Skip to main content
A Play watches your customers for an outcome you care about: keeping onboarding moving, preparing for renewals, or following up on commitments. It follows a skill to investigate customer context, plan the work, and carry out permitted changes or request review. Follow its activity and results in Action History and Actions. Manage Plays from the Agents page. Use the Plays filter to find them.

Creating a Play

Choose Start from scratch → Play → Create Play, or Create from template to start with a configured Play and its skill. Templates cover renewal preparation, touchpoint follow-up, growth and expansion, onboarding, collections, account monitoring, and project updates. Before publishing, connect the data sources your skill needs. If you want to target a particular group, create a customer segment under Configuration → Customer Segments first. In the Builder tab:
1

Define the successful outcome

Describe what the Play should help achieve. For example, “Keep onboarding on track by identifying blocked tasks and commitments that need a follow-up.” This guides which changes are worth turning into Actions. You can rename the Play and edit its description in Play settings.
2

Choose the audience and cadence

Target All customers or a Specific segment. A customer can belong to multiple Plays.Under When to run, choose When data changes or On a schedule. For data changes, describe a trigger such as “deal lost” and choose Add. Quivly matches it to available records and fields, including custom fields. Review the plain-language conditions before publishing.Use Suggest triggers to start from the successful outcome and Skill, or Improve triggers to review suggested changes to an existing list. Improvement suggestions take effect only when you apply them. Supported sources include calls, Slack threads, health changes, customers, deals, tickets, invoices, subscriptions, and contracts; only available sources can start a Play.Scheduled Plays can run hourly, daily, weekly, or monthly; scheduled times use whole UTC hours. Use a schedule for recurring summaries or finding inactivity. Switching to a schedule replaces the data triggers. With no data triggers and no schedule, the Play runs only through Run now.
3

Select a skill

Choose a published skill that defines what to investigate and which data to read. Use the preview icon to inspect it, or Create to describe a new skill with AI. Publish the skill before using it in a live Play.
4

Choose actions

In the Actions panel, turn on the Quivly areas or connected apps the Play may use. Open a group’s details to inspect its operations, or use Suggest actions for recommendations. Suggestions take effect only after you accept them.Turning on an app or Quivly area enables all its available actions, including actions added later. Slack sends and webhook sends remain individual choices. Review these choices before publishing, especially for automatic Plays.A blank Play starts with no actions enabled; templates keep their configured choices. You can leave actions off to surface findings without proposing changes.Read access alone does not enable sending emails or creating calendar events. A Skill’s drafted message or invitation can remain a manual next step for you to review and use; it is not a sent message or scheduled event.Under Apps this Play can read, add other apps for lookups without allowing changes. Apps with enabled actions can also be read. A Skill supplies instructions; the Play’s own configuration determines its permitted reads and actions.
5

Configure delivery

Choose where new Action alerts go: Slack, email, or both. Actions also appear in the Actions inbox. Delivery settings are explained below.
6

Choose approval

Select Manual to review changes before execution, or Auto to run permitted changes using the publisher’s current workspace access. The customer owner reviews Manual changes; if there is no owner, the authorized person who ran or published the Play can review them.Auto still requires review for destructive, high-stakes, customer-visible, or unclassified changes, and messages to customers or outside the company. Three changes in a row that need review switch the remaining run to review. Ordinary edits and lookups do not trigger this switch, regardless of how many the run makes. If the Play needs a person’s answer, it asks and waits in either mode.
Check the saved results before treating a change as complete. A Play can finish with unresolved work for a person; its wrap-up explains what happened and what remains.

Delivery

Send test Slack message and Send test email send a sample alert without running the Play or creating Actions. Slack owner-delivery tests use your DM. Email tests go to all selected destinations, with your address standing in for Account Owner; live alerts use each customer’s owner. Test samples contain no customer data.
A custom email address can be outside your workspace. The email includes Action details, but opening the Action still requires workspace access.

Previewing and publishing

Open Preview, choose a call, Slack thread, or customer in the audience, and select Run preview. Inspect the plan, activity, questions, and simulated changes in the Quivly and Slack views. New notebook proposals with authored text show their full body and source links before approval. Preview uses real customer data without saving the proposed changes or sending messages. Preview runs remain available in Action History. For call-triggered work, the selected call and permitted records actually read support that run. Verification can include other recordings of the same meeting; selecting one meeting does not authorize using unrelated calls’ transcripts or summaries. Summary-only calls can run when the available evidence supports the work; having transcript readers available does not itself require a transcript. Older calls can be previewed without making them eligible for automatic follow-ups. An administrator can choose Run Play for an eligible selection after publishing. This runs the published Play for real and follows its Manual or Auto setting. Explicitly selected historical calls can be run after the current audience, access, and source checks; this does not enable automatic historical delivery. If your skill requests a combined digest, you can preview a report across the audience. Automatic digests require a scheduled cadence and a Slack channel; owner DMs deliver customer-specific Actions. Portfolio reports in Slack can include relevant people, profile and evidence links, material qualifications, and concrete next actions. Read the qualifications alongside the findings. Processing limitations stay in the approval review; older saved reports keep their separate next steps. Click Save draft, then Publish Play. Publishing requires a published skill that identifies the data the Play needs, plus a valid destination for each enabled delivery channel. The Play evaluates matching customers when the required data is ready. Edits stay in a draft while the published version keeps running. Save and publish again to make changes live. Use the Active toggle or Play settings to pause the Play. If a published Play targets a segment but its draft no longer has one, publishing asks whether to Publish for all customers. Choose a segment instead if you intend to keep the audience restricted.

Reviewing results

Open Action History to inspect live and saved runs, previews, plans, activity, questions, and results. Select a waiting run to answer its question or review changes. Runs retain progress across approval pauses. Where offered, use Pause to hold after the current step, or Stop to end the run and withdraw pending review requests. Changes already completed remain saved. Resume continues unfinished work using the current Play instructions; Retry rechecks a selected failed or uncertain change. Review partial outcomes before retrying. Use Run again to start a fresh run after completed work. For example, Automate project updates can read its source conversation and propose changes to projects, tasks, and milestones. With Manual, an authorized reviewer approves those changes before they are applied. Use a connected app’s own actions for follow-up work in that app; the former built-in call-commitments operation is no longer available. Replace that selection in older Plays and run them again to prepare current proposals.

FAQ

A workflow follows a defined sequence of steps on a visual canvas. A Play uses a skill to assess customer context against your desired outcome and decide which findings and next steps to surface.
Check that the Play is published and active, the customer is in its audience, and the required data sources are connected and ready. Preview the customer to see whether data is missing, the issue has already been surfaced, or there is no new action worth taking. A scheduled evaluation does not guarantee a new Action.
No. Connecting an app for actions lets the Play propose supported operations in that app. It does not automatically add the app’s incoming events as triggers.
Manual requires an authorized reviewer’s approval. Auto permits ordinary changes with the publisher’s current workspace access, while destructive, high-stakes, customer-visible, or unclassified changes, and messages to customers or outside the company still need review. Slack and email alerts about Actions follow the Play’s delivery settings.