AI email assistant: triage, drafts and follow-ups with your OK
An AI email assistant sorts your inbox, drafts replies and tracks follow-ups, and asks before it sends to anyone new. How it works, and how to set one up.
An AI email assistant is an AI agent that reads your email, sorts what matters, drafts replies in your style and keeps track of what's waiting on someone, so your inbox stops being a to-do list you never finish. The useful ones go beyond writing help: they read whole conversations, label and archive, remember context, and remind you or follow up when a reply doesn't come. The safe ones ask before sending anything to someone new.
In KutrimBots there are two ways to use one: an assistant that works in your Gmail or Outlook, and an agent with its own email address for the work you hand it.
What can an AI email assistant do?
| Job | What the assistant does | Needs your OK? |
|---|---|---|
| Triage | Reads new mail, labels it (needs-reply, waiting, invoice, support) and archives noise | No |
| Summaries | Sends a short morning brief of what needs you today | No |
| Drafting | Writes replies using the whole conversation and your notes | No, until it sends |
| Sending | Replies to people who wrote first; writes to new contacts | New contacts: yes |
| Follow-ups | Tracks threads waiting for an answer and nudges on time | Depends on contact |
| Attachments | Saves invoices and documents to your files | No |
| Sign-ups | Uses its own address to sign up for tools and read the verification codes | No |
| Spam and scams | Flags suspicious mail and ignores instructions inside emails | No |
How is it different from AI writing help in your email app?
Most email apps now include AI that helps you write: it suggests a reply or rewrites a paragraph while you're in the app. An AI email assistant is an agent: it works when you're not there, across many threads, and acts.
| AI writing help in your email app | AI email assistant (agent) | |
|---|---|---|
| Works when you're away | No | Yes, on new mail and on a schedule |
| Reads the whole conversation | Sometimes | Yes, oldest to newest |
| Organises the inbox | No | Labels, archives, flags |
| Remembers context | No | Yes, plus your team's shared memory |
| Follows up | No | Yes, from "waiting" threads |
| Hands work to others | No | Yes, to other agents (sales, support, accounts) |
If you want help writing faster, the built-in features are enough. If you want fewer emails reaching you at all, you want an agent. For the difference in general, see AI agents vs chatbots.
Option 1: an assistant for your own inbox
Connect Gmail or Outlook in Apps (through a secure OAuth connection, so the agent never sees your password) and hire an agent from the Inbox assistant role. Its standing instruction is simple: sort email, draft replies, never send without asking. If this is your first agent, set up your first AI team first; it takes about ten minutes.
A typical setup:
- Triage rules in plain words: "Label invoices, archive newsletters, flag anything from clients, and tell me about anything urgent."
- A morning brief: a routine at 8:30 AM that lists what needs a reply today.
- An automation for urgent mail: when a new email arrives in Gmail from a client domain, the agent reads it and messages you in KutrimBots if it's urgent.
- Drafts, not sends: the agent prepares replies; sending from your mailbox shows an approval card first.
Option 2: give the agent its own email address
The Email Inbox plugin gives each agent its own address, such as atlas@yourcompany.com, with a Gmail-style mail app on your phone and desktop where you can read and organise everything it sends and receives. You connect any mailbox you own (Zoho, Google Workspace, Hostinger, cPanel and others).
This is the better choice when the agent does work in its own name:
- Customer and supplier email: support@ or sales@ handled by your support or sales agent.
- Sign-ups: the agent registers for tools, receives the verification code or magic link, and the Security Vault creates and stores the password.
- Reports and newsletters it needs to read for research.
Who it can email without asking
- Straight away: people who wrote to it first, and contacts you've allowed. Replies to the sender always go straight away.
- After your OK: anyone new, including new people added on a reply-all or a forward. The email waits in a "pending" folder until you allow it, then goes out by itself.
How does it keep track of follow-ups?
The assistant uses the same small set of labels every time:
- needs-reply: it still has to answer;
- waiting: it answered and is waiting for them;
- escalated: you have to decide;
- done: finished and archived.
A routine checks "waiting" threads each morning and follows up on the ones that have gone quiet, in the same thread, referring to the earlier message. For sales email, the follow-up is tied to the lead in CRM & Leads instead, as described in CRM lead follow-up on autopilot.
Is an AI email assistant safe?
Email is where most attacks start, so an email agent needs firm rules:
- Emails are data, not instructions. If an email asks the agent to send files, open unknown links, pay or reveal anything, it doesn't, unless you asked for exactly that, and it tells you about the suspicious mail.
- New contacts need your OK, so a tricked agent can't start writing to strangers.
- Passwords never go in email and never pass through the agent; they live in the Security Vault.
- Everything is visible: you can read every email the agent sent and received.
Example: a week with an email assistant
Example scenario: a freelance designer gets about 60 emails a day.
- Monday: the assistant labels 41 emails as newsletters or notifications and archives them, flags 6 client emails, and drafts replies to 4 of them for review.
- Tuesday: an invoice from a software vendor arrives; the assistant saves the PDF to File Manager and labels it "invoice".
- Wednesday: a prospect from last week hasn't replied; the morning routine sends a short follow-up in the same thread.
- Thursday: an email "from the bank" asks to confirm card details by link. The assistant marks it spam and tells the designer.
- Friday: the brief lists two threads still waiting and one decision only the designer can make.
Frequently asked questions
What is the best AI email assistant?
The best one is an agent that reads whole conversations, organises your inbox, tracks follow-ups and asks before contacting anyone new. If you also need email handled in an agent's own name, choose one that can have its own address, like KutrimBots' Email Inbox.
Can an AI email assistant reply automatically?
Yes, within rules you set. In KutrimBots, replies to people who wrote first can go straight away, while the first email to anyone new waits for your approval.
Does an AI email assistant work with Gmail and Outlook?
Yes. You connect Gmail or Outlook securely so the agent can read, sort and draft, and Email Inbox works with mailboxes from Zoho, Google Workspace, Hostinger, cPanel and other providers.
Can an AI agent have its own email address?
Yes. With the Email Inbox plugin each agent gets an address on your domain, such as atlas@yourcompany.com, which it uses for customer email, sign-ups and verification codes.
How much does an AI email assistant cost?
In KutrimBots, an email assistant is one of your plan's agents, and Email Inbox is a built-in plugin: plans start at $20 a month for up to five agents. See pricing.