The AI-Human Collaboration Framework

AI Slop is a Growing Problem in the Workplace

I read an insightful article recently about the problems that can arise when co-workers don't collaborate with AI responsibly. It discusses times when someone asks a co-worker for help and they get back AI slop. AI slop is an output from AI that doesn't add value and wastes your time.

This is becoming a big problem for organizations. Leadership often pressures employees to use AI and move fast. This pressure results in workers using tools they believe can improve their efficiency.

I've seen complaints about co-workers using ChatGPT for everything. For example, they'll paste a long essay into Slack when a single sentence was all that was needed. Instead, they received copied and pasted 1,500-word essays that lack understanding of the situation at hand.

Receiving an AI generated reply instead of a hand-crafted reply sends a dismissive message: 'I didn't want to think about your problem myself.' This type of behavior can make colleagues feel devalued and damage team morale.

These kinds of interactions end up hurting work relationships because they can feel disrespectful to the person on the receiving end. Did the co-workers who used ChatGPT not care enough to put their own thoughts into the matter? Do they not care for the quality of the work they are sharing with their colleagues? This can erode trust among team members.

Organizations are drowning in AI slop. This makes it harder to find the signal in the noise of all this newly generated content and data.

A Framework to Use AI Responsibly

To solve this, I created a framework. It's designed to help teams use AI responsibly. The goal is to reduce AI slop, improve quality, and show respect for our co-workers.

The framework is a decision matrix that helps you choose the right way to use AI based on two factors: task complexity and the level of strategic thinking required.

1. Automate & Monitor

• When to use: For tasks with Low Importance and Low Complexity. These are repetitive, low-risk activities.
• AI's Role: Executor. The AI does the work.
• Human's Role: Monitor. Set the rules and periodically check the results.
• Examples: Categorizing expenses, scheduling internal meetings, organizing files.

2. AI-Assist & Human-Verify

• When to use: For tasks with High Importance but Low Complexity. The cost of an error is high, but the task itself is straightforward.
• AI's Role: Drafter. Creates the first version.
• Human's Role: Approver. Edits, fact-checks, and gives final sign-off.
• Examples: Writing social media posts, summarizing meeting notes, drafting standard customer service replies.

3. Ideate & Experiment

• When to use: For tasks with Low Importance but High Complexity. This is a low-risk creative sandbox.
• AI's Role: Brainstorming Partner. Generates ideas and possibilities.
• Human's Role: Curator. Selects the best ideas for further development.
• Examples: Brainstorming blog post topics, generating name ideas for an internal project, creating draft visuals for mockups.

4. Human-Led & AI-Enhanced

• When to use: For tasks with High Importance and High Complexity. These require deep strategic thought, empathy, and accountability.
• AI's Role: Research Assistant. Provides data, summarizes information, and acts as a sounding board.
• Human's Role: Strategist & Creator. Owns the thinking, makes the final decisions, and is fully responsible for the outcome.
• Examples: Developing a marketing plan, setting business strategy, writing a major grant proposal, handling a sensitive client issue.

How to Align AI to the Framework?

But what good is a framework if AI can't be aligned to it? That's why I developed and am releasing a system prompt to control your AI's behavior. It will review user requests and decide how best to respond given the framework. This could be really powerful and prevent you and others from using AI inappropriately on your team.

There's a bonus instruction to limit the AI's verbosity as much as possible, but prioritize clear responses. This should help cut down on unnecessarily long outputs.

How to Introduce This to Your Team

1. Try using it yourself so you form your own opinions about what works and what doesn't. Download the system prompt and add it to your AI chat of choice.
2. Discuss the framework and the original Descript article with your team. What resonates with them? How do they feel about AI slop? What do they think about the framework as a solution?
3. Modify the system prompt together to fit your team's specific needs.

Future of AI-Human Collaboration -> AI-Team Collaboration

I think we're at the point where we need to think deeply about how AI is used within teams more so than at an individual level. After all, that's how most work is done in the office: by collaborating with others.

I welcome feedback if you try this out. You are also free to modify the system prompt to fit your needs, since it's licensed under Creative Commons BY 4.0.

CC by 4.0 Conditions

• Attribution: You must give appropriate credit to the original author, provide a link to the license, and indicate if you made changes.
• Indicate changes: You must indicate if you have modified or built upon the material.

Work output transparency: This framework was created by AI (Gemini 2.5 Pro) inspired by the article written by the Descript team and reviewed by myself. The image was created manually by myself using Figma. The system prompt was created by AI (Gemini 2.5 Pro) using the framework as input, with minor changes by myself. This blog article was written manually by myself with grammar edits by AI (Gemini 2.5 Pro).