Agent Scope Contract (Onboarding)

Agent Scope Contract (Onboarding)

1. 01

   Write the agent's job description as a single sentence with a measurable output

   For example: 'Draft first-touch replies to inbound demo requests within 15 minutes, routed to the correct account executive.' If the sentence needs a comma splice to stay honest, the scope is too wide for a first deployment.

2. 02

   List every system the agent must read from and write to, with the credential owner named

   A white-label receptionist agent such as Vendasta's touches calendar and CRM; a research agent such as Dust's pulls from shared knowledge sources. Each connection is a client-side approval, not an agency-side assumption.

3. 03

   Define the escalation path for anything the agent cannot resolve in two attempts

   Name the human, the channel, and the response window. Agents that fail silently are the fastest way to lose a retainer; agents that hand off cleanly become the reason the retainer renews.

4. 04

   Set the review cadence and the kill switch before the first live run

   Agree on a weekly output review for the first month, then monthly. Document who can pause the agent and how, in writing, inside the client's own workspace.

5. 05

   Price the engagement as wiring plus oversight, not as a license pass-through

   The agent subscription is a line item; the integration, prompt tuning, exception handling, and monthly review are the billable service. This is where agencies hold margin when the underlying model changes.

6. 06

   Record the baseline metrics the agent is expected to move

   Capture current response time, throughput, or error rate before launch. Without a baseline, the 90-day review becomes a debate about impressions instead of a conversation about numbers.