Leadership in the Age of AI

There is a strange pattern in many conversations about AI.

The technology arrives first, and the problem comes second.

Organizations buy licenses and then ask employees to find use cases. Leaders hear about agents and start asking where they need an agent. A new model launches and suddenly teams are looking for reasons to use capabilities they did not know they needed a week earlier.

It is understandable. AI is moving quickly, the possibilities are genuinely impressive, and there is pressure on leaders to demonstrate that their organizations are keeping up. But starting with the technology creates a predictable problem: a lot of activity without enough clarity on what any of it is supposed to improve.

The better place to start is much less exciting.

Start with the work.

Where are people spending time they should not be spending? Where does a process repeatedly slow down? Where is information difficult to find, combine, or act on? Where does the same question keep reaching the same few experts? Where do customers experience unnecessary friction? Where are teams producing something manually that could be produced differently?

Those questions tend to lead somewhere more useful than asking, “Where can we use AI?”

They also leave open an important possibility: AI may not be the answer.

Sometimes the problem is a broken process. Sometimes the information is poorly structured. Sometimes a simple automation would do the job. Sometimes two teams are doing the same work because nobody has fixed the handoff between them. Adding AI to a bad process can make it faster without making it better.

This is one of the harder disciplines for leaders in the current environment. The technology is capable of so much that it becomes easy to confuse what is possible with what is valuable.

A compelling demonstration can make almost any use case look transformational. The real test comes later, when the tool meets the organization’s data, systems, controls, employees, customers, and existing ways of working. What looked simple in a demo may require integration, governance, behavioral change, or more human oversight than anyone expected.

That is why choosing the first use cases matters.

There is often an instinct to begin with the biggest opportunity: the most ambitious agent, the most complex process, or the use case that promises the largest theoretical return. But an organization’s first serious AI initiatives have another job to do. They need to help the organization learn how adoption actually works.

A good starting point is therefore not necessarily the largest opportunity. It is a problem that matters enough to be worth solving, is contained enough to experiment with, has the information required to work on it, and can produce an outcome that people can actually observe.

That first project teaches you far more than whether the technology works. It shows you where employees need support, which assumptions were wrong, how much oversight is necessary, where governance creates friction, how easily AI fits into existing systems, and whether people actually want to work differently once the capability exists.

Those lessons then make the second use case better.

This is where structured discovery becomes valuable.

Many organizations do not have a shortage of ideas. They have too many. The challenge is separating interesting possibilities from problems that are actually worth solving, then prioritizing them against value, feasibility, risk, available data, and organizational readiness.

That is a big part of the work we do with organizations before any build begins. Our AI Discovery engagements bring business, technology, and leadership teams together to examine how work is actually happening, identify the areas where AI could create meaningful value, and turn a long list of possibilities into a smaller number of practical opportunities.

The aim is not to produce another use-case catalog.

It is to leave the organization with a clearer answer to three questions: where should we start, why should we start there, and what would we need to learn before we scale?

That creates a much stronger foundation for pilots, training, investment, and broader transformation.

It is also why an AI strategy should not become a catalog of tools. Models will change. Vendors will change. Capabilities that feel remarkable today will become standard, while entirely new ones will appear.

The problems the organization is trying to solve are more durable.

Leadership therefore needs to keep the conversation anchored there. Instead of beginning with, “Should we use an agent?” begin with, “Why does this process take three days?” Instead of, “Which model should we buy?” ask, “What would materially improve this team’s ability to do its job?” Instead of collecting fifty possible AI use cases, identify the few problems where solving them would actually matter.

Once the problem is clear, the technology becomes much easier to evaluate.

The question is not what can we do with AI?

It is where do we have a problem worth solving, and what is the best way to solve it?