Assistant
Responds to a request
Usually waits for a prompt and returns information or guidance.
Works from an instructionPersonal AI explained
What is a personal AI representative?
A personal AI representative is an AI entrusted to handle defined interactions on someone’s behalf, using context, permissions and boundaries established by that person.
It does not replace personal judgement. It manages the interactions that do not always require it—and returns control when they do.
Assistants answer. Agents execute. Representatives handle relationships and situations.
The distinction is not a matter of branding. It changes what the system must understand, what authority it needs and when it must return control.
Agent
Completes a defined goal
Uses tools and follows a workflow to deliver a task or outcome.
Works from a goalRepresentative
Handles an interaction
Uses context, delegated authority and boundaries to handle or escalate a situation.
Works from context + permission
The four foundations
Representation requires more than intelligence.
A system may be capable of producing a fluent answer without being qualified to represent a person. Four foundations must exist together.
Context
The people, responsibilities, preferences and circumstances that change what an interaction means.
Permission
What the representative may access, disclose, confirm, change or complete.
Boundaries
Where authority ends and when the person must be involved.
Accountability
A reviewable record of what happened, why it happened and what can be corrected.
Levels of delegated authority
A representative should not begin with unrestricted authority.
- 1
Observe and summarise
Receive an interaction and explain what happened.
- 2
Ask and collect
Gather information without making a commitment.
- 3
Respond within approved rules
Give information or decline requests inside defined boundaries.
- 4
Complete routine actions
Confirm, book, rearrange or follow up where permission is explicit.
- 5
Escalate exceptions
Bring the person in when judgement, consent or urgency requires them.
Representation is not autonomy
Authority must never be assumed.
A personal AI representative should be designed to pause at the edge of its authority—not improvise past it.
- Access does not automatically permit disclosure.
- Previous consent does not apply forever.
- Routine authority does not extend to consequential decisions.
- Every relationship should not be treated the same.
- Confidence is never a substitute for asking.
- The user must always be able to withdraw permission.
Where representation may apply
The model extends wherever the outside world asks for your attention.
These are category examples, not a claim that every representative system should act across every channel.
Calls and messages
Understand purpose, collect details and decide whether an interruption needs the person.
Appointments and availability
Check approved options, confirm routine arrangements and escalate conflicts.
Household administration
Handle low-risk confirmations, changes and everyday coordination.
Service interactions
Gather information, wait on hold and return the useful outcome.
Bookings and changes
Complete straightforward requests where the rules are already clear.
Follow-ups and coordination
Track routine next steps without making the person manage every exchange.
Accountability and review
Every action should leave a clear path back to the person.
A representative should make it possible to see what permission was used, what action was taken, what outcome was produced and when control returned to the user.
That record should be understandable, reviewable and correctable—not hidden inside an opaque system.
Questions people should ask
Any AI claiming to represent you should answer these clearly.
What does it know about me?
Who may it speak to?
What information may it share?
What may it confirm or change?
When must it ask first?
Can I review what it did?
Can permission be withdrawn?
Who is responsible when it acts?
One implementation of the category
How Cocoon approaches personal representation.
Cocoon is a consumer personal AI representative designed around relationship-aware context, permission-led action and protection of attention.
It begins by learning how the user wants everyday situations handled, then works within those boundaries across calls and other approved interactions.
Common questions
Personal AI representation, clearly defined.
What is a personal AI representative?
A personal AI representative is an AI entrusted to handle defined interactions on someone’s behalf, using context, permissions and boundaries established by that person.
How is a personal AI representative different from an AI assistant?
An assistant normally responds to a prompt. A representative can receive an interaction, interpret it in context and handle it within delegated authority.
Is a personal AI representative the same as an AI agent?
No. An agent is usually organised around completing a goal. A representative must also understand relationships, disclosure, urgency, boundaries and when control must return to the person.
Can a personal AI representative make decisions for me?
Only within authority you have explicitly provided. Consequential, sensitive or uncertain decisions should be returned to you.
What permissions does a personal AI representative need?
Permissions should be specific to the information it may access, the people it may interact with, what it may disclose and which actions it may complete.
Can I review or reverse its actions?
A trustworthy representative should provide a clear record of what it did and why, and allow permissions and reversible actions to be changed where possible.
What happens when it is unsure?
It should ask, pause or escalate rather than hide uncertainty or exceed its authority.
Who is responsible when it acts?
Responsibility depends on the service and circumstances, but the provider should clearly identify the accountable company, the user’s controls and the process for reviewing or correcting an action.
Can permissions be withdrawn?
Yes. Delegated authority should be changeable and withdrawable rather than assumed to be permanent.
Is Cocoon a personal AI representative?
Cocoon is the consumer product and protected environment in which a personal AI representative learns the user’s context, works within permissions and handles approved interactions.