How long does setup actually take?

About 5 minutes for the connection, plus 60-90 seconds for the initial tone analysis. You can start drafting replies immediately. The 'feels like you' tone settles in over the first week.

Does it work on the Gmail/Outlook mobile apps?

Yes — because drafts appear in your normal drafts folder, you'll see them in any client that accesses that inbox (mobile Gmail, Outlook on iOS/Android, third-party clients). No separate app needed.

Can I disconnect at any time?

Yes. Disconnect in your account settings or directly via Google's OAuth permissions page (or Microsoft's equivalent). Once disconnected, ReadandReply loses all access to your inbox and your tone data is deleted within 30 days.

What if my organisation uses Microsoft 365 with strict admin policies?

ReadandReply offers an admin-approval flow for 365 tenants — your IT admin can pre-approve the app for specific users or groups. We have documentation you can share.

How to Set Up an AI Email Assistant (Gmail & Outlook)

Setting up an AI email assistant takes about 5 minutes. The hard part isn't the setup — it's deciding what permissions to grant, how to think about privacy, and what to actually do in the first week to get drafts that feel like you. Here's a straightforward walkthrough for both Gmail and Outlook.

Before you start

You'll need two things: an active Gmail or Outlook account (workspace or personal, both work), and 5 minutes. There's no software to install — ReadandReply runs entirely on the email account you connect, and drafts appear in your normal drafts folder.

Step 1 — Sign up and choose a plan

Sign up at readandreply.com. Every plan starts with a 14-day free trial, no credit card required. If you're not sure which plan, start on Starter — you can upgrade later without losing tone learning.

Step 2 — Connect your inbox

For Gmail, you'll be redirected to Google's OAuth consent screen. For Outlook, it's Microsoft's equivalent. In both cases the system will ask you to grant ReadandReply specific permissions:

Read your emails — required so the AI can understand the email it's replying to and learn your tone from your sent items.
Create drafts in your drafts folder — required so the assistant can write the suggested replies.
Send emails on your behalf — only required if you want to enable auto-send (off by default). If you leave auto-send off, you can decline this permission.

These are the standard OAuth permissions for any AI email tool. The permissions are scoped to your individual account — not your whole organisation's inbox.

What if I don't want to grant read access?

Without read access, the AI has nothing to read and nothing to learn from, so it can't function. This is the unavoidable trade-off of any AI email assistant. If reading your email is a hard limit for your situation (highly sensitive role, regulated industry without exemption), an AI email assistant isn't the right tool.

Step 3 — Tone calibration (the important bit)

Once connected, ReadandReply analyses your last 500-1,000 sent emails to learn your tone. This takes about 60-90 seconds. You don't need to do anything during this step except wait.

After the initial analysis, you'll see a tone summary — sign-offs you typically use, formality level, average sentence length, the kinds of openings you favour. If anything looks wrong (maybe you've recently changed your tone deliberately, or you have a 'work voice' and a 'personal voice' that the AI has averaged together), you can correct it in the dashboard.

Step 4 — Configure auto-send (recommended: off, at first)

Auto-send is the feature that lets ReadandReply send drafts without you reviewing them. It's off by default and we strongly recommend leaving it off for the first month. The reason is simple: you need to build trust in the drafts before you let them go out unsupervised.

When you're ready, you can enable auto-send for specific categories (e.g. meeting confirmations, where the worst case is a small typo) rather than enabling it globally. Most users never enable global auto-send — and that's fine. The time savings come from drafting, not from sending.

Step 5 — The first-week tuning

For the first week, do three things:

Edit drafts before sending — not because they're wrong, but because every edit teaches the AI a preference. Two weeks of editing typically removes 90% of the small tone mismatches.
Use the 'this is wrong' button — if a draft is genuinely off (wrong context, missed the point), flag it explicitly. The AI learns faster from explicit corrections than from silent edits.
Don't expect perfection on edge cases — the first time you handle a really unusual situation (an angry escalation, a sensitive personal note), the AI will probably draft something safe rather than something brilliant. That's fine — write those yourself.

The single biggest predictor of success with an AI email assistant is using it actively in week one. Users who keep editing drafts and providing feedback land on a 'feels like me' baseline within 7-10 days. Users who set it up and then ignore it never get there.

What about Outlook specifically?

Setup is identical to Gmail apart from the consent screen. ReadandReply supports both Outlook personal accounts (Hotmail, Outlook.com) and Microsoft 365 work/school accounts. For 365 admin-managed accounts, your IT admin may need to approve the app — we have a one-page document you can share with them.

Common first-week issues

'The drafts don't sound like me yet.' Give it a week. Tone learning compounds — by day 7 the difference is usually dramatic.
'I'm getting drafts for emails I don't want drafted.' Use the per-category filters in settings to exclude sensitive labels, specific senders, or particular threads.
'Drafts are referencing wrong context from old threads.' Increase the context window in settings, or link your CRM so the AI gets the right relationship history.

Once setup is done and tone is dialled in, the experience becomes invisible — drafts just appear and you skim-and-send. Most users forget the tool is running by week three. Want to try it? Start a 14-day free trial.