What this policy covers and where separate customer terms apply.
Approvals and responsibilities stay visible throughout the workflow.
Responsibilities, dependencies and limitations stay explicit.
Discovery and written scope
Each engagement begins with the intended outcome, selected agent responsibilities, inputs, systems, delivery model, approval gates, dependencies and exclusions. A catalogue entry alone does not activate a live provider or integration.
Implementation stages
A typical engagement may include discovery, configuration, connection checks, a bounded test, customer review, agreed corrections, training or handover, and controlled activation. The accepted quotation or SOW sets the actual stages.
Customer inputs and changes
Customers provide approved knowledge, accounts, permissions, content and decisions. Delays or incomplete inputs may affect the plan. New responsibilities, channels or integrations require written change control and may require a revised quotation.
Release and acceptance
Live release should follow verified permissions, safety and connection checks, customer approval and appropriate recovery planning. Acceptance evidence should be tied to the approved scope rather than an undefined promise that AI will always produce a particular outcome.
Hosting responsibilities
Managed, hybrid and customer-hosted models allocate infrastructure, patches, backups, monitoring, domain/DNS, provider accounts and incident response differently. The order documents must include a responsibility matrix.
Operational details pending confirmation
Standard milestones; acceptance criteria; correction allowance; target dates; customer-delay rules; maintenance/update entitlement; training and documentation; hosting matrix; and support start/end.