The Hidden Cost of AI Workslop

As organizations move quickly to adopt artificial intelligence, many are measuring how frequently employees use the technology. They track licenses, logins, participation, and time saved. Far fewer are examining the quality of the work employees produce with AI or the additional burden that poorly reviewed output can place on colleagues.

Recent research from Stanford’s Social Media Lab and BetterUp Labs describes one form of that burden as “workslop.” The researchers define it as low-effort, AI-generated material that appears polished but requires the recipient to complete the thinking, verify the information, or reconstruct the work.

The document may be well formatted. The language may sound authoritative. Yet when someone attempts to use it, the weaknesses become apparent. Sources may be missing. Numbers may be inaccurate. Conclusions may not follow from the evidence. Important context may have been overlooked.

What initially appears to be greater productivity may simply be shifting time and responsibility from one employee to another.

When Saving Time Creates More Work

In a survey of 1,150 full-time employees in the United States, 41% said they had received workslop during the preceding month. Each incident required an average of one hour and 56 minutes to resolve.

Nearly two hours of a colleague’s time was spent reviewing, correcting, or rebuilding work that was supposed to make the organization more efficient.

Using participants’ reported salaries and estimates of the time involved, the researchers calculated an average cost of approximately $186 per employee each month. For an organization with 10,000 employees, that could represent more than $9 million a year in lost productivity.

The precise cost will vary by organization and role, but the underlying issue is clear. Time saved by one employee has not necessarily been saved by the organization. It may simply have been transferred to someone else.

Consider a competitive analysis that arrives looking complete but contains inconsistent figures, unsupported claims, and conclusions that do not reflect the team’s actual data. Before using it, the recipient must check the sources, resolve the contradictions, and determine which portions can be trusted.

The recipient is no longer reviewing a colleague’s completed work. That person is finishing it. AI-generated work becomes a collaboration problem when the sender transfers responsibility for evaluation, verification, and judgment to the recipient.

The Consequences Extend Beyond Productivity

The cost of poor AI-assisted work is not limited to time.

Among employees who received workslop, 53% said they felt annoyed. Forty-two percent viewed the sender as less trustworthy. Approximately half considered the colleague less creative, capable, or reliable than they had before receiving the material.

These findings point to an important consequence of careless AI use. When employees cannot be confident that colleagues have reviewed or can stand behind the work they share, trust begins to weaken.

Trust is closely connected to the way teams collaborate and perform. RW3 CultureWizard’s High Performing Teams research found that 97.4% of professionals who strongly agree that trust exists on their team also agree that culture drives performance.

This is more than an interpersonal reaction to one poor document. Team members depend on one another to contribute sound thinking, exercise judgment, and take responsibility for their work. When those expectations are not met, colleagues may begin questioning whether they need to verify every statistic, recommendation, or conclusion they receive.

That uncertainty creates friction. Work is reviewed more cautiously. People may become less willing to rely on one another or delegate important assignments. Decisions take longer.

AI is not causing employees to lose trust in one another simply because it was used. The problem arises when AI allows unfinished thinking to be presented as completed work.

Access Is Not the Same as Capability

It may be tempting to treat poor AI-assisted work solely as an individual performance problem. Individual accountability is certainly important. However, the researchers caution against focusing only on the person while overlooking the conditions shaping the behavior. This is an example of the fundamental attribution error: attributing a problem to someone’s character or competence while discounting the influence of the environment.

Many employees are being encouraged to use AI without receiving clear guidance about what responsible use looks like. Organizations may celebrate rapid adoption without explaining when AI is appropriate, what must be verified, or how much human review is expected.

Logging into an AI platform demonstrates access. It does not demonstrate that an employee knows how to evaluate the output, apply relevant context, or decide when the technology should not be used.

When usage becomes the primary measure of success, employees may feel pressure to use AI even when it does not improve the work. They may also assume that speed is valued more highly than accuracy, context, or careful judgment.

Leaders should therefore look beyond how often employees use AI and ask more useful questions:

• Is the work accurate?
• Can the evidence be verified?
• Does the material reflect the organization’s actual objectives?
• Can the recipient use it without having to reconstruct it?
• Is someone prepared to take responsibility for the final result?

These questions measure capability and quality, not simply adoption.

Leadership Sets the Conditions

Organizations need clear expectations for AI-assisted work before it reaches a colleague, customer, or client.

Leaders can begin by clarifying which tasks are appropriate for AI support. Artificial intelligence may be helpful for organizing information, generating an initial structure, comparing possible approaches, or identifying questions that need further exploration. It should not be treated as an unquestioned authority or as a substitute for knowledge of the organization, its customers, or the consequences of a decision.

Teams also need standards for verification. Employees should understand when sources are required, how statistics must be checked, and what kinds of information require additional review. Confidential, legal, financial, medical, or personnel-related work may require particularly careful oversight.

Most importantly, responsibility for the final work must remain with a person. AI may contribute to the process, but it cannot be accountable for the result. The employee sending the material should be able to explain the reasoning, identify the sources, and defend the conclusions.

Leaders must also create a climate in which employees can question polished, confident-looking output. Raising a concern about AI-generated work should not be interpreted as resisting innovation. It should be recognized as responsible judgment.

When speed is rewarded without equal attention to quality, people will naturally focus on producing more. When careful review, sound reasoning, and constructive questioning are recognized, employees learn that responsible use is the real expectation.

Skill Building: Helping Team Members Question AI Downloads

A few disciplined habits can help teams prevent incomplete or unreliable work from being passed to colleagues.

1. Verify the evidence

Every important statistic, quotation, factual claim, and research finding should be traceable to a credible source.

AI systems can generate references that sound plausible but do not exist. They can combine separate findings, misstate conclusions, or present outdated information as current. If the source cannot be located and reviewed, the claim should not be shared as fact.

“The AI provided it” is not a source.

2. Review the work for usefulness

Read AI-assisted work from the recipient’s perspective.

Is the purpose clear? Does the reasoning make sense? Are there unexplained gaps? Does one section contradict another? Can the recipient act on the information without having to seek extensive clarification?

Encourage team members to reread AI output as if they had never seen the topic before. Gaps, contradictions, and vague phrases are much easier to catch from that angle.

3. Apply human judgment

Grammar and formatting are only part of a meaningful review. Someone with relevant knowledge must determine whether the ideas are sound, the conclusions are justified, and the content fits the situation.

A person should ask:

• What would have to be true for this conclusion to hold?
• What information may be missing?
• What assumptions shaped the response?
• Who could be affected if the answer is wrong?

Normalize asking "what would this need to be true?" A simple prompt like this pushes people to check assumptions instead of accepting confident-sounding language at face value.

4. Make questioning AI part of the culture

Encourage everyone to identify weaknesses and challenge assumptions before unreliable work progresses further.

This requires trust in the company and basic psychological safety. People must feel able to say that a conclusion is unsupported; a source cannot be verified, or a document is not ready without being viewed as difficult or resistant to change. Reward the person who flags a problem, not just the person who ships fast.

The employee who prevents a flawed recommendation from reaching a customer may be contributing more value than the employee who produced the recommendation quickly.

5. Pause and Reflect Before Sending

RW3 CultureWizard’s Pause-Reflect Model offers a practical way to approach AI-assisted work.

Pause. Before forwarding or submitting the material, stop and consider why it appears convincing. Is it accurate and relevant, or does it simply sound polished and confident?

Reflect. Examine the assumptions, evidence, and context. What information was used? What may have been omitted? Does the output reflect the realities of the organization and the people affected?

Evaluate. Verify the sources, test the reasoning, and consider the possible consequences of an inaccurate or incomplete answer.

Act. Revise the work, seek another perspective, conduct additional research, or decide not to use the output. Share it only when a person is prepared to take responsibility for it.

This process does not require elaborate new technology. It requires people to slow down long enough to bring judgment back into the work.

Measuring What Matters

AI can help teams save time, improve access to information, and perform more effectively. Those benefits depend on how thoughtfully the technology is used.

Organizations need to look beyond adoption and ask whether AI-assisted work is accurate, useful, and ready to be shared; that it’s not workslop. They need to measure not only the time employees believe they have saved, but also the time colleagues spend correcting, interpreting, or rebuilding AI-generated material.

And, pay attention to the effects on trust and friction among colleagues.

When team members know that colleagues have carefully reviewed their work and can explain the reasoning behind it, AI can strengthen productivity and collaboration. When unfinished thinking is passed along as completed work, the technology can add cost, frustration, and doubt.

The real measure of successful AI adoption is not how often employees use it. It is whether the technology helps people produce better work together.