Start with one lighthouse domain. Not fifty.
The most common mistake in AI strategy is also the most understandable one: trying to move everywhere at once.
The logic sounds reasonable. AI has broad applicability. Competitive pressure is real. The board wants to see progress across the business. So the program launches twenty use cases simultaneously, spreads the team thin across every function, and six months later has twenty mediocre results and no compelling story to tell.
The faster path. Counterintuitively. Is one domain done properly. A lighthouse.
What a lighthouse is
A lighthouse domain is a single area of the business where you go all the way. Full operating model redesign, full governance build, full measurement infrastructure, genuine accountability. Not a pilot. Not an experiment. A production implementation that changes how the work actually gets done and produces a P&L result that you can point to.
The lighthouse does three things that broad pilots cannot. It produces a real result, not a demo result. It builds organisational capability. The people who did it know how to do it again. And it creates a template that can be replicated across the business faster than the first instance took.
How to choose the right domain
The lighthouse domain is not chosen on the basis of where AI is most impressive. It is chosen on the basis of where three things align.
Value. The domain has to be connected to a material P&L line. Not a nice-to-have efficiency. A genuine lever. Revenue, cost, risk, or capacity. Where a significant change in performance would be noticed at the board level. If the value at stake is not large enough to justify the organisational disruption of a real implementation, choose a different domain.
Feasibility. The domain has to be technically achievable given your current data, systems, and AI capability. This does not mean it has to be easy. It means it has to be winnable in a twelve-to-eighteen month timeframe without building every piece of infrastructure from scratch. Choose a domain where you have the data, the access, and the technical capability to go deep. Not just wide.
Willingness. The domain has to have a leader who is genuinely willing to own the outcome. Not willing to participate in a program. Willing to change how their function operates, accept accountability for the P&L result, and tell the truth when things are not working. One genuinely committed leader in a good domain is worth more than ten slightly interested leaders in perfect domains.
What you do with the lighthouse
When the lighthouse works. When it produces a measurable result, in production, with a human accountable for it. It changes the conversation in the rest of the business. It is no longer theoretical. It is no longer a vendor claim or a conference anecdote. It is something that happened in your organisation, to a real P&L line, run by people your colleagues know.
That changes the adoption dynamic in every subsequent domain. The second implementation takes less time than the first. The third takes less than the second. The governance framework exists. The operating model patterns are established. The organisation knows how to do it.
Fifty mediocre pilots produce fifty mediocre learnings and one exhausted program team. One lighthouse done properly produces a replicable template, a proven P&L story, and an organisation that knows how to build the next one.
Pick the domain where value, feasibility, and willingness line up. Then go all the way. The breadth comes after the depth works.