Assistant Scope Lock and Access Provisioning (Onboarding)

Assistant Scope Lock and Access Provisioning (Onboarding)

1. 01

   Write the assistant's job description before anyone connects an inbox

   Name the three tasks it owns (scheduling, first-draft replies, meeting notes) and the three it never touches (pricing, contracts, client escalations). Ambiguity here is what produces off-brand replies in month two.

2. 02

   Map every data source the assistant will read against a named owner

   For each connected mailbox, calendar, or drive, record who at the agency can revoke access and how quickly. A Windows-native assistant such as MyHandler indexes screen content and messages into a local database, so the revocation path differs from a cloud tool.

3. 03

   Set the retainer-hour ceiling the assistant is allowed to influence

   Decide whether automated scheduling and drafting count against a client's retainer or sit in agency overhead. Agencies that skip this step discover the ambiguity during the first invoice dispute.

4. 04

   Run a two-week shadow period where drafts are queued, not sent

   An assistant like Jace can draft replies in a named person's voice, but voice fidelity on a new client account is unproven until a human reviews 20 to 30 real threads.

5. 05

   Define the escalation phrase that forces human takeover

   Pick one trigger (a client writing 'urgent', a thread with more than four participants, any message containing a dollar figure) and document it where the delivery team can find it.

6. 06

   Confirm the fallback path for the day the underlying model is unavailable

   Simultaneous outages across major providers have already disrupted agencies running single-provider workflows, so name the manual process and the person who runs it.

7. 07

   Log the go-live decision with a named approver and a review date

   Thirty days out is a workable first checkpoint. The review asks one question: did the assistant reduce coordination overhead without a client noticing a change in tone?