How to automate intake & triggers
Start a workflow the moment something happens — a form, an email, a file, a record, a schedule.
Written by Max Zeshut
Founder at Agentmelt · Last updated Sep 12, 2026
In short
What intake & triggers means in practice.
Intake is the first step of every automation and the one most people underestimate. It answers a simple question — how does the workflow know something happened? — and the answer decides how fast, how reliable and how expensive everything after it will be.
There are four ways in: an event pushed to you (a webhook from a form, a CRM, a billing system), an inbox watched for new messages and attachments, a folder or table polled for new rows and files, and a clock. Most real processes use two of them, because the thing you want to react to arrives in more than one way — a claim comes as a form, an email and a phone note.
Where it shows up
34 automations use it.
Each link opens the full blueprint — trigger, every step, guardrails, template and what it costs to have it built. The step names below are the ones that do this job.
AI Competitor Monitoring
Snapshot monitored pages · Collect feeds
OperationsAI Appointment Reminders
Pull upcoming appointments
MarketingAI Social Listening
Collect mentions
EngineeringAI Unit Test Generation
Receive the PR and diff
EngineeringAutomated Code Review with AI
Receive the PR event
InsuranceAutomated Insurance Underwriting Intake
Capture and classify attachments
Security & ITAutomated Security Alert Triage
Receive and normalise the alert
HRCandidate Outreach Automation
Receive the candidate
OperationsClient Intake Automation
Receive the enquiry
Finance & accountingClient Reporting Automation
Pull the numbers
OperationsContract Review Automation
Capture and identify
Customer successCustomer Feedback Analysis Automation
Collect feedback
Customer successCustomer Onboarding Automation
Register the new account
OperationsDocument and Proposal Generation Automation
Receive the request
OperationsEmail Triage Automation
Receive and de-duplicate
HREmployee Offboarding Automation
Receive the termination record
HREmployee Onboarding Automation
Receive the new-hire event
SalesInbound Lead Qualification with AI
Receive the submission
InsuranceInsurance Claims Intake Automation
Capture the notice
HRInterview Scheduling Automation
Read the stage
Finance & accountingInvoice Processing Automation
Capture the invoice
Security & ITIT Helpdesk Automation
Receive the request
SalesLead Nurture Automation
Detect the reason
Supply chain & procurementLogistics Optimization with AI
Receive the shipment
SalesMeeting Prep Automation
Find upcoming external meetings
HRPayroll Automation
Collect period inputs
Customer supportProactive Customer Outreach
Receive or detect the signal
Supply chain & procurementPurchase Order Automation
Receive the request
MarketingReal Estate Listing Description Automation
Receive the listing
HRResume Screening Automation
Receive the application
MarketingReview Response Automation
Collect new reviews
SalesSpeed-to-Lead Automation
Receive the lead
Finance & accountingSubscription Dunning Automation
Receive the billing event
Customer supportSupport Ticket Deflection with AI
Receive the ticket
How it works
Step by step.
Prefer pushed events over polling
If the source system can send a webhook — HRIS hires, billing events, pull requests, form submissions — use it. It fires within seconds and costs nothing to wait. Polling an inbox or a table every few minutes is the fallback for systems that cannot push, and it needs a cursor (last seen id or timestamp) so nothing is processed twice.
Normalise at the door
Whatever arrives — JSON from a webhook, an email with three attachments, a CSV row — the intake step turns it into one shape the rest of the workflow understands: source, received-at, the payload, the raw original kept for audit. Every later step then only has to handle one input format.
De-duplicate and make it idempotent
Sources retry. Webhooks fire twice, emails get forwarded, files get re-uploaded. The intake step keys every item (an external id, or a hash of sender + subject + attachment) and drops or merges repeats, so a retried event never becomes two invoices or two tickets.
Fail loudly, not silently
Intake that swallows a malformed payload hides the problem for weeks. Anything that cannot be normalised goes to a dead-letter list with the raw item and a reason, and somebody is told.
By tool
How it is built, by tool.
The step is the same whichever tool runs it; this is where it lives in each. We build in the one that fits your accounts and budget, and you own the result.
- n8nWebhook, Gmail Trigger / IMAP, Schedule Trigger, and app triggers (HubSpot, Stripe, GitHub); a Code step for the idempotency key.
- MakeCustom Webhook, Watch Emails, Watch Records; scheduling on the scenario; a data store for seen keys.
- ZapierWebhooks by Zapier (Catch Hook), New Email, Schedule by Zapier; Storage by Zapier or Formatter for de-duplication.
- CodeA small HTTP endpoint plus a queue (SQS, Cloud Tasks) with a unique-key constraint on the inbox table.
Where it goes wrong
The failure modes we design against.
No idempotency key
The single most common cause of duplicate records downstream. Decide the key before the first run.
Trusting the payload
A form field can hold anything, including instructions aimed at the AI steps later in the workflow. Treat intake content as data, never as commands.
One trigger per source
A process that starts from three sources but has one trigger quietly drops the other two. Map every entry point on day one.
See what we automate
The automation libraryEvery process, by department, each with a price and a free template.
Or get a plan for your own process
Three sentences; one option and its price appear in minutes. No call.
FAQ
Questions people ask.
Can an automation start from an email?
Yes. A mailbox trigger watches an inbox (or a label) and starts the workflow for each new message, with attachments available to the extraction step. For shared inboxes it is the most common intake there is.
How do I stop the same item being processed twice?
Give every incoming item a stable key — an external id if the source has one, otherwise a hash of the fields that identify it — and check it against what has already been seen before doing anything else.
Webhook or polling?
Webhook whenever the source offers one: instant and free. Poll only when it does not, and keep the interval honest — a five-minute poll on a high-volume inbox is fine; a ten-second poll on a rate-limited API is not.