Most companies are slowly yet deliberately adding AI to improve business operations, building on the systems, people, and customers they already have. AI-first companies, however, rip up that script and rethink everything. It's the riskiest approach, but one with extraordinary potential.
At the dawn of the internet age, it made no sense for a video rental company to pay $50 million to acquire an online startup. People wanted their videos now, not three days later by mail. And they wanted the top titles, not some limited selection of obscure art films. The internet then was still slow, unsafe, unreliable, and plagued by copyright infringement issues.
They made the right business decision, but it cost them the company. They weren’t alone either. Many retailers, publishers, travel agencies, map companies, and video arcades closed their doors, unable to compete in the new reality.
Disruptive technologies have consequences. The new thing can be cheaper, but initially performs worse than what you currently have. You have sunk costs or fixed ones to cover, an existing customer base that likes things the old way, and a trusted brand to maintain. But the new thing gets better, especially AI. This year, for example, we’re taking a big step from AI chatbots to agentic AI systems.
Worse yet, from the standpoint of established enterprises, there are no barriers to entry. Anyone with internet access can use AI, modify open-source AI models, build on top of free tools, get a free cloud account, and deploy minimal viable products at no cost while they look for customers. The way the internet disrupted access to information, AI is disrupting access to expertise.
An AI-first company is one that’s identified a business opportunity centering on the expertise bottleneck, and creates new products and services from the ground up to replace traditional, usually human-based systems, with ones built on top of AI. These startups will first compare poorly to the existing companies they aim to replace, and many will never secure footing, which will result in a massive AI crash. But some will succeed and become the next Amazon, Google, Netflix, or Uber.
Today, we’re in a period of such a transition and CIOs have a big role to play to help guide their companies through it, sort hype from reality, and lead the reinvention of tasks, processes, and even the entire business to an AI-first approach.
First of all, AI-first
As AI democratizes access to expertise, any business process, or business model, based around the scarcity of expertise will be disrupted. That expertise could come in the form of writing software, creating marketing campaigns, understanding legal nuances, answering customer questions, identifying security threats, creating text or graphics, analyzing data — anything currently done by a human expert. It won’t necessarily happen immediately nor all at once, but it will at some point and to some degree.
“What’s unique about AI is it does something no other technology has ever done before, which is emulate thinking processes,” says Gartner analyst Mark McDonald. “If an AI is emulating thought processes, it challenges every process in the organization, from the education your people have and whom you hire, to the data you collect and the software you use. There isn’t a thing that goes unchallenged.”
An AI-first company, business unit, or business process starts with the premise that AI can handle every task requiring expertise, is organized from the ground-up around that idea, and then uses human experts where the AI technology isn’t yet ready. For example, Netflix launched as an online-only company, but used the postal service to deliver DVDs until internet technology became good enough. This is different from the AI augmented approach, where a company starts out with what it already has, and slowly replaces or augments individual steps in their processes with AI.
“The approach of becoming AI-first starting tomorrow doesn’t work if you’re an established company,” says McDonald. “So you take a more modest approach.”
Most companies go gradually, he says, not only because of the time and effort it takes to change processes, but because it can be daunting for leaders to take drastic steps. An AI-first approach is risky and can alienate existing customers, employees, and the public. But it also has dramatic potential for growth. For example, Charles Schwab had $7.6 billion in client assets under management in 1985, according to its official history. Then in 1996, it became one of the first brokerages to offer online trading and hit two million online accounts just two years later. Today, Schwab has $11 trillion in total assets under management, and, according to Investopedia, is still the largest brokerage in the US by that metric.
Brokerages trade in access to information, whether about the stock market or a customer’s stock portfolio. In the past, this information was available in person by visiting a human stockbroker at a physical office, or by mail. By virtue of these methods, traditional brokerages invested heavily in machines, human stockbrokers, and physical locations. As a discount brokerage, Schwab had less of this overhead to slow it down. But still, it still had some, so its online bet was not fully without risk.
Whether a company needs to go all-in or not is very much a top-level decision. It has to do with how much of a company’s business model is exposed to AI, how big a risk would be involved switching to AI too early, and what the potential market opportunities are.
According to a June IBM survey of nearly 3,000 executives, about a quarter of companies are already embracing what they say is an AI-first approach, and those companies attributed more than half of their revenue growth and operating margin improvements to their AI initiatives.
“This isn’t about plugging an agent into an existing process and hoping for the best,” says Francesco Brenna, VP and senior partner for AI integration services at IBM Consulting, in the report. “It means re-architecting how the process is executed, redesigning the user experience, orchestrating agents end-to-end, and integrating the right data to provide context, memory, and intelligence throughout.”
And Constellation Research’s principal analyst and founder Ray Wang says the companies that go beyond augmentation to being fully and autonomously AI-first have the potential of seeing more than $5 million in profit per employee.
Wang expects to see the first 100-person company to hit $100 billion in annual revenues within the next three years, and we’re already seeing the emergence of such companies in the software development space. According to TinyTeams, Cursor has $200 million in revenues with just 20 people, and Midjourney hit that milestone with just 10.
With that kind of growth, CIOs are integral in choosing and deploying AI tools, helping employees get the most out of them, fighting for appropriate budgets, and rethinking the entire technological backbone and business processes of the company.
At the intersection of AI and partnerships
The build-vs-buy decision is changing quickly due to AI, as is the selection of tools and platforms available. Nearly every business function can be AI enabled, and there are plenty of vendors offering to help you do that. A lot of it is hype, of course, and the person best positioned to see through the noise is the CIO.
“CIOs should be able to draw the line between what’s pure hype and what’s actually realistic,” says Mark Hughes, CTO at Insight, a global solution integrator that provides hardware, infrastructure, software, and services. AI vendors have invested huge amounts of money in AI, he adds, and are doing everything they can to convince people to sign up for what they’re selling.
“It’s important for us to be the guides as to where it makes sense to invest and where it doesn’t,” he says.
And at Booz Allen Hamilton, CTO Bill Vass says his organization is adopting AI as fast as possible. “We were early adopters of large language models,” he says. “We started about five years ago, and built a team of around 3,000 AI experts to deploy it, and we’ve embedded it in a lot of places.”
That doesn’t mean the AI always works, though, and the company has invested significantly in guardrails and failsafes. For example, in code development, AI can make a lot of mistakes and isn’t at the point where companies can vibe code entire applications and replace all their development teams.
“We’re adopting it as fast as it works,” Vass says. “It’s getting better and we’re leaning into it a lot. But like any tool, you’ve got to understand what it does and doesn’t do.”
An AI-first talent pool
It’s not enough to just throw money at some enterprise-safe AI chatbots and copilots, sign up for the new AI features in all your enterprise software, and expect a company to instantly become AI-enabled. People need to be able and willing to use all this stuff.
“The first and probably most important thing is to change the way people think about the technology,” says Hughes. At Insight, they started with a bottoms-up approach, he says, with its 16,000 global employees.
“People should know and be enthusiastic about AI and understand what it can do for them, rather than be competition for them,” he says. But that can be a challenge for some companies. Not every employee will welcome the new reality. Many will fear they’ll lose their jobs, or their jobs will become unrecognizable. The numbers can sometimes back up that unease. Consulting firm BCG released a survey in June of more than 10,000 leaders, managers, and employees and found that 41% of employees think their jobs will certainly or probably disappear in the next 10 years.
But a different perspective can reveal something very different. Take for example, tech occupations, ones expected to be among the first hit by AI. In July, the unemployment rate for tech occupations in the US was just 2.9%, down from 3.2% the same time last year. Meanwhile, the total unemployment rate for all jobs was 4.6% in July, according to the Bureau of Labor Statistics.
And back in January, the World Economic Forum predicted that 92 million jobs would be gone by 2030 because of AI, but counterclaimed that 170 million new jobs would be created due to AI.
“In order to really become AI-first, we had to make it as much a marketing exercise as a technology exercise,” says Hughes. “It’s cultural change. We’re educating and introducing people to what’s possible. We’re getting people to think about the technology as an aid to them, and that’s a pretty significant mindset change, and there’s fear that has to be overcome.”
We’re going to need a bigger budget
The transition to AI is expensive, too. The best bets are those that won’t pay off for a while, especially those involving big architectural reimaginings.
Explaining the value of the investment is something that a CIO can and should do, especially when that investment doesn’t immediately pay off.
“If you’re going to be playing in this area, the most important thing you have to be is agile,” adds Hughes. “You have to be willing to fail, cut your losses, and move on to other things. That gives you the freedom to experiment.”
In addition to making internal bets, companies can also look outside, at startups. During the dot-com era, many large enterprises invested in startups to experiment even more with new, unproven technologies and business models. That’s happening today as well, with AI-focused startups seeing investment not just from venture capital firms but also from the big, established players.
“We have a whole portfolio of startups we invest in where we see advantages occurring, and things that we couldn’t develop quickly enough,” says Booz Allen’s Vass. “We also do an awful lot of prototyping and experimentation.” But at some point, the CEO is going to show up and ask to see the results, he adds. “So we’re putting in the metrics to understand and measure the ROI,” he says.
But many organizations don’t have a good set of metrics to track productivity, even without AI, so getting hard numbers to show that AI is making an impact can be difficult.
“You manage what you measure,” says Vass. “Put measures in place to understand it so you can justify the investments to your board or your shareholders so they can see the productivity and benefit to the bottom line.”
Rebuild the foundation
By showing early results, and demonstrating AI’s value and potential, technology leaders will be able to ask for more aggressive investments — the kinds necessary to rebuild data infrastructure and technology backbones to support fully autonomous agentic AI systems and other transformative projects. That’s what it will take to truly become AI-first.
“We need a bigger wallet to do this stuff,” says Faruk Muratovic, US AI and engineering strategy and services leader at Deloitte. “We need to rearchitect the core, and not focus on step-function improvements but rather build a significantly different technological backbone for the enterprise.”
CIOs and CTOs have significant roles to play in influencing and shaping that strategy, he adds. It’s different for small companies, though, since they can move quickly, starting out with incremental improvements, efficiency gains and proven ROI. They also postpone large investments to replace legacy systems. But at some point, that strategy might stop working.
“At the point where I have to replace the legacy system or I’m dead, I have to forget about the earnings,” says Muratovic. “I’ll let the street know it’s going to take huge investments. I’ll have to take a step-function improvement and rearchitect my core enterprise, the data layer, infrastructure layer, and the applications layer, and all the agents that go on top of that.”
The companies that have the most at-risk business models may be more willing to do this, and do it earlier, saying there are CIOs taking more progressive strategies around AI.
Does AI-first make sense?
The highest probability and potential to go AI-first are the small, nimble companies that don’t have baggage, says Muratovic. For larger enterprises, he adds, “very few are brave enough and can afford to go and burn the boats in the world of AI because they’re responsible to shareholders.”
And if the adoption of the internet taught us anything, it’s that slow and steady is sometimes a perfectly fine approach. And not every company or business process that could be replaced by AI should. Or it might be too early for really big projects and large enterprises that can’t turn on a dime.
“The idea of a company on the scale of Insight and taking something like ERP or sales tools and replacing those with AI, that’s a huge undertaking,” says Hughes. “That’s years and millions of dollars. I don’t know whether the market is quite ready for that yet. You have to make some clear decisions about where you’re going to introduce AI into your technology stack and draw boundaries. That’s probably going to be the way a lot of organizations are able to move forward.”
Gartner’s McDonald agrees that most companies take an incremental approach, and it often works. “Walmart didn’t shut down its stores just because they were opening an online presence,” he says. But in other areas, there are companies evolving to be AI-first, he says, such as in the software space. In fact, in some at-risk areas, the biggest companies are the most vulnerable because of their entrenched thinking.
“An overemphasis on ROI is the kiss of death,” he says. “If you don’t leave room to think about how you’re changing your business with AI, all you’ll be keeping up is the cost savings that everyone else is delivering. You might have a cost advantage, but that’s only a six-month advantage. But everyone’s using AI to lower their costs.”
Truly AI-first companies are doing it differently, McDonald adds. “They’re challenging their model. The companies I’m talking to think less about how to save money and appease the board or CEO, and more about how to set up to take advantage of AI in their marketplace.”
There’s a practical, mathematical limit to how much companies can lower their costs, but, he adds, there’s no limit to the amount of revenue growth you can get.
