Stop AI Workslop to Unlock ROI

Many organizations see little ROI from generative AI because employees are producing workslop—outputs that look good but create rework for others. Survey data show frequent occurrence, substantial time and cost burdens, and interpersonal fallout that undermines collaboration. Leaders must replace blanket AI mandates with clear guardrails, purpose-driven use, and collaborative norms to capture value.

Key Points

AI adoption is high but ROI is low because employees produce workslop—polished yet shallow AI outputs that shift effort to recipients.
Workslop is common and costly: 40% encounter it monthly; it consumes nearly two hours per incident and amounts to an estimated $186 per month per employee and over $9 million annually for a 10,000-person firm at 41% prevalence.
The interpersonal toll is significant, with recipients annoyed or confused and viewing senders as less competent and trustworthy, which harms collaboration.
Indiscriminate AI mandates encourage indiscriminate usage; leaders must provide nuanced guidance, guardrails, and norms tied to strategy and values.
Mindsets matter: pilots (high agency and optimism) use AI to enhance creativity and outcomes, while passengers use it to avoid work; organizations should foster pilot behaviors and integrate AI into collaborative workflows.

Sentiment

The discussion overwhelmingly agrees with the article's diagnosis of AI workslop as a real and growing problem. However, commenters are considerably more cynical than the article about solutions — where the article recommends cultivating a pilot mindset and setting guardrails, commenters see this as naive corporate consulting language that misses the structural dysfunction driving workslop. The community views the problem as rooted in perverse corporate incentives and management hype rather than individual worker behavior.

In Agreement

Corporate mandates to use AI and document only positive outcomes mirror Soviet Stakhanovism — workers are pressured to fabricate success stories while negative outcomes are ignored
Workslop transfers effort from creator to recipient, applying Brandolini's law: the energy to review and refute AI-generated work vastly exceeds the energy to produce it
A tech consultancy manager saw zero measurable efficiency gains despite team-wide AI adoption, finding AI useful only for prototyping vanilla applications not enterprise maintenance
AI-generated code from non-technical managers creates a massive review burden, producing work that appears functional but contains elaborate non-working workarounds
AI is creating an arms race of junk — AI generates content that more AI parses and summarizes in an infinite regression that burns energy without adding value
Much of corporate work that AI is meant to replace was already meaningless busywork, connecting the problem to David Graeber's Bullshit Jobs thesis

Opposed

Most corporate roles can find at least some modest AI benefit, and claiming total uselessness borders on bad-faith refusal to engage with the technology
AI tools require genuine skill to use effectively, and many people haven't invested the effort to move past their productivity local maximum
Using AI-generated code as exploratory reference then rewriting manually can accelerate delivery while maintaining quality
The real problem is poor change management rather than the toolset itself — organizations struggle to re-skill resistant employees at the pace required
AI productivity research is flawed when test subjects lack unlimited tokens; the true ROI should compare token cost to hourly wage times time saved