Over the past few years, I’ve been asked one question more than any other:
“What can we do with AI?”
It’s an understandable question. AI is powerful, visible, and evolving quickly. But it’s also a misleading starting point. Like a hammer looking for a nail, it tends to create activity without direction and solutions without clearly defined problems.
I’ve seen this dynamic before.
A Familiar Patterns from Earlier Transformations
When organizations introduced Agile ways of working, the ones that succeeded didn’t start with frameworks, ceremonies, or tooling. They started by getting clear on what they actually wanted to improve.
- Where does work slow down?
- Where do handovers break?
- Where do decisions get delayed or diluted?
Agile created value when it stayed close to real work and real teams. It failed when it became an abstract operating model rolled out top-down.
AI adoption follows the same pattern.
Start With Better Questions
Instead of starting with AI, I’ve consistently found it more effective to start with three grounded questions:
- Is there actually a problem to be solved?
Not a hypothetical opportunity, but a real source of friction, inefficiency, or poor decision-making. - Is it shared across the organization?
Or is it a local irritation that doesn’t justify broader attention or investment? - Is it worth the effort and investment to solve it?
Including the cost of change, not just the technology itself.
These questions are not new. I’ve worked through them with teams and leaders throughout my career, long before AI entered the conversation. What has changed is how often the answer to the third question has shifted.
What AI Has Actually Changed
For a long time, many problems were well understood but simply not worth tackling. They were too manual, too fragmented, too slow, or too expensive to address with traditional approaches.
AI has changed that feasibility threshold.
Not because it magically solves everything, but because it allows certain types of improvements that were previously impractical. It can reduce cognitive load, support better decisions, and automate coordination-heavy steps that once required disproportionate effort.
This is why I often find myself revisiting existing value chains and processes instead of searching for “AI use cases.” The opportunity is rarely about inventing new work. It’s about improving work we already know is broken, but couldn’t reasonably fix before.
Why Experimentation Still Matters
Experimentation is essential, but not as an end in itself.
AI is still too broad and too new to be meaningfully captured upfront in rigid templates or best-practice frameworks. Experimentation allows organizations to learn where AI genuinely helps, where it doesn’t, and where it introduces new constraints or risks.
Strategy should follow those learnings, not precede them.
When experimentation is grounded in real problems and real roles, alignment and adoption don’t need to be forced. They emerge naturally.
A Tool Worth Using Well
AI isn’t a silver bullet, and it doesn’t replace leadership, judgment, or clarity of intent.
It is a tool.
But unlike many others we’ve seen over the years, it’s a genuinely powerful one. Used deliberately, grounded in real problems, and introduced close to the people doing the work, it meaningfully expands what’s possible.
That’s where AI starts to deliver real value.
