Governance
Data governance
Knowing what data you hold, who may see it, and where it is allowed to go — the prerequisite most AI projects skip.
AI makes existing data problems visible. A model with broad access will happily surface information to whoever asks, including things a particular person should not see.
The groundwork that pays off: know what you hold and where, define who may access what, decide what may leave your infrastructure, and keep an audit trail of what the AI was given.
This is unglamorous and it is regularly the reason AI initiatives stall. Doing it first is faster than doing it after an incident.