Tomorrow, an AI agent joins your company.
Would you give it access to your CRM, email, customer files and internal systems immediately?
Most leaders would never do that with a new employee.
Yet with AI agents, many organisations are getting surprisingly close.
Agents can already read customer data, update records, send messages, trigger workflows and interact with internal systems.
Once they can act inside the organisation, access becomes a leadership question as much as a technical one.
Microsoft is already moving in this direction with Entra Agent ID, giving AI agents dedicated identities, security roles and controlled permissions.
That makes sense.
We already know how to onboard people.
We define their role. Limit initial access. Assign responsibility. Supervise important decisions. Expand authority as trust grows.
AI agents may need the same discipline.
I increasingly think every serious enterprise agent should have an identity, an owner and something resembling a probation period.
Start by observing.
Then recommend.
Then act with approval.
Only after the organisation understands the risks should autonomy increase.
Because an agent with access to your systems is no longer just a tool sitting outside the organisation.
It has become an actor inside it.
And every actor with authority needs boundaries.
If an AI agent joined your company tomorrow, what would you let it touch on day one?