Overview
AI (branded “Ami” in-product) is AmDital’s cross-module AI surface: a general assistant, five domain-specific agents, a visual automation builder, a prompt library, run history, and usage/cost tracking. Because agents and automations act across modules, this is the one module with no single dependency — it can read and act on data from any other module you’ve enabled, plus Integrations for webhook and MCP connections.Where to find it
Core workflows
Chat with the assistant or a domain agent
- Open AI for the general assistant, or pick a domain agent (Sales, HR, Finance, Support, Dev) for one scoped to that module’s data and actions.
- Memory stores facts the assistant should remember across conversations (e.g. your name or a standing preference) — add or edit entries directly.
- Every conversation is recorded in History, searchable later.
Build an automation
- Go to Automations → Builder to define a trigger and the steps that run when it fires, or start from a Template.
- Triggers lists the trigger rules across all your automations; Webhooks manages automations kicked off by an inbound webhook rather than an in-app event.
- A failed or paused run shows up filtered at
/ai/automations?status=failedor?status=pausedfor triage.
Manage prompts
- Prompt Library stores reusable prompts; Test a prompt before saving it into the library.
Track usage and cost
- Usage and Costs report AI call volume and spend, useful for keeping an eye on a workspace’s AI budget.
Permissions
Workspace-scoped. Individual domain agents are expected to respect their parent module’s own permission gates (e.g. the Finance agent shouldn’t surface data a viewer couldn’t see in Finance directly) — this is not yet broken out into a separate, explicit permission matrix per agent.Known limitations
- Per-agent permission gating (beyond the general workspace-scoped access) is not yet a distinct, documented matrix — don’t assume a domain agent enforces the exact same field-level restrictions as its parent module without checking directly.