Buying AI Is Not Like Buying Software
The modern CEO is an expert buyer of software. You know the drill: You identify a process that is slow, expensive, or error-prone. You seek out the best software solution and listen to all the vendor pitches. Then you sign off on the purchase. It could be a new CRM, a new ERP, or a new applicant tracking system.
Now that AI has arrived, CEOs are reaching for the same playbook. But the old method of buying software doesn’t fit this new category of technology. The CEOs who treat AI as the next generation of process automation will only capture a fraction of its value while wondering what all the fuss was about.
AI Is a Different Kind of Purchase
I suggest a different mental model for CEOs. When you deploy AI in your company, you aren’t buy a tool. You’re hiring someone.
AI has much more in common with a smart, hyper-educated employee than it has with traditional software. A CRM has a defined function that does not change unless it’s on the vendor roadmap. But AI can absorb information, reason through problems, respond to direction, produce original work, and perform virtually unlimited types of assignments.
If AI is more like a hire and less like a software purchase, that means we have to deploy it differently. AI may have the equivalent of a degree in marketing, finance, technology, law, management and every other domain, but it does not automatically know that your board is focused on gross margin this year, that a major customer is at risk, or why your company has chosen one strategic priority over another. Without that context, AI can produce plenty of work that seems impressive while missing what actually matters to the business.
AI also needs guidance about what you expect it to do. Traditional software arrives with the job already defined. The reason you bought payroll software was to run payroll. AI can research, analyze, write, plan, compare alternatives, review work, and help solve problems across almost any function. Someone has to decide where those capabilities are useful, what standards should apply, what information the AI needs, and where human judgment remains necessary.
This makes working with AI much more of a two-way relationship than the software tools we bought in the past. The important question for CEOs is therefore broader than which processes can be automated. We need to figure out how to make these new forms of intelligence productive inside the organization.
That is largely a management question. Fortunately, CEOs already know a great deal about managing intelligent resources.
Managing AI
There are three areas I think companies need to consider as they begin putting AI to serious work:
Onboard it. Every company is unique, which is why I have always believed that even experienced employees require training when they join a new organization. AI needs the same kind of business context. If you want it to contribute to important work, it needs to understand the company’s strategy, goals, customers, organization, and current priorities. This requirement may expose a problem for some CEOs: You cannot give AI a strategy that has never been clearly articulated. Important knowledge that exists only in the heads of a few senior executives will have to become more explicit if AI is going to use it.
Give it the right assignments. We are also beginning to see that AI can have different kinds of expertise, much like people do. A general-purpose model can draw on an enormous amount of information and may be useful across many subjects. Specialized systems can be built around a particular body of knowledge, methodology, or company context. In my own work on ChatCEO, for example, the goal is to give the AI a specific methodology for the CEO role rather than simply having it synthesize all the leadership advice available on the Internet. The broader point is that CEOs will need to understand what kind of intelligence they are using and where it can contribute the most value.
Manage its performance. Once AI is doing meaningful work, managers need to evaluate the results. Does it have the information it needs? Are the instructions clear? Where is the work strong or weak? What should change the next time? Good managers already do these things with people. They provide context, clarify expectations, review performance, and make adjustments that help employees become more productive over time. Companies will need to develop the same discipline around AI.
The Meaning of Management Is Expanding
I have long argued that a company’s management competency can become one of its greatest competitive advantages. At scale, the quality of an organization depends heavily on the ability of its managers to take talented people, align them around a strategy, provide the proper resources, and turn their efforts into results.
AI does not change that principle. It expands the set of resources management is responsible for.
Managers will increasingly have both human and AI resources available to them. They will have to determine which work belongs with people, which work can be handled by AI, and how the two work together. They will also have to make sure that AI has the same things any productive resource needs: clear objectives, the right information, defined responsibilities, and accountability for results.
I expect companies to vary widely in how well they do this. We already see enormous differences between organizations that have access to similar pools of human talent. Some consistently turn that talent into exceptional performance and others do not. AI will create another opportunity for strong management systems to separate one company from another.
The technology itself will continue to improve and become broadly available. The management challenge is making sure your organization improves along with it—and handles it differently from software rollouts of the past.