than traditional software development."

There is no historic technology parallel to the effort necessary to support AI, "which is going to be much more complex, much more expensive," Linthicum\ detailed. 

Also: Employees are hiding their AI use from their managers. Here's why

This requires "cleaning and managing their data, getting the skills they need, doing the strategic planning, mapping out the use cases, and mapping out to the ROI." Then, "you'll get to a state where you're using AI as a strategic differentiator for your business. You're able to do something your competitors can't – providing a better customer experience, higher productivity, lower prices, and better efficiency."