Email Triage Automation Every message classified, routine replies drafted, a short list for the person
Your inbox is sorted before you open it: replies drafted for approval, tasks extracted, and one short list of what needs you.
- Installed for you · one-time$249
- Run for you$197/month
- Custom build · this process$2,000–4,000 one-time
- Starts whenGmail / Outlook trigger (new message)
Per inbox, up to 3,000 messages a month. Multiple inboxes or a custom set of reply playbooks beyond the standard ten are a custom build. Custom builds start at $2,000 site-wide; this band is for this process and its integrations.
Email Triage Automation is an automation Agentmelt installs in Slack. The free template is a file you import yourself; the kit ($49) adds the setup guide and prompts. Installed for you costs $249 one-time and is live within two working days of access, with a recorded walkthrough and 30 days of fixes. Run for you costs $197/month, with setup and maintenance included. A custom build of this process is $2,000–4,000 one-time; custom builds start at $2,000 site-wide. No call is needed for any option. Written by Max Zeshut; last revised 2026-09-21.
Reference stack: Claude + n8n. Node names are on each step; the file is under the template.
Founder at Agentmelt
Reviews every build before it goes live; the workflow escalates to a person, never past one.
What is email triage?
Email triage automation turns an inbox into a queue with the decisions already made: every incoming message is classified — needs a reply, for information, a task, a meeting request, a newsletter, spam — the context is fetched from the CRM, tasks and deadlines are extracted into the task tool, routine replies are drafted from a small set of playbooks for a one-click approval, and what reaches the person is a short list ranked by urgency plus a daily digest. It runs on shared inboxes (info@, sales@, support@) and on an executive’s own, and it never sends a reply on its own until a playbook has earned auto-send over weeks of unedited drafts.
- The shared inbox has 400 unread and three people who each think someone else owns it.
- The important email is in there somewhere, under the newsletters and the automated notifications.
- The same ten questions get answered by hand, differently, every day.
- Deadlines mentioned in emails live only in the emails.
- The executive spends the first hour of every day reading, not deciding.
If two of these are yours, this is the process to look at first. Free audit — or read on.
What it looks like once it runs.
Email triage automation turns an inbox into a queue with decisions already made: every incoming message is classified (needs reply, FYI, task, meeting request, newsletter, spam), routine replies are drafted for approval, tasks and deadlines are extracted into your task tool, and what reaches the person is a short list ranked by urgency. It runs on shared inboxes (info@, sales@, ops@) and executive inboxes alike.
- 7:30 Overnight mail is classified: reply needed, task, meeting, FYI, invoice, newsletter.
- 7:35 Routine replies are drafted from your playbook; tasks with deadlines are created in your task tool.
- 8:00 One digest: handled, waiting for you, extracted tasks, and what the sorter was unsure about.
- During the day You tap approve on a draft and it sends from your address.
Shared inboxes are where requests go to be forgotten, and executives spend two hours a day deciding what an email is before doing anything about it. Replies to routine requests take days because they are queued behind hard ones.
Messages are answered or assigned within the hour. The inbox owner sees a daily digest: what was handled, what needs their decision, what was extracted as tasks. Routine requests get a drafted reply the same morning; nothing waits because it was invisible.
Trigger, then 8 steps.
Seven of the eight steps run without anyone. Step 07 waits for a person.
Gmail / Outlook trigger (new message)
Receive and de-duplicate
New messages and thread updates; automated notifications and newsletters are recognised by sender patterns and handled by their own rules.
Classify
Intent (reply needed, task, meeting request, FYI, invoice, complaint, spam), urgency, and the internal owner if the inbox is shared. Rules for known senders override the model.
Fetch context
CRM record for the sender, open tickets or deals, prior thread history, and the calendar for meeting requests.
Extract tasks and dates
Any request with a deliverable and a date becomes a structured task with owner and deadline; ambiguous ones are flagged for confirmation.
Draft routine replies
For categories with an approved playbook (pricing questions, scheduling, document requests, status updates) the agent drafts a reply in the owner's voice, grounded in the context fetched.
Route
Draft to the owner for one-click approval; tasks to Asana/Linear/Notion; meeting requests to a scheduling link or calendar hold; invoices to AP; complaints straight to a person with the thread summarised.
Approve and send
Drafts sit in a Slack or Gmail 'ready' label; approval sends from the owner's address. Nothing is sent without a click unless a category has been explicitly set to auto-send.
Daily digest
One message at 8am: handled, waiting for you, extracted tasks, and anything the classifier was unsure about.
Messages are answered or assigned within the hour
Messages are answered or assigned within the hour.
One decision, at step 07. Seven steps ran without you.
Each pattern page explains it in plain words and lists every automation that uses it.
The template is built on n8n (how the n8n AI Agent node works); the same steps run on Make, Zapier or plain code. The tool is chosen to fit your accounts, you own the result, and the template comes in all three formats.
Data it touches
- Mailbox (Gmail, Google Workspace, Microsoft 365)
- CRM contacts and deals
- Calendar
- Task tool (Asana, Linear, Notion, ClickUp)
- Reply playbooks per category
Guardrails
- Drafts require approval by default; auto-send is enabled per category by the inbox owner and logged.
- Complaints, legal, HR and anything from unknown senders with attachments go to a person unread by the agent's reply logic.
- The agent never invents facts about availability, pricing or commitments; it drafts from the context fetched or asks.
- Every automated action is visible in the thread and the digest.
Classification is where the time goes
Reading an email to decide what it is takes almost as long as dealing with it, and it happens for every message. The classifier does that once, reliably, and attaches the decision to the thread — so the person opens an inbox that is already sorted into 'needs me', 'drafted for you', 'done' and 'noise'. That alone recovers an hour a day for most people who run a busy inbox, before any reply is drafted.
Which replies to automate
Start with the ten most common routine requests the inbox receives — you can find them by classifying a month of history — and write a playbook for each: what the reply should contain, what context to pull, what it must never promise. The agent drafts within those playbooks. Everything else gets a summary and a suggested action, not a draft. Expanding the playbooks over time is how the automation rate climbs without the risk climbing with it.
Executive inboxes
For an executive the same workflow runs with a stricter posture: nothing auto-sends, the digest is the primary interface, and the assistant (human or agent) works from the classified queue. The extracted-tasks step is the most valued part — commitments made in email finally land in the task system with a deadline, instead of in memory.
What it connects to.
| Tool | Role in this workflow |
|---|---|
| n8n | Polling, classification pipeline, routing, digest |
| Gmail / Outlook | Mailbox and sending |
| Claude | Classification, extraction, reply drafts |
| Asana / Linear / Notion | Task destination |
| Slack | Approvals and digest |
Want this running without building it?
$197/month
We set it up in your tools, run it on our servershost it on our n8n cloud (or deploy it to your own n8n, Make or Zapier), watch it, and fix it when something changes. Setup included, cancel monthly; the automation stays yours if you leave.
$2,000–4,000 one-time
Your systems, your rules, your edge cases. A one-off build delivered into your accounts to keepon Claude and n8n, delivered into your accounts to keep, with documentation and a recorded walkthrough.
Per inbox, up to 3,000 messages a month. Multiple inboxes or a custom set of reply playbooks beyond the standard ten are a custom build.
Also sold one-time, installed in your own accounts, as part of the executive assistants package and the founders package and the operations managers package and the educators package and the project managers package and the graphic designers package and the freelancers package and the operations teams package — do it yourself $49 · installed for you $249 · done for you $490.
When this becomes a custom build.
The subscription covers the standard version — $197/month, setup and maintenance included. Three things move it to a custom build ($2,000–4,000 one-time), delivered into your accounts to keep. If one of them is you, say so on the audit and the plan prices both.
Several inboxes — a shared support address, an executive’s and a team alias — each with its own rules.
Reply playbooks beyond the standard ten, such as contract-specific answers or regulated disclosures.
Above 3,000 messages a month, or triage that has to create records in a system without a standard API.
Running the standard version first is the usual path: the exceptions it logs in the first month are the scope of the custom build, with your own numbers. How custom builds work.
Frequently asked questions.
Does the agent send emails on its own?
Not by default. Every draft is approved with one click. The inbox owner can enable auto-send for specific low-risk categories once the drafts have proven reliable, and those sends are logged.
Will it work with our CRM?
HubSpot, Salesforce, Pipedrive and Attio are supported for sender context and logging. Without a CRM the workflow still classifies, drafts and extracts tasks.
How does it handle attachments?
Invoices go to accounts payable with the data extracted; documents are summarised for the digest; anything unexpected from an unknown sender is flagged, not opened by the reply logic.
Does an email triage agent send emails on its own?
Not in any version worth running. It classifies, extracts and drafts; a person approves each reply from Slack or the inbox until a playbook — the invoice-copy request, the meeting confirmation — has gone out unedited for weeks. Anything mentioning money, a commitment or a complaint stays a draft by rule.
How accurate is email classification?
High on the categories that matter operationally — reply needed, task, meeting, newsletter, spam — because the categories are few and the model sees the thread and the CRM context. The ambiguous 5–10% are routed to the person with a suggested category, and their corrections tune the classifier weekly.
Which inboxes and tools does it work with?
Gmail and Outlook, shared or personal; HubSpot, Salesforce or Pipedrive for the context; Asana, Notion, Jira or Todoist for the extracted tasks; Slack or Teams for approvals and the digest. Several inboxes with their own rules, or reply playbooks beyond the standard ten, are a custom build.
What this blueprint is built on.
Vendor documentation the steps rely on, and the rules the guardrails are written against.
- Google — Gmail API guides developers.google.com/workspace/gmail/api/guides
- Microsoft Graph — Outlook mail API overview learn.microsoft.com/en-us/graph/api/resources/mail-api-overview
- n8n docs — AI Agent node docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/