AI has never been more accessible. The technology works, there are frameworks to follow, real-world use case examples to draw inspiration from, and solution vendors ready to help almost any organization get started. And yet most organizations are still searching for where and how to begin — overwhelmed by the abundance of options rather than held back by a lack of them.
Getting started with AI and building something lasting with it are two different challenges, and they require different things from leadership. The first demands the courage to start before conditions are perfect. The second requires discipline to build direction before momentum dissipates. Neither is primarily a technology challenge, and both come down to decisions leaders have to make before they have all the answers.
When getting started with AI, the challenge isn't a lack of options. It's the opposite.
Whenever I talk with leaders about AI, there’s no shortage of ideas: use cases suggested by teams, proposals from vendors, inspiration from conferences, examples from competitors. The possibilities seem endless. And that abundance, paradoxically, makes it harder to act. When everything could be the priority, committing to any one thing feels like giving up on the rest.
Faced with that abundance and no obvious place to begin, organizations tend to do one of two things. Some keep analyzing, hoping that more research will bring clarity. Others sidestep the difficulty entirely — the data isn't clean enough yet, an ERP project takes all the focus, the internal processes are too fragmented. All understandable responses. But both tend to push the decision forward indefinitely. The irony is that those prerequisites rarely get resolved in the abstract. A real use case, with real stakes, is usually what finally forces the data quality issue to get fixed, or makes the process question concrete enough to answer.
I'm not immune to this myself. It's easy to keep analyzing — to convince yourself that one more round of research or iteration will bring the clarity I need to commit. I've caught myself doing that. What usually snaps me out of it is watching someone else acting before they had all the answers. There's genuine inspiration in that. And a little bit of envy, too.
Starting isn't about having the right conditions. It's about making a decision to learn. It means picking something small and specific, and being honest about what you're trying to learn from it. The perfect moment doesn't come from waiting — it gets built by doing.
Getting started is one challenge. What comes after is a different one.
Early experiments are supposed to be scattered. You're testing ideas quickly, learning across a wide front, getting the organization's hands dirty with real AI work. That's healthy.
But only up to a point. When the scattering continues without a shared direction, the picture starts to look different. Teams building toward slightly different goals. A technology stack growing in multiple directions without a shared logic. Individual wins that are hard to explain to each other, let alone to leadership. Nobody can quite answer "why this, why now?" without a long backstory. That's not momentum. That's noise accumulating.
The move from experiments to direction takes discipline. Not the discipline of slowing down or writing a comprehensive vision or strategy document before starting. That would just recreate the analysis paralysis of the starting phase at a higher level. The discipline here is less visible: saying yes to this, not yet to that, and sticking to it even when something new and shiny arrives.
Direction is often formalized through tools like vision statements or strategy documents. That framing can make it feel heavier than it needs to be. What you actually need is closer to a compass than a full map.
In practice, it means having a small number of priority directions tied to real business value — specific enough that people can use them to evaluate ideas — and being equally clear about what is out of scope for now. Saying no is how you protect the space for the things you've actually chosen.
This doesn't need to be perfect on day one. The important thing is that the direction exists in the first place, and it is shared clearly enough that people can make their own decisions without escalating every choice.
Worth noting though: a good compass is relatively stable. It should evolve as you learn and as the world around you changes — but with intention, not as a reaction. The whole point is that it gives people something steady to navigate by.
When you start moving from a series of disconnected AI initiatives into building organizational capability, the questions change. It's no longer just "which project do we run next?" It becomes "what kind of organization are we building, and what does that enable?"
It's about building a repeatable, disciplined process for choosing, doing, and learning — so that each step makes the next one better informed. That's what a capability actually is: not a set of skills people need to acquire, but an organizational muscle that gets stronger with use.
Building a capability has to start somewhere. The question is whether the first experiments are treated as the beginning of something intentional, because that's when the work starts to compound.
Both challenges share something uncomfortable: neither allows you to wait for certainty before acting. The courage to start and the discipline to build direction both require making consequential decisions with incomplete information. And AI moving at its current pace has only raised the stakes on that.
The organizations I most enjoy talking to and working with aren't necessarily the ones with the most projects or the most sophisticated technology. They're the ones where leadership has a clear forward energy: curiosity, a willingness to act, and genuine excitement about where it leads. That energy is contagious and impossible to prescribe — but immediately recognizable.
Do you recognize it in yours?