What is a meat proxy, and how do you avoid becoming one?
A meat proxy pastes AI output at colleagues without adding a thing. Here's how nonprofit teams use AI to make messages shorter before they hit send.
What is a meat proxy?
A meat proxy is a person who pastes AI output into an email, a chat or a document and passes it along without adding anything of their own. They act as a go-between for the machine, and the only human part of the exchange is the person, the meat, doing the copying and pasting.
I didn't coin the term. It hit the zeitgeist in late summer, and when I shared it with Colleen in September, it named something every inbox was already feeling. Here's how I described it to her:
Scott Williams · Recorded conversation, Sep 10, 2026This is a person who pastes AI output into Slack, anywhere, shoots an email, whatever, without adding anything to it.
You've probably received one. A colleague asks AI a question, gets back a long answer and forwards it with a line like "Here's what ChatGPT said." The reader now has a pile of text, no point of view and no idea which part the sender actually believes.
Why is everyone drowning in AI-written email?
Because AI flipped the shape of communication. Before AI, you started with a pile of ideas in your head, did the hard thinking yourself and sent one clear note. Now it's easy to start with a few words, skip the thinking, and send a wall of text that someone else has to sort through.
I walked Colleen through it with two drawings. The old shape was wide at the top and tiny at the bottom. The new one is upside down.
Scott Williams · Recorded conversation, Sep 10, 2026AI showed up and the shape flipped because now what happened is, is that you have six words, you skip your brain, throw it into AI. Then you go 600 words.
The cost lands on the reader. Colleen feels it every day. When we talked about our first year of running If Possible, this is what she said she was most tired of:
Colleen Cook · Recorded conversation, Sep 3, 2026you feel like the one thing you're most tired of at the end of the day is reading. Just feel like I'm reading so many words every day.
I told her that in the age of AI, being a strong reader has become an incredible skill, because the content keeps coming whether you asked for it or not.
What does it feel like to receive meat-proxy messages?
It feels like homework. The sender saved time, and you inherited it. You have to read the whole thing, guess which parts matter, figure out what the sender actually thinks and then decide what to do, all without the context they had.
I've been on the sending end of this. Colleen reminded me that last year I sent her an idea as a five-page ChatGPT output, and she was annoyed. She'd much rather have spent 25 minutes just talking about it with me. She was right, and it's the feeling I was describing when I said this:
Scott Williams · Recorded conversation, Sep 10, 2026Can you just get to the point, you know, like, don't give me more homework to do.
Inside a nonprofit, this gets expensive fast. Your program director's attention is limited. Your board chair reads email on a phone between meetings. Your volunteers give you an hour a week. Every bloated message spends their time on your thinking.
How do you use AI to make a message shorter?
Do the thinking with AI, then ask it to shrink the result before anyone else sees it. Tell it who the reader is, how much time they have and what they need to do next. The goal is one sharp note that holds everything that matters and nothing else.
Here's where the idea started for me. I spent about an hour in one long chat with Claude, working through a frustrating problem with one of our recording tools. When I finally had it solved, I still had to tell Colleen what happened, and I dreaded it. I call that the communication tax. So I asked Claude to tell her briefly, without talking her ear off. I wanted her to get it in about 20 seconds. It drafted the note, I changed two things and it was done. I paid the strategy tax, figuring out the problem, and skipped the communication tax entirely.
good AI compresses before the send, you know, that you spend time whittling it down. Now, this is the five things you need to know instead of the 50,000 that I could tell you.
Copy this prompt into the end of any long AI conversation:
``` We've talked about this for a while. Now write a message to [name, role] that they can understand in 20 seconds. Lead with what they need to know or do. Include only what changes their next step. Keep it under 80 words, in my voice. ```
Why is shorter communication an act of empathy?
Because a short message means you did the work so the other person doesn't have to. When you whittle a message down, you're thinking about the person on the other end, their time and their attention. A long message dumped on someone does the opposite.
I said this to Colleen in the same conversation. It's really about empathy, about thinking of the person receiving it. I reached for the old line attributed to Einstein about making everything as simple as possible, but no simpler. And some things deserve no AI at all. Write your mom's birthday card yourself.
This connects to how we think about AI slop. When a person stops deciding and lets AI do the talking, you get generic content in public and meat-proxy messages in private. Both come from the same place. If that's familiar, read AI slop happens when AI sits in the director's chair.
What should a nonprofit team agree on about AI-written messages?
Agree on a few simple norms and write them down. Nobody forwards raw AI output. Every message leads with the sender's own point. Long AI work gets summarized before it's shared. These take five minutes to adopt and save hours of reading every week.
Use this checklist before you send anything AI helped you write:
- Did I add my own point? If the message has no sentence only you could write, it isn't ready.
- Can the reader get it in 20 seconds? If not, ask AI to cut it in half and try again.
- Is the next step obvious? Put what you need from them in the first two lines.
- Did I cut what I can't stand behind? Delete anything you didn't check or don't believe.
- Would a link or attachment do better? Share the full AI output only when someone asks for it.
Norms like these belong next to your first AI policy, and they pair well with batching your communications so your team writes fewer, better messages. You'll find more in our strategy articles.
For one practical idea each week, sent in a short note, join the If Possible newsletter. If you want your whole team using AI to work smarter together, save a spot on the cohort waitlist.
Frequently asked questions
What is a meat proxy?
A meat proxy is someone who copies AI output into an email, chat or document and sends it on without adding their own thinking. The term describes a person acting as a pass-through for the machine, leaving the reader to sort out what matters.
Why is AI-written email so long?
AI expands short requests into long answers by default. When people send that output without trimming it, a few words of thought become hundreds of words for someone else to read. Asking AI to compress before you send fixes most of it.
How do I stop being a meat proxy?
Add your own point, cut the message to what the reader needs and ask AI to shrink it before you send. If the reader can't understand it in about 20 seconds, it isn't ready yet.
Should our nonprofit have rules about AI-written messages?
Yes, a few simple ones. Agree that nobody forwards raw AI output, that every message leads with the sender's own point and that long AI work gets summarized first. Add them to your AI policy so new staff learn them on day one.
Know an executive director who needs this?
