The unanswered mail in an expert's inbox is not a pile of things they refused. It's a pile of things they couldn't finish in the gap they had. The messages that get replies fit inside a few minutes of expertise they were already carrying. The rest aren't harder questions — they're questions that arrived without their framing, and framing is the expensive part.
This isn't the advisor case, where a relationship exists and someone has already agreed to be available to you. This is someone who owes you nothing and will never owe you an explanation for silence.
Their scarce resource is attention, not knowledge
What you want from an expert is cheap for them to hold and expensive to spend. They know the answer whether or not you ask. What costs them is loading your situation, working out what you actually want, deciding what level to pitch a reply at, and then writing it.
Most requests fail on that middle stretch. They arrive with facts but no shape, so the recipient has to do the framing before they can start answering — work with no defined end. Unbounded work gets deferred rather than declined, which is why good requests die without ever being refused. And every second you save by not framing gets spent by them at worse odds, since they know less about your situation than you do. That's the attention arithmetic applied to a stranger.
Show your work so they correct instead of construct
Correcting is cheap. Constructing is expensive. An expert reading "here's what I think and why" can answer in a line: your third assumption is wrong. An expert reading "what should I do about X" has to build the whole answer from an empty page.
So arrive with a position — what you tried, what you concluded, which option you already rejected and why. Then name the assumption you're least sure of, explicitly, because that's the one you want shot at. Stating an assumption plainly is what turns a vague worry into something a specialist can knock down in seconds.
Being visibly wrong is the price of this move, and it's a bargain.
Ask one question, not a list
Five questions aren't five times the work. They're worse, because the recipient now has to decide which ones to answer and whether a partial reply reads as rude. Faced with that, most people defer, and deferred means gone.
Pick the question whose answer changes what you do next. The others can be a second message if there is one. If you can't identify which one that is, you aren't ready to send.
"Can I pick your brain" is the degenerate case: zero questions, and it asks the recipient to generate the question, the scope and the format, usually with a calendar request smuggled in. The same rule governs first contact generally, in outreach that isn't spam.
Make it answerable from expertise, not investigation
The line is whether they could answer without opening anything. "Does this approach have a failure mode I'm not seeing" is answerable from expertise. "Could you look through my repo and tell me what's wrong" is not — that's consulting, and you should either pay for it or not ask.
Anything that requires them to read a long attachment, run something, or look up a fact you could have looked up yourself has crossed that line. So do the lookups first. If the answer sits in something they already published, finding it there and asking the next question is the entire difference between a request that flatters and one that works.
Give them an easy exit
Give them a way to decline that costs nothing and needs no explanation. Say directly that no reply is a complete answer. Otherwise declining requires composing a decline, and a small tax on saying no is a large tax on replying at all.
Then say what you'll do without them — "I'll go with the second option if I don't hear by Friday" — which turns silence from an awkward loose end into a real choice with a known consequence. It also forces honesty about the deadline, and setting an honest one is harder than it looks.
A thorough-looking request is now free to produce
An assistant will turn a rough situation into a polished, well-structured request in seconds. It cannot make that request faster to read. So the bar goes up, not down: a message that cost you nothing and costs them ten minutes is a worse trade than the scruffy paragraph you'd have typed yourself. That gap between cheap writing and unchanged reading is the whole of respecting the reader's time.
There's a sharper failure underneath it. An assistant will confidently fill the gaps you couldn't — a fluent account of your situation that includes the parts you were guessing at. The expert can't tell which parts you vouch for, so they either correct something you never claimed or trust something you invented. The second one spends their credibility, not just their time.
The rule is tedious and it works: read the draft as the recipient. Anything that isn't a fact you'd defend, cut it or mark it as a guess. That's why provenance is a field rather than a tone you strike.
What a briefing pins to a field
A RelayLink briefing turns three of these moves into structure rather than good intentions. Assumptions each carry a tag — stated by the sender, or inferred by the sender's AI — so the expert sees which claims you personally stand behind. Options you considered carry a status of chosen, leaning, open or rejected with a reason attached, so "I rejected this, here's why" is a field rather than a paragraph you hope they reach. And the ask is one field, capped at 400 characters.
Two limits worth stating. Open questions are a separate list permitting up to ten — the cap is generous, it is not a target, and the argument above still says use one. And the stated-by-sender tag is self-reported by the drafting model, not verified by the server; it marks where to push rather than guaranteeing anything. The note is the one field with a real gate behind it.
None of this makes an expert owe you an answer. It removes the failure where the answer was cheap and the framing wasn't. If you'd rather your ask arrive as a field than a hope, connect your assistant.