Does Your Company Need a Chief AI Officer?

With the advent of the internet came the rise of the chief technology officer and chief information officer. As mobile applications proliferated in the next decade, some companies added a chief digital officer. Now, a similar phenomenon is playing out as organizations weigh whether to appoint a chief AI officer, or CAIO.

While most business leaders could do with more AI expertise on their teams, the benefit of hiring a CAIO may vary, says Birju Shah, a clinical assistant professor at Kellogg School who was the head of AI at Uber for three years.

“Not every business needs a chief AI officer,” Shah says. “A majority of businesses all the way up to the Fortune 500 need to either train or change their current executives to gain AI capability, but that doesn’t necessarily mean creating a chief AI officer position.”

Here, Shah offers some advice on determining when an organization may need a CAIO, how to structure the role, and how businesses small and large can optimize their AI investment for the long term.

How do you know if you need a CAIO?

With all the attention on artificial intelligence, the competition for top-line Chief AI Officers is fierce—and expensive. The position currently holds a median salary north of $350,000. Top companies have thrown around seven-figure signing bonuses to attract top talent. And that’s before the long-term cost of investing in the infrastructure needed to execute a cutting-edge AI strategy.

Given the potential expense of such a role, organizations should use a three-pronged threshold to determine if they need a CAIO, or if another arrangement of resources will suffice, according to Shah.

First, Shah says, the threshold for a business considering hiring a CAIO is a million customers or more. “If you have under a million-customer scale, it’s easier and cheaper just to have humans handle it. If you’re over a million-customer scale, things get more nuanced.”

Second, is your company able to offer the same product to all customers, or is it moving into personalization? If your company is betting on personalized products and services, AI will be a fundamental investment.

“Netflix, from a consumer standpoint, is the gold standard of personalization using machine learning data for a consumer’s streaming statistics,” Shah says. “This is based on user behavior—directors they like, actors they like, shows they watch repeatedly, data on when they stop or start a stream—they have been best in class at recommending shows.”

With AI, Netflix will increasingly be able to build custom and personalized content on an individual level. “This is something studios have dreamt of but couldn’t execute until now.”

Third, at a practical level, organizations need to have the resources and expertise in place to implement AI.

“This is the biggest threshold companies miss,” Shah says. “You need people that do math at your company. You need people to have a bioinformatics background, a diagnostic background. Most companies don’t have those skill sets in-house.”

For companies that meet the three-prong threshold for needing a CAIO, ensuring a long-term focus begins with finding the right person.

“A chief AI officer is quite a rare capability set,” Shah says. “They need competitive intelligence. They need to know infrastructure, network equipment, real estate. And then on top of that, they need to know AI’s problems and limitations.”

Prospective CAIOs need to be familiar with where data comes from across all business units—and be able to glean the most important data for their purposes, Shah says. This requires being networked enough internally that they can identify and execute on use cases that will help the company financially and operationally.

What should a CAIO do?

Too often at big companies, a new CAIO will bring in third-party “tiger teams” that give leaders of business lines an opportunity to use AI—but not to decide if the AI tools are better than their current processes, Shah says. He points to a recent MIT report indicating that about 95 percent of companies’ generative AI pilot programs brought little or no benefit to the bottom line.

So it’s critical to go into the process with a clear idea of how to integrate the CAIO within the company. Shah sees two common versions of this implementation.

The first version is what Shah describes as a “platform” version. Here, the CAIO works horizontally across the organization, supporting all use cases. The CAIO’s customers, in this view, are the internal executives whom they approach to learn what problems need solving at scale and not just while experimenting.

“You have to have a sound understanding of company workflows, period,” Shah says. “Those designing the strategy have to talk to everyone internally that owns these workflows. If it’s a pharma company, for example, they need to discuss with the head of R&D, the sales manager, and the people who lead interactions with doctors to unlock potential new drug discoveries and aid the adoption of those drugs.”

Shah was this type of AI leader during his time at Uber. Uber Rides, the company’s ride-hailing division, would come to him with ideas for using AI in dynamic pricing, reducing cancellations, or better serving customers. He would run trip data through AI to make incremental improvements.

“Most of what I did with Uber Rides was then copyable for the Uber Eats and Uber self-driving use cases,” Shah says.

The second type of implementation is what Shah calls a “partnership” strategy.

“In private equity, for example, the approach to turning around companies is not more labor but better partners,” Shah says.

In this setup, the organization picks the leader of an underperforming business line and makes them the CAIO. This person is given a budget to partner with AI service providers such as Microsoft, OpenAI, or Palantir to take an underperforming outcome like sales growth and supercharge it using AI tools to make sales deals close faster.

The CAIO creates a playbook that is then given to the company’s other leaders, and the position rotates every two years or so to the different business units, he says.

“In Chicago, for example, Shore Capital is working on providing a platform for its portfolio companies to come under its partnership umbrella with the outcomes they are trying to achieve,” Shah says. “Shore Capital then provides a partner and a playbook to improve that outcome, whether that is profit, revenue growth, sales growth, or cost cutting.”

How can small and medium-sized businesses implement AI?

Just because a company isn’t large enough to need a CAIO doesn’t mean it should avoid creating an AI strategy. But Shah says that the approach must be more targeted for small and medium-sized businesses as they look to capitalize on the potential of this new technology.

As AI costs rise, bigger businesses can generally negotiate with their vendors to support them at scale. Smaller businesses don’t have the same leverage, so they may need to get creative.

“Small and medium-sized businesses have to realize that they’re not going to compete on the vendor side or get a deal with OpenAI, but they can compete in execution on the customer side,” Shah says.

For these businesses, then, implementing AI could mean seeking out customers as partners in developing valuable new tools that benefit both groups.

“Where smaller businesses do so well,” he says, “is they just call a customer and ask, ‘Can we do AI with you? It may not work great right away, but can we do it together?’”

Because of the close link with the client, smaller businesses are often able to extract more-specific personalization requirements from the company and provide more value. Those smaller businesses that punch above their weight in AI may then be able to reap reputation gains by going public about it.

This edge in execution is often in highly specialized functions. For example, a human-resources outsourcing firm that supported small companies in California was able to demonstrate how it was using AI to handle legal claims more efficiently. The app the company built for this function was recently sold for 10 times more than the value of the company.

“Smaller businesses don’t have the luxury of treating AI in a general way, as big companies too often attempt to do,” he adds. “Instead, they focus on an annoying problem, and when humans can’t solve it, that’s when they try AI. So they get specialized fast.”