AI · Agents
Agents that carry out multi-step work — through the same workflow a person would use.
AI Agents go a step beyond the Assistant: given a goal, an agent can look across related records, propose a set of actions, and create the follow-up tasks or approvals needed to carry them out — semi-autonomously, and always inside your existing permissions and workflow engine.
A worked example: stalled purchase requisitions
A procurement lead asks the Assistant a question; an agent takes it from analysis to a concrete next step.
"Which purchase requisitions are stuck waiting on approval?"
Agent identifies 12 requisitions unapproved for more than 3 business days
Agent cross-checks vendor lead times against each requisition's need-by date
Agent flags 5 requisitions now at risk of missing their need-by date
Agent drafts an expedite request and creates a follow-up task for the buyer to review and send
Illustrative worked example — not a live query against a connected tenant.
Agents act inside your permissions and workflow
An agent reads and writes through the same entity model, workflow engine and role-based access control as every other user — it does not have a side channel into your data. A task an agent creates goes through the same approval a person would need, and shows up in the same audit trail as a human-driven change.
Other domains agents apply to
HR
Identify onboarding tasks at risk of missing a start date and create reminders for the owning manager.
Finance
Surface invoices likely to be disputed based on prior history and draft a review task before they are paid.
Sales & CRM
Flag deals with stalled activity and draft a follow-up task for the account owner.
See how automation and agents fit together.
Agents propose and prepare multi-step work; AI Automation layers smarter triggers on top of the processes that already run it.