Financing the AI Revolution Recap: Adopting AI in the Enterprise
LLMs are improving rapidly, with new capabilities and record-setting scores on reasoning and coding benchmarks seemingly announced each week. And yet there remains a significant adoption gap, with organizations of all sizes still in the early days of translating these improvements into measurable business impacts.
During a panel discussion at The Information’s Financing the AI Revolution event at the NYSE, reporter Stephanie Palazzolo discussed how companies are adopting AI with two leaders in the field:
- Dave Rosenberg, SVP, product marketing, Oracle
- Matt Fitzpatrick, CEO, Invisible Technologies
Challenges to Adoption
While some startups have used AI to fuel explosive growth, Rosenberg said, existing enterprises are having difficulty integrating their legacy tools with generative AI platforms. “It’s not easy to be AI native yet,” he said. “We can all be AI enabled, but getting to AI native is a different ballgame.”
One of the most significant hurdles to adoption, Rosenberg said, is data. He noted that organizations must not only feed massive amounts of information into AI models, but also design systems that do not use their proprietary data to train public LLMs. “Without good data, AI becomes basically useless,” he said.
This need for accuracy, Fitzpatrick said, has kept AI tools from following the off-the-shelf, plug-and-play model of past SaaS successes. “You need that model to be 99% accurate—not 80% accurate,” he said. “That paradigm shift has been harder than people expected.”
Early AI Wins
Rosenberg noted that organizations are using AI tools embedded within Oracle NetSuite to provide advice and assistance to employees by generating content and providing instant, context-rich answers to complex questions. Over time, he said, these early wins will be augmented by AI agents that can tackle tasks autonomously. “You have areas where you wanted just some advice: ‘What should I be looking at? Maybe this is an anomaly, this may be important,’” he said. “Then you have the assistant: ‘Help me do something.’ Those features are where we’ll see a lot of the first enterprise apps. Then we’ll start to think about agents: ‘Now do it for me.’”
Looking ahead, Fitzpatrick said, many of the most valuable AI use cases will be highly specialized rather than broadly applicable across industries. “A lot of applications are going to be what I’d call long-tail use cases,” he said.
Invest in AI or ‘Fall Behind’
Fitzpatrick and Rosenberg noted that the cost of LLMs continues to drop, and many enterprises are setting aside tens of millions of dollars per year for dedicated AI budgets. “Over the last couple of years, these budgets have kind of appeared out of thin air, which reflects a belief that if you’re a public company, you need to invest in AI or you’ll fall behind,” Fitzpatrick said.
Rosenberg stressed that cloud business systems, such as cloud ERP, can offer organizations an opportunity to invest in AI without incurring large capital costs. “With the rise of the cloud, you don’t have to buy all this hardware anymore,” he said. “You can free up tens to hundreds of millions of dollars of maintenance. At some point, you start to say, ‘Why would I want to run this myself?’ If you want to be in the game, you need to move to a system that allows you to do that, which is going to be in the cloud.”
The Role of Human Workers
Both Fitzpatrick and Rosenberg stressed that they largely see AI tools as augmenting human employees rather than replacing them. “The magic is, if you’re just a user, you have these advantages now, and that really goes a long way,” Rosenberg said. “Now we have a service that can help people progress with what they’re doing.”
“I think humans are more important than ever,” Fitzpatrick said. “Our whole company’s belief is that humans in the loop are actually the way you benefit from AI. We’re hiring more and more people.…Our view is that [with AI] you will be able to get a lot more velocity from hiring great people.”
While AI tools are obliterating one coding benchmark after another, Fitzpatrick added, they still struggle with independent software development. “Without humans, it’s actually pretty hard to build anything,” he said. “You need great engineers to accelerate with AI. You can’t just rely on the technology on its own.”
