The leader's job is not to tell everyone to use AI.
That is too vague to be useful.
The leader's job is to set the standard for how AI changes the work.
Start by choosing the first workflows carefully. Good early workflows are frequent enough to matter, bounded enough to test, painful enough that people care, and low enough risk that the team can learn without turning every step into a governance debate.
Set expectations in plain language. What can people use AI for? What data is not allowed? What requires human review? What outputs can be shared? What does good look like? What does "not good enough" look like?
Create permission and pressure.
Permission without pressure creates hobby projects. People experiment when they have time, which usually means they do not. Pressure without permission creates hidden use, bad incentives, and careless shortcuts. The team needs both: room to test and a clear expectation that the work should improve.
Build internal champions, but do not confuse champions with strategy. Champions should transfer capability. They should help other people learn, publish patterns, and raise the standard. If every AI improvement depends on the same few people, the organization has created a new bottleneck.
Decide what to centralize and what to leave local. Standards, risk boundaries, shared patterns, and measurement should be centralized enough to keep the team safe and coherent. Experimentation should stay local enough that people closest to the work can move.
Turn experiments into repeatable systems. A useful experiment should end with a decision: adopt it, adapt it, or kill it. If adopted, it needs an owner, a workflow, a review path, and a way for others to learn it.
Communicate without hype or fear. People do not need another speech about how AI changes everything. They need to know what changes for their work, what stays human, what is expected, and how they will be supported.
Most importantly, make subtraction part of the plan.
If the AI-assisted workflow works, what stops?
If the answer is nothing, the strategy is not finished.