One growing area of employment involves "workslop" cleanup—essentially the task of fixing errors, inconsistencies, and mistakes made by artificial intelligence (AI) systems. Large language models (LLMs), the technology behind AI tools like chatbots and automated writers, are marketed as efficient, low-cost solutions for tasks such as coding, report writing, and data analysis. However, while these systems can generate vast amounts of content quickly, they often produce work that is “almost right,” with errors, inaccuracies, or gaps in logic that only a human can catch. This has created a new class of jobs known as AI remediation, where workers review and correct AI-generated outputs to ensure quality. These jobs vary widely in skill level and pay. Some companies now hire professionals from diverse fields—such as finance, tech, and retail—to review AI output and improve it when necessary. However, the need to constantly check and correct AI work has led some white-collar workers to report that AI is actually increasing their workload rather than reducing it. This can also strain trust between colleagues, as the reliability of AI-generated information becomes a concern. A 2025 Harvard Business Review report highlighted the frustration of a retail director who spent significant time verifying AI-provided data and coordinating with other supervisors to correct errors. The public sector is also beginning to face similar challenges. Governments, including the UK’s, are experimenting with AI for tasks like summarizing public feedback or drafting official letters. However, guidelines for civil servants emphasize the need for caution, as AI outputs can be misleading and must be independently verified. If AI is used extensively in public services, it may lead to the creation of new roles focused on monitoring and correcting AI-generated content before it reaches citizens or influences policy. This is crucial, as seen in the Netherlands, where an algorithm used by tax authorities wrongly accused thousands of families of fraud, causing financial ruin for many. The issue was not just the AI’s output but the lack of human oversight to question its decisions before they caused harm. Beyond the immediate need for human oversight, there is a deeper economic concern. The data that AI systems process—known as "tokens"—is often subsidized by companies with heavy debt, while the human labor required to review and fix AI output is not. Many AI firms offer their tools at below-cost prices to encourage dependence, and once companies become reliant on their systems, the firms can raise prices or reduce access. This could shift the cost-benefit balance of AI in the future, making it more expensive or less reliable over time. The human cost of AI systems is also a growing concern. Much of the “hidden” labor behind AI—such as moderating chatbots or labeling data—has long been outsourced to low-wage countries, where workers often face psychological strain from reviewing disturbing content. If AI workslop becomes widespread, it’s possible that the cleanup work itself could be outsourced, turning into a new form of low-paid, repetitive digital labor. This is already visible in translation services, where AI does the initial work and human translators are paid less to correct errors. This role is often seen as tedious and devaluing, with concerns that it may eventually be replaced by further automation. Efforts to resist this shift are already underway. In 2023, Hollywood writers went on strike in part to prevent studios from using AI to generate first drafts of scripts and then hiring low-paid humans to polish them. The resulting agreement with the Writers Guild of America set boundaries, such as not crediting AI as a writer or using it to undercut human pay. However, many other workers—such as translators and freelancers in gaming—lack the collective power to negotiate similar protections, making it easier for AI-driven models to normalize repair work as the new standard.