How to spot workslop and stop forwarding AI work you haven't read

Someone runs a task through AI, glances at what comes back, decides it is good enough, and sends it on. The person who receives it can tell within a minute that nobody read it. Both halves of that exchange now have names: the person forwarding the draft is a meat proxy, and what lands on the next desk is workslop.

In Episode 112 of The Biggest Goal Podcast, Alane Boyd and Micah Johnson admit they hate both terms and think both are accurate. They cover what each one costs, how to recognize it, and four changes that stop it, and both hosts confess to doing it themselves.

Two names for one bad handoff

A meat proxy takes what the AI produced and pushes it down the line without thinking about it. Workslop is what they push: the work version of the AI slop everyone scrolls past online. It rarely stops at one desk. Sales sends something to marketing, marketing builds on it, and a careless draft becomes careless execution across departments.

Micah admits he has looked at a draft, called it good enough, and passed it along. Alane once sent her team a set of her ideas that AI had pulled together and asked them to take it from there. Micah points out that she told the team she had worked with AI and that the ideas still needed refining, which puts her case in a gray area each company has to settle for itself.

"If you're listening to this and you're a leader, be aware that you could be the one doing this, not just your team or somebody on your team. I've seen it both ways. I've been guilty of it," Micah says.

What workslop costs the person who sends it

The hosts share how people react when they receive AI work the sender clearly never read:

• 53% feel annoyed and keep it to themselves.
• 42% see the sender as less trustworthy.
• About half see that colleague or leader as less creative, less capable, and less reliable.
• About a third do not want to work with that person again.

Nobody tells the sender. The recipient inherits the problem instead. "Now I have two choices. One, to send it back to you and have to manage you and tell you, hey, please update this because it's crap. Or two, the work now falls on me," Alane says. A junior team member does not even get the first option, since managing up is not realistic for them, so they are stuck with the work.

Knowing the outcome separates good AI use from workslop

Alane uses a legal plugin to draft contracts and terms, and she is not a lawyer. Fifteen years of owning a company tells her when a clause is wrong. "I know what I'm looking for. I know what I'm trying to put together," she says. She changes the terms she does not like, then sends the draft to a legal expert.

Micah runs the same scenario with an intern: same contract, same plugin, same model, very little context. Alane's version might reach 80 or 90 percent before the lawyers see it. The intern's might reach 40 or 50 percent, and a nervous intern may just go with whatever Claude says. Delegating still works if the draft finishes in the right hands. "There's different levels internally, but no matter what, it always ends in an expert," Micah says.

Most meat proxies are not lazy. As Alane describes it, the boss asked for something, the person does not know what outcome the boss wants, so they have AI execute it and never check, because they would not know what to check for. Micah ties it to an old management rule: give someone a task and they will do it, but give them the reason behind it and everything changes, for people and for people using AI.

The tells that give workslop away

• It does not match reality. A domain expert sees it immediately.
• It is long for no reason. Verbose writing that covers every corner means nobody decided what mattered. "Why are we talking about these eight other things? We only need to talk about these two things," Micah says.
• It sounds like AI. The structure and phrasing everyone now recognizes on LinkedIn.
• It ignores what the company already knows. Marketing that skips the brand guidelines or fails to promote what is currently running.

One host describes a former assistant who ran the team's call debrief skill, which has a deliberate human-in-the-loop step, and handed over the drafted follow-up email. Three items in it were things the team had decided against later in the call, and reading the email once would have caught them.

Four ways to fix it

1. Train the team, including on culture. A license and a feature tour beats nothing, but real training also covers what the right way to use AI looks like at your company and whether workslop is acceptable there. Alane describes their own team sharing use cases in a Slack channel and meeting by department to learn from each other. Micah adds training on a single source of truth that everyone works from.

2. Make AI use visible. If using AI is shameful, nobody admits it, and the reviewer loses the context that would change how they read the work. Micah wants AI treated as a tool, not a partner or a co-worker: Claude does in five minutes what would take an hour or two, and part of the saved time goes to checking it. "I'm going to spend 10 to 15 or 20 minutes getting it right. That's not an error in the process. That's not an issue with the system. That is the freaking feature of the whole thing," he says. He also wants the process itself to define checkpoints for review and validation.

3. Reward quality, not volume. Building a skill takes extra time up front before it saves hours, so the first weeks may not feel faster. "Generally speaking, putting out volume means you're shortcutting something. And AI is a beautifully easy shortcut to take," Micah says. The better measure is reaching the finish line sooner on the same number of daily tasks.

4. Require a human-written summary. Asking for a person's own written take on the AI draft means they have to read it. If it is too long to summarize, they work with AI to cut it down until they understand it. One team member already works this way: she uses AI to draft speaking proposals, then comments in the document wherever she wants input, so the draft has been read and edited before anyone else sees it.

Alane closes with the book Radical Candor. "At the end of the day, you have to have radical candor. You have to say, this is workslop," she says. Telling someone it is not okay to forward unread AI work is still part of managing people.

Start with your own work

Every fix comes back to one habit: know the outcome you want before you ask AI for it, and read what comes back before it leaves your hands. You can build that habit on your own tasks, where you already know what good looks like. Start with the free Claude Cowork Masterclass. It is self-paced, so you can work through it on your own job and check every draft against what you know.

Someone on your team ran a task through AI and forwarded the output without reviewing it. That person was a Meat Proxy and just generated Workslop. Yuck!

Alane Boyd and Micah Johnson break down these two terms, which sound like they should be in food manufacturing, and describe exactly what can go wrong inside AI-forward companies.

In this episode, you’ll learn:

• How to tell a meat proxy from someone using AI well and why knowing the desired outcome is the whole difference
• What your colleagues actually think when they spot unread AI output, backed by the numbers on trust and credibility
• The tells that give workslop away, from verbose answers to ignored brand guidelines
• Why “train your team” isn’t enough unless you define what the right way to use AI looks like at your company
• How making AI use visible turns a shameful shortcut into a useful handoff
• The one requirement that forces people to read the output before it leaves their desk

If your team is moving fast with AI and you’ve started to notice the quality slipping, this conversation names the problem and gives you several things to change today.