Somewhere in the middle of a good session with your assistant, it happens: the next step needs a person. Your cofounder has to weigh in, your accountant has to check the assumption, your friend has to say whether the plan is mad. Now you have to get an hour of context out of your chat window and into their head.
There are three ways to do it. Two of them share the conversation. One of them shares what you meant.
What a share link actually shares
Most AI apps can mint a link to a conversation. For some jobs that's exactly right — you're showing off an output, walking a colleague through a bug reproduction, or preserving an exchange as a record of what happened.
But the link is the conversation: every retry, every tangent, every position you tried on and abandoned. It's all-or-nothing — there's no version of the link that includes the conclusion but not the three wrong drafts before it. And it arrives without an ask. The other person has to excavate what you want from them, which in practice means they skim, guess, and answer a question you didn't ask.
Copy-paste fails in three separate ways
Pasting the transcript into an email inherits all of that and adds problems of its own:
- It doesn't survive the trip. Long sessions get compacted and outputs get capped. The "full conversation" is not reliably a thing you can extract, however much you want to.
- Half of it isn't theirs to read. A session holds half-formed ideas and things you said because nobody was watching. Sharing your thinking shouldn't mean surrendering the room you did it in.
- It hijacks their assistant. If they feed your transcript to their own AI, it inherits your framing, your conclusions, and your blind spots — instead of helping them form their own view.
That third one is subtle and matters most when the point of sharing was to get an independent opinion.
When a transcript is the right thing
A fair rule: share a transcript to show what happened; send a briefing to get input on what to do.
Transcripts are right when the artifact is the point — a striking output, a support repro, a record of exactly what was said in what order. If nobody needs to answer anything, a share link is fine, and it's less work.
The moment you need something back — a decision, an opinion, a review — the transcript starts working against you.
The briefing: send the part that needs an answer
A briefing is what RelayLink sends instead. Your assistant composes it from the session; the session itself stays home. What crosses:
- The ask, with a response shape ("decision", "opinion", "review") and an honest urgency.
- A TL;DR and a context brief — 150–400 words written for someone with zero context.
- Assumptions, each labeled stated by sender or inferred by sender's AI.
- Options considered, and what's already been ruled in or out.
- Open questions, so they know what's genuinely unsettled.
- Your note, in your own words, with a provenance label that says exactly whose words they are.
You review the rendered package — precisely what the recipient will see — and confirm. They get a normal email with your note, the TL;DR, and a link to the full briefing. No account, no app. Their reply, typed anywhere, lands back on the same thread.
The practical difference shows up in the reply. Answers to transcripts are vague because the question was; answers to briefings tend to come back in the shape you asked for.
If it's specifically a shared chat link you were about to send, RelayLink vs. sharing a ChatGPT link compares the two side by side.
If you've got a real question stuck inside a chat session right now, that's the ideal first run: connect your assistant and have it draft the briefing. Sending it takes less time than reformatting the transcript would have.