Agent Automations are standing jobs you hand the Enterpret Agent to run in the background. You define an objective once, say a weekly readout of a product area or a recurring triage of new tickets, and the agent runs it on its own on a schedule. It researches your feedback, does the work, and delivers the result where your team works: a report in Slack or email, or an action taken in a connected tool, like a filed ticket or an updated record. No one has to be at the keyboard.
The same agent, a different mode
The Enterpret Agent works in two modes. In a Session, you drive it turn by turn: you ask a question, it researches your Context Graph, and you steer it toward the answer. An Agent Automation is the same agent running the same kind of work in the background, on a cadence, with no one watching.
The capabilities carry over. An automation researches the Context Graph, cites every count and quote to a real feedback record, runs the platform skills, and produces the same structured report a Session does: an executive snapshot, themes that moved with week-over-week deltas, representative quotes, and a link back to the source feedback behind every claim. The difference is who starts the run and what happens with the result. A Session ends in your chat window. An automation runs on the schedule you set and delivers on its own, whether that is a report to your Slack or inbox or an action in a connected tool.
This is why a good automation often begins life as a Session. Once a question becomes one you ask every week, "what changed in my product area?", it is a candidate to hand off. Prove the prompt works in a Session, then set it to run on its own.
What an automation does with the result
An automation can do two kinds of work with what it finds.
Deliver a report. Most automations research the feedback over the window and send a report where your team already works, in Slack or email. The report carries the structure the agent's skills produce: a short executive snapshot, the themes that moved with volume and week-over-week deltas, representative verbatim quotes, and a link back to the source feedback behind every claim. Citations survive into the delivered report, so a finding in a Slack message is as verifiable as an answer in a Session. For how to read one end to end, see Reading automation reports.
Act in a connected tool. Because an automation runs the full Enterpret Agent, it can also take action in any third-party system you have added as a connector, not just send a report. It can file or triage a ticket, update a record, or post into the tool where that work already lives. Actions run under the connector's access, the same as in a Session.
Example automations
Each automation is a recurring job scoped to a subject. Some deliver a report; others take action in a connected tool.
Automation | What it does | Who it's for |
Product Area VoC | A recurring feedback digest for one product area. Leads with what changed, classifies themes as emerging, trending, or persistent, and carries representative quotes and a driver hypothesis for each. | The PM who owns the area |
Account Health Monitoring | A health digest across a configured account roster, whether top accounts, a CSM book, or a segment. Surfaces which accounts are getting louder or quieter, the risks worth escalating, and the wins worth amplifying. | CS and post-sales leadership |
Org-Wide VoC Report | A whole-company readout of what got worse, what got better, and what is new across the full feedback corpus, built for a five-minute read before the recurring review. | Product, support, and CX leadership |
Ticket triage | On each run, finds the new high-priority issues in your feedback and files or tags them in your connected tracker, such as Linear or Jira, so the queue is triaged before the day starts. | Support and triage owners |
CRM account flags | Flags accounts whose feedback signals rising risk and updates the account record in your connected CRM, so the owner sees it in the tool they already use. | CSMs and account owners |
For the full catalog and how to start from a pattern, see Automation templates.
How to get started
Decide what to automate. A good first automation is a question you already ask on a cadence and that the agent has the context to answer on its own, such as a weekly product-area readout or a recurring account-health check.
Create the automation. Set the prompt, the schedule, where the result goes, and the connectors and skills it should use. See Create a scheduled automation for the full walkthrough.
Let it run, then tune it. Read the first few deliveries, adjust the prompt or cadence, and pause or edit it any time from Manage your automations.
To start from a known-good pattern instead of a blank prompt, browse Automation templates for ready-made jobs like the ones above.
What to expect
An automation runs after its trigger fires, and the report arrives once the run completes, so the trigger time and the delivery time are not the same. A scheduled automation set for Monday 9am delivers its report shortly after, not at the stroke of the hour.
Connectors and skills that an automation uses run under the same model as a Session. Actions an automation takes in a connected tool are attributable to the user who created it and scoped to that person's access.
FAQs
Is an Agent Automation a different product from the Enterpret Agent? Agent Automations run the same Enterpret Agent in the background. A Session is the foreground mode where you drive the agent live; an Agent Automation is the background mode where it runs on a trigger and delivers on its own.
Where does an automation's output go? Reports are delivered to Slack or email, which you choose when you create the automation. An automation can also act directly in a connected tool, such as filing a ticket or updating a record, instead of or in addition to sending a report.
Are the reports cited? Yes. Every count and quote links back to the source feedback record, the same as an answer in a Session. Quotes are pulled word for word from the source rather than paraphrased.
Can I see what an automation found without leaving Slack? Today the report is delivered to your Slack channel or inbox.
How is this different from Enrichment workflows? Agent Automations are standing jobs the Enterpret Agent runs to research feedback and deliver reports. Enrichment workflows are a separate Enterpret feature.
Have a question about Agent Automations, or an automation you would love to run? Reach out at [email protected] and we will jump in.