What Companies Are Really Doing With AI

 

“Companies say artificial intelligence will fundamentally change business forever. But what does that change look like, and what are they doing to make it happen? Reporters for the WSJ Leadership Institute's CIO Journal talk it out.

 

Isabelle Bousquette: Three years into the AI boom, it seems like American corporations are more gung-ho on this than ever. CEOs constantly tell us they're "all in" on AI. But I think we're still seeing a lot of mixed signals and vagueness about how effectively they're adopting AI and where. What do you guys think?

 

Belle Lin: I think there's one version of doing AI which is like using Microsoft Copilot: purchasing the licenses and giving them to employees.

 

Tom Loftus: That's kind of a low-hanging fruit. But then, how is this eventually, as they say, going to be incorporated across all processes in the workplace?

 

Steven Rosenbush: I feel like at a high level, AI in business is ahead of where people generally think it is, but behind where they imagine it could or should be. It's sort of hard to tell. But I just got off a call with the CIO of Lockheed Martin. They are using AI in many different ways. But it's behind the scenes. And to the extent that there are gains, they aren't completely transforming the company. It's just steady usage, steady payoff.

 

Lin: What I'm hearing is that the gains are in automating workflows that were already on their way to becoming more streamlined with robotic process automation. They're able to take unstructured data or data that sits in Word documents and PDFs and emails, and then combine it with what they were doing before.

 

Bousquette: I think that's right. It is happening inside companies, but it's not that exciting. It's summarizing a lot of content. It's doing a lot of really boring research.

 

It's taking customer service calls.

 

Rosenbush: It kind of reminds me of what people always say when they experience a driverless car for the first time. It's really exciting for maybe a second or two, and then it's really boring because they get in the car and nothing happens.

 

Bousquette: What about "agentic." Are they doing that? Because I think a lot of companies will still use a chatbot and then call it an agent, and that drives me crazy.

 

Rosenbush: I agree there's a lot of exaggeration, but actual specific companies that we've reported on like Walmart and Bank of New York are using agents. But I also think that we've yet to really get to the kind of complex, multipart tasks that function or can function with a lot of autonomy.

 

Bousquette: So why haven't we seen that?

 

Lin: I think technology still has a ways to go in terms of memory. That's a really important aspect of models and agents: having the capacity to literally remember what the previous context or the previous actions it was taking.

 

Loftus: I think also, do you want to have agents be truly agentic? Even without the technical issues, it's really a big step to actually let an agent make decisions over days, weeks, months, your whole business.

 

Bousquette: That's such a good point. I was talking to Ford the other day and they were making a big deal out of keeping a human in the loop. And I was like, "Well, how is it an agent? How is it taking autonomous action?" And they said: "Well, we don't want it to take autonomous action."

 

Lin: It speaks to the question of: Do we trust agents? We know they hallucinate. We know they can make mistakes. That goes for chatbots, too. Even if it's a low-stakes customer chatbot, you don't want a situation where it promises a deal that it shouldn't have or where it starts talking about ultra right-wing propaganda, which some have been hacked to do.

 

Bousquette: Only recently in the last couple of months have I started to see companies actually leaning into AI chatbots that actually talk to customers. The technology has been there to do that for years, but I think companies were afraid. Now Ralph Lauren has one, Jo Malone London has one. Carvana has a really funny one that talks in the persona of Shaquille O'Neal.

 

Bousquette: I also wanted to ask, where do you think companies are at in terms of getting an actual return on some of these AI investments? And do they still care about getting a return? You read studies like MIT saying 95% of AI pilots are not delivering meaningful value, but weirdly CEOs seem as bullish as ever?

 

Rosenbush: I think companies do care about ROI, but I don't think that they care about it as urgently as maybe we've been trained to think. CEOs have a lot of confidence that AI is absolutely worth the investment and it will pay off. But I think they have a pretty long-term time horizon. The benefits they see are often expressed in nonfinancial terms. It's like, well, of course it's worthwhile. It's saving us time and that's really useful.

 

Lin: Some of the metrics that have come to the fore are things like "cost avoidance." When you talk about: well, we don't need to hire 100 more engineers in the future. Maybe we only need to hire five because we can use Claude Code or GitHub Copilot.

 

Bousquette: Maybe we can take this opportunity to address a common fear. Are people losing jobs to AI because of AI?

 

Loftus: Yeah, I'd say so. We've seen all these earnings reports, where companies would talk about their investments in AI and also brag about how they're keeping hiring low to zero. With the stagnation in the job market, there's a lot of factors involved. It's not just AI, obviously. But it's in the mix.

 

Lin: And that's where I hesitate a little because CEOs are saying that AI is taking jobs, but it's sort of an easy scapegoat for layoffs when times are tough and when the markets are uncertain. You can just say, "We're doing good things with AI" and then get a bump in the share price as a result.

 

Rosenbush: What I hear from analysts on Wall Street is that if you look at many big corporations, revenue per employee is trending down. I think that that's probably indicative of where hiring is headed.

 

Bousquette: So are we seeing any particular AI vendors, whether it's OpenAI, Anthropic, Google or anyone else, become real winners in the enterprise yet?

 

Rosenbush: What I hear from a lot of companies is that they're designed to be model-agnostic.

 

Lin: Right, there's a lot of remaining trauma from the days of cloud lock-in where if you were in the Google ecosystem or the Amazon ecosystem or the Microsoft ecosystem, you were stuck there. Companies have learned lessons from that era.

 

Rosenbush: So then one question I have, where do you think companies go to obtain competitive advantage?

 

Lin: People. I think it's getting your people on board.

 

Bousquette: And that's a hard code to crack. You can put out a bunch of trainings, appoint so-called champions inside the company. But it's hard to know how to really break through. I actually think the best thing companies can do is just give people the tools.

 

Loftus: You have to trust your employees. You have to provide them with a sandbox to play with it. But then how do you scale that? We've all had our own magic moment with AI. I used AI to build a running app. But have we brought it into the workplace?

 

Bousquette: Right, I think for people a lot of those magical moments are more on the consumer side. So many people are using it to plan their vacations. Shopping is going to be huge, clearly. But at work, I don't know. Why doesn't it feel as magical?

 

Rosenbush: It's probably more on the consumer side because, as consumers, no one is telling us that we can't use this or that model. They're not telling us that we can only do certain things. We can do whatever we want. And so we come up with all sorts of crazy ideas.

 

Bousquette: So is there a company you think has done AI really successfully inside? And how did they make it work?

 

Lin: One is Nvidia. They use AI in their chip design. They use AI to write code. They use AI in their customer service, for all the use cases that we've talked about prior, and probably more, because it also comes from the very top. And so you have to have a leader that is a true believer, so that sort of mentality trickles down.

 

Bousquette: I think that point really hits the nail on the head. The companies that we've seen be the most successful with AI are the companies where the CEO cares a lot, and it's not just the CTO or the CIO.

 

Rosenbush: Also one of the things that really characterizes Nvidia and its culture and its leadership is all about learning. Jensen [Huang] is incredibly curious. Everyone around him, everyone that I've spoken to, they're incredibly curious.

 

Loftus: It's in their DNA. But for 99.9% of the companies, it's not at that level, and that's where the challenge is.

 

Bousquette: It'll be really interesting to see how companies try to tackle that in 2026.” [1]

 

1. What Companies Are Really Doing With AI. Bousquette, Isabelle; Lin, Belle; Loftus, Tom; Rosenbush, Steven.  Wall Street Journal, Eastern edition; New York, N.Y.. 31 Dec 2025: B1.