How to avoid AI workslop

Work slop is AI-generated work that looks polished but does not meaningfully advance the task. It includes reports with invented statistics, meeting summaries that miss the decision, and slide decks full of confident language but little substance.

Researchers from Stanford Social Media Lab and BetterUp Labs coined the term in their workslop research. They found that 40% of U.S. desk workers had received it from a coworker. Each instance took nearly two hours to untangle, with an estimated cost of $9 million a year for a 10,000-person company. Another 32% of recipients said it made them less likely to work with the sender again.

Work slop is eroding quality and trust simultaneously. But the root cause isn't lazy employees or bad AI. It's a fundamental misunderstanding of what AI actually does well.

AI is not a database

A different study from Searchlight Institute asked Americans what happens when you submit a question to ChatGPT. The results should alarm any business leader investing in AI adoption:

“[We asked employees] what actually happens when you ask a tool like ChatGPT a question. 45% think it looks up an exact answer in a database, and 21% think it follows a script of prewritten responses.”

That's two-thirds of the population who incorrectly think AI works like a search engine or FAQ bot.

This misconception explains why work slop is so pervasive. Employees who believe AI retrieves facts from an authoritative source have no reason to verify outputs. They copy, paste, and send, trusting that the confident-sounding response must be accurate.

We produced a report, "AI for Startup Leaders," to help companies understand exactly how AI functions, so you can manage the strengths and pitfalls of the technology. In the guide we quote OpenAI CEO Sam Altman who diagnoses the problem businesses have with AI:

"Using these models as databases is sort of ridiculous. It's a very slow, expensive, very broken database."

Altman argues that the real power of AI is in reasoning, not retrieval. That’s why business leaders should give AI models access to their company data and tools to take action, while also enabling staff with an interface to use the models collaboratively. Then instead of treating AI as an oracle that dispenses facts, business professionals can use it as an intelligence layer that can reason through business problems and perform tasks.

A mental model that prevents work slop

The practical fix is to ground AI in source systems, make its reasoning visible, and keep review close to the work. An integrated AI coworker can collect evidence across tools and show the team what it used before anyone acts.

Before employees can use AI responsibly, they need to understand what it actually does. In "AI for Startup Leaders," we present a framework called A-R-C that captures the three core capabilities of AI:

Agency: AI can work with tools, run code, and complete tasks on an employee’s behalf. For example, it can query data from your CRM, draft documents, or update a spreadsheet.

Reasoning: AI can plan ahead and think through problems step-by-step. The latest models can break down complex challenges, consider multiple approaches, and work through logic chains.

Context: AI can base its answers on the information you provide in each conversation. It can understand natural language, identify objects in photos, and parse data in spreadsheets.

Notice what’s missing from this framework. A-R-C doesn’t claim that AI knows everything or can replace the judgment of a team of professionals. The public internet AI was trained on lives inside the model as statistical patterns, not a library of facts. So when you ask AI a question, you’re prompting it to generate what a correct answer usually sounds like.

Often it lands beautifully. And then, at the worst possible moment, it confidently makes something up. And that’s how work slop is created.

AI native employees

When employees understand A-R-C, they approach AI outputs differently. They recognize that AI is best used to reason from their business context rather than look up answers. And they use AI to speed up how they collaborate with other teams, rather than dumping work slop over a cubicle wall.

Adapt is built on this principle of empowering cross functional teams to become AI native. Instead of giving individuals just a private AI assistant, Adapt works in shared Slack channels where teams prompt AI together.

Everyone sees the questions and the outputs, and can participate in the critical evaluation that follows. This enables interactions where AI power users model appropriate skepticism and demonstrate best practices, while novices to AI learn through participation.

When AI is a shared teammate rather than a solo productivity hack, teams produce higher quality work, rather than useless slop.

FAQ

What is work slop?

Work slop is AI-made output that looks professional but doesn’t contain the substance needed to produce results. It creates extra cleanup work and quietly damages trust between coworkers.

What is a poor use of AI at work?

A poor use of AI is treating the models like an authority or “fact machine,” then copy-pasting its confident answer without checking it against real sources or your company’s data. That mindset makes it easy to ship polished nonsense that sounds right but isn’t grounded in reality.

What can AI do well?

AI shines when it’s used to think through a problem and complete tasks based on the company data and tools it has access to. It’s strongest as an intelligence partner that helps teams think and create faster, and fails as a database that guarantees truth.

How does Adapt help teams produce high quality work?

Adapt helps teams collaborate with AI so the prompts, outputs, and critiques are visible to everyone, which encourages verification and better judgment. That transparency lets experienced teammates model good AI habits and makes low-quality “slop” harder to slip into the workflow.