Automation Recipes

Perspective automations let you push conversation data into the tools your team already uses. Below are the most common patterns, with enough detail to get each one running.

Built-In Channels#

Before reaching for external tools, check whether a built-in channel covers your use case. Perspective supports direct webhook/email delivery plus connected-app channels through Slack, HubSpot, and Gmail:

Automations are configured from the Perspective's Design page or by asking the perspective assistant. If you are not already in Design, open the Perspective actions menu and choose Edit. In the right-side panel, switch to Settings and expand Automations. Connected-app channels guide account connection, and the Automations list reports recent delivery errors. Direct webhooks are retried automatically for transient failures.

Developer note: MCP automation tools can create webhook, email, and connected-app (composio) automations. Participant Email and recurring invite automations are configured from the in-app automation assistant, though existing automations can appear in automation lists and tests.

Recipe 1: Sync Leads to Your CRM#

Goal: When a concierge conversation completes, create or update a contact in your CRM using the participant's details and the form fields the conversation captured.

Using the built-in HubSpot channel:

1. Open your Perspective. If you are not already in Design, open the actions menu and choose Edit. In the right-side panel, switch to Settings and expand Automations.
2. Add a new automation with trigger Per Interview.
3. Select HubSpot as the channel and connect your account.
4. Choose the HubSpot action and map conversation fields or participant metadata to the action inputs.

Using a webhook with any CRM:

1. Set up a webhook automation pointing to your server or a Zapier/Make webhook URL.
2. In your handler, read structured_output (the form fields the conversation captured) and participant_metadata (URL params, embed attributes, invite context, etc.) from the payload.
3. Use your CRM's API to create or update the contact.

// Example: Zapier webhook handler pseudocodeconst {  event,  interview_id,  research_id,  participant_metadata,  structured_output,} = payload;
// Webhook automations fire on partial/abandoned events too.// only sync completed conversations to the CRMif (event !== "interview.completed") return;
await crm.upsertContact({  email: structured_output?.email ?? participant_metadata?.email,  source: "perspective",  interview_id,  campaign: participant_metadata?.utm_campaign || "unknown",});

Recipe 2: Slack Notifications on Every Conversation#

Goal: Post a summary to a Slack channel each time a conversation completes.

Using the built-in Slack channel:

1. Add an automation with trigger Per Interview.
2. Select Slack as the channel and authorize access to your workspace.
3. Pick the target channel (e.g., #research-updates).
4. Optionally add an instruction to customize the message format -- the automation agent will use it to shape the Slack post.

Using a webhook with a relay:

Slack Incoming Webhooks expect their own JSON shape ({"text": "..."} or {"blocks": [...]}), so a Perspective webhook can't be sent directly to a Slack Incoming Webhook URL. To use a webhook for Slack delivery, point Perspective at a relay (Zapier, Make, or a small service of your own) that turns the interview payload into a Slack message and forwards it.

To split low-trust conversations into a separate channel like #escalations, set up a second automation on the same perspective with a trust-score filter capping the maximum.

Recipe 3: Scheduled Digests#

Goal: Send a daily or weekly summary of completed conversations to your team via email or Slack.

1. Add an automation with trigger Scheduled (daily or weekly).
2. Select your channel: Email for an inbox digest, or Slack for a channel post.
3. The automation agent aggregates completed conversations from the period since the last digest and sends a single summary.

Scheduled automations track their own cadence -- nextRunAt advances automatically after each run. You can set the timezone in the automation settings so digests arrive at a reasonable hour.

Recipe 4: Recurring Participation Invites#

Goal: Send a daily or weekly prompt that asks a fixed audience to start a new conversation, such as a standup, reflection, or recurring feedback check-in.

1. Ask the perspective assistant to create a scheduled invite automation.
2. Choose a daily or weekly schedule and timezone.
3. Use Email for a recipient list, or Slack for a connected Slack destination.
4. Provide instructions for the invite copy and include any required context.

Invite automations use the perspective's share URL. When an invite automation is created, Perspective enables multiple conversations per participant so the same person can respond on each recurrence.

Recipe 5: Conditional Follow-Up Emails#

Goal: Send different follow-up emails to participants based on how the conversation went.

1. Add an automation with channel Participant Email.
2. Create up to 5 email templates, each with a name, subject, body, and an intent description explaining when to use it.
3. Set a default template as the fallback.
4. The automation agent reads the conversation and selects the most appropriate template based on the intent descriptions.

Example templates:

• Positive feedback: Thanks the participant and offers a referral link.
• Feature request: Acknowledges the request and links to the public roadmap.
• Escalation needed: Loops in a human with context from the conversation.

Recipe 6: Data Export to a Warehouse#

Goal: Sync each conversation's captured fields and metadata to Snowflake, BigQuery, or another data warehouse for downstream analysis.

1. Set up a webhook automation with trigger Per Interview.
2. Point it to a lightweight ingestion service (e.g., an AWS Lambda or Cloud Function).
3. In the handler, transform the payload and write to your warehouse.

# Example: Lambda handler writing to Snowflakedef handler(event, context):    payload = json.loads(event["body"])    row = {        "interview_id": payload["interview_id"],        "research_id": payload["research_id"],        "event": payload["event"],        "status": payload["status"],        "delivered_at": payload["timestamp"],        "structured_output": json.dumps(payload.get("structured_output")),        "participant_metadata": json.dumps(payload.get("participant_metadata")),    }    snowflake_cursor.execute(INSERT_SQL, row)

For batch workflows, use a scheduled automation to trigger a nightly export instead of processing each conversation individually.

Recipe 7: Zapier and Make (Integromat)#

Both Zapier and Make can receive Perspective webhooks as triggers.

Zapier:

1. Create a new Zap with trigger Webhooks by Zapier > Catch Hook.
2. Copy the webhook URL into a Perspective webhook automation.
3. Test the trigger by completing a conversation.
4. Add actions: create CRM contacts, send emails, update spreadsheets, or anything else Zapier supports.

Make:

1. Create a new scenario with a Webhooks > Custom webhook module.
2. Copy the URL into Perspective.
3. Map the JSON fields to downstream modules.

Both tools handle retries and error notifications on their end, so you get two layers of reliability (Perspective retries + Zapier/Make retries).

Tips#

• Use conditions to reduce noise: Filter automations by tags or trust score so you only process the conversations that matter.
• Use execution mode intentionally: Direct mode sends a raw webhook payload without an LLM and currently supports webhook channels only. Agent mode reads the conversation and drafts a message or action for Slack, email, HubSpot, Gmail, and participant follow-ups.
• Monitor failures: Perspective tracks consecutive failures per automation. If an automation fails repeatedly, check the error in the Design page's Automations section in the Settings view and verify your endpoint or connected app is still reachable.