If you built a Relay.app workflow that quietly handled lead routing, meeting notes, or client onboarding, you already know the bad news. The Relay.app shut down and took your automations with it. The frustrating part isn’t losing the tool. It’s losing the logic you spent hours tuning.
This guide is a rebuild manual. I’ll walk through three common Relay.app workflows and show you how to recreate them step by step in Zapier and n8n. No fluff, no “just use another tool” hand-waving.
Before You Rebuild, Write Down the Workflow
Open a blank doc and answer four questions about the old automation:
- What starts the workflow? A new row in Airtable, a form submission, an email landing in a shared inbox.
- What decisions does it make? Is this lead qualified? Which team owns this ticket?
- Where does a human approve something? Before an email goes out, before a deal stage changes.
- What is the final output? A Slack alert, a task, a draft reply, a weekly report.
Example: a Relay.app workflow took Typeform responses, asked an AI model to score fit from 1 to 100, and sent a Slack alert only for scores above 80. That maps to trigger, decision, human approval if needed, and output. Once you have that map, you can rebuild it anywhere.
Rebuild 1: Lead Triage With AI Scoring and Slack Alerts
The original Relay.app logic
Typeform submission came in. An AI step scored the lead. If the score was above 80, a Slack message went to #sales. If not, the lead landed in a nurture list. A rep could approve or edit the Slack message before it posted.
Step-by-step in Zapier
Step 1: Create the trigger. In Zapier, choose Typeform as the trigger app and “New Entry” as the event. Connect your form.
Step 2: Add an AI scoring step. Use Zapier AI or an OpenAI action. Prompt it with: “Score this lead from 1 to 100 based on budget, company size, and timeline. Return only the number.” Feed in the form fields as variables.
Step 3: Add a filter. Set the filter to continue only when the score is greater than 80. Everything else can go to a Google Sheet or your CRM as a nurture lead.
Step 4: Send the Slack alert. Choose Slack “Send Channel Message.” Put the lead name, score, and a one-line reason in the message.
Step 5: Add human approval. If you want a rep to approve the message first, use Zapier’s “Human in the Loop” step or send the draft to a review channel where someone reacts with a checkmark. For judgment-heavy steps, the approach in Zapier AI for work that needs judgment is worth copying.
Rebuild 2: Support Ticket Summaries and Routing in n8n
The original Relay.app logic
A new Zendesk ticket arrived. An AI step summarized it in two sentences. A routing rule sent billing questions to the finance Slack channel and technical questions to a Linear project. The ticket owner got a notification.
Step-by-step in n8n
Step 1: Add a webhook trigger. In n8n, use the Webhook node and paste the URL into Zendesk’s webhook settings. Set it to fire on new tickets.
Step 2: Add an AI Agent node. Connect your OpenAI or Anthropic credentials. Give the agent a system prompt: “Summarize the ticket in two sentences. Then classify it as billing, technical, or other. Return JSON with summary and category.”
Step 3: Add a Switch node. Route on the category value. Billing goes to one branch, technical to another, other to a catch-all.
Step 4: Create the task. In the technical branch, use the Linear node to create an issue with the summary as the description. In the billing branch, send a Slack message to #finance.
Step 5: Notify the owner. Add a Slack node that mentions the ticket assignee with a link back to Zendesk.
n8n is the better fit when you want self-hosting, complex branching, or AI agents that can call tools. The n8n AI agents guide covers memory, tool use, and the mistakes that break agents in production.
Rebuild 3: Weekly Client Reporting With an Approval Gate
The original Relay.app logic
Every Friday at 9 a.m., the workflow pulled Google Analytics and Stripe data, asked AI to write a plain-English summary, sent the draft to the account manager for approval, then emailed the client after approval.
Step-by-step in Make or Zapier
Step 1: Add a schedule trigger. Set it for Friday at 9 a.m. in your timezone.
Step 2: Pull the data. Use Google Analytics and Stripe actions to get sessions, conversions, and revenue for the last seven days.
Step 3: Generate the summary. Send the numbers to an AI step with a prompt like: “Write a 150-word client update. Mention wins, one concern, and one recommendation. Use plain language.”
Step 4: Create the approval draft. In Gmail, create a draft to the client with the AI summary. Send a Slack message to the account manager with a link to the draft and two buttons: approve or edit.
Step 5: Send after approval. When the approval comes through, send the Gmail draft. If the manager edits it, send the edited version.
That approval gate was one of the features people loved most about Relay.app. The rise and sudden disappearance of Relay.app is a reminder that human-in-the-loop automation is valuable when it’s built into the workflow, not bolted on at the end.
Where Human Approval Actually Belongs
Not every step needs a human. Adding approval to everything creates a bottleneck that makes the automation slower than doing the work manually. Use a human check when the output is customer-facing, financial, or hard to undo.
- Customer emails and proposals: approve before sending.
- Internal Slack alerts: no approval needed.
- CRM deal stage changes: approve if it affects forecasting.
- Weekly client reports: approve if the client will see it.
- Refunds or credits: always approve.
Test With One Real Record, Not a Demo
Don’t switch the workflow on for every lead or ticket. Pick one real form submission from last week. Run the workflow manually. Check the AI output. If the score is nonsense, shorten the prompt and give the model a few examples of good and bad leads. If the routing is wrong, add a keyword filter before the AI step. Repeat with three edge cases: a lead with no company name, a ticket written in all caps, and a client with missing Stripe data. That small test catches most of the failures you’d otherwise discover in production.
What to Do With Workflows That Need More Than a Rebuild
Some Relay.app workflows leaned on features that don’t map cleanly to Zapier or n8n. Multi-step approvals, shared team inboxes, and AI memory are possible, but they take more setup. Be honest about the cost. If a workflow saves two hours a week, rebuilding it is worth a Saturday morning. If it saved five minutes a month, let it go. Relay’s team joined Google’s Chrome team, so there’s no comeback tour to wait for. You can read more about Relay’s team joining Google’s Chrome team if you’re curious about the aftermath, but your time is better spent on the workflows that move revenue.
Start with the automation that touches money or customers. Map the trigger, rebuild one action, test it with a real record, then add the approval step. The rest is iteration.