What to do when you receive an AI briefing

An AI briefing lands in your inbox as a normal email — a note, a TL;DR, and a link, nothing to install. Here's what to check before you answer, and how to reply either way.

4 min read

An email shows up. The subject line reads like a person wrote it. Below the fold there's a short note, a two-line summary, and a link — nothing asking you to install anything or create an account. That's what it looks like on the receiving end when someone's AI assistant sends a RelayLink briefing your way. If this is your first one, here's what's actually in it, and what you're supposed to do with it.

What actually arrived

The email carries three things: the sender's note, a TL;DR, and a link to the full briefing in a web view. The note is short — the part the sender personally wrote or approved. The TL;DR is the ten-second version of what they want.

Follow the link and you get the whole package: the ask itself, with a stated response shape — a decision, an opinion, a review — and an honest urgency level; a context brief written for someone with no history on the topic; the assumptions behind it; options already considered; and a reply box. None of this requires an account. Reading it doesn't sign you up for anything or connect anything to your inbox.

Check the provenance labels before you check the content

Before you weigh what's being asked, look at who's actually saying it. Every field in a briefing is labeled with where it came from — the detail that makes an AI-mediated message checkable in a way a normal email, where you just trust the byline, isn't.

The sender's note carries one of two labels if a human stood behind it: verbatim, human-authored means they typed those exact words themselves, different from whatever their assistant first drafted. AI-drafted, approved unchanged means the model wrote it and the sender signed off without changing it — still something they vouch for, just not their own sentence construction. If the note carries neither label, and the rest of the package leans on assumptions marked inferred by sender's AI, you're reading the assistant's best guess about what its owner wants, not a considered statement from the person. Provenance labels exist so that distinction is something you can check, not something you have to guess at.

Find the ask before you read anything else

A briefing opens by stating what's actually being requested — a decision, an opinion, or a review — plus how urgent the sender says it is. Read that line first. It tells you the shape of a sufficient reply before you've spent five minutes on the context brief, so you're not left reconstructing "so what do they actually want from me" out of three paragraphs.

If the stated urgency doesn't match what the context brief actually contains, that mismatch is worth noticing too, before you decide how quickly to respond.

How to reply

Two ways, and both are equally valid. Type directly into the reply box in the web view, or just hit reply in your email client the normal way. Either lands your words back on the sender's thread, labeled typed via magic link so they know exactly how your reply arrived. There's no wrong choice here — use whichever is already open.

Bring your own assistant into it, if you have one

The sender worked their side out privately with their AI before any of this reached you. You can do the same thing in reverse: if you have an assistant connected to RelayLink, point it at the briefing and think through your answer with it before you type a reply — the same private-session pattern, just running on your side this time. Neither assistant gets to answer as you. Both get to help you think.

What you don't owe anyone

No obligation to respond. Opening the email or the web view doesn't create an account or start anything on your end. Ignoring the message costs you nothing.

None of this tells you whether the ask itself is worth your time — that's a judgment call the format can present cleanly but can't make for you. If the message came from someone you've never corresponded with, that judgment call has its own checklist. If it came from someone you already know, you likely have everything you need on the page in front of you already.

If you end up wanting the same thing for your own outgoing correspondence — a reviewed note, clear provenance, a reply that lands back on the same thread — connect your assistant.

Frequently asked questions

Do I need an account to read or reply to an AI briefing?
No. The email and the linked web view are both readable without signing up for anything, and replying, whether you type into the web view or just hit reply in your email client, also requires no account. Either way, your words land back on the sender's thread.
How do I know which parts of the message are the sender's own words?
Every part of a briefing carries a provenance label. The sender's note is marked verbatim, human-authored only if they typed it themselves and changed it from their assistant's draft, or AI-drafted, approved unchanged if they signed off on the model's wording without changing it. Assumptions in the context brief are separately marked stated by sender or inferred by sender's AI, so you can tell a stated fact from a guess at a glance.
What happens if I just ignore the message?
Nothing happens on your end. No account is created by opening the email or the web view, there is no obligation to respond, and ignoring it costs you nothing. If you would rather not hear from that sender again, every message includes a one-click block that stops future delivery and does not require an account either.