Startups are clamping down on internal AI slop: ‘You’re losing so much’

As AI-generated writing becomes more common in startups, internal communication problems are arising due to the quirks and shortcomings of large language models. Companies are implementing policies to prevent employees from relying solely on AI for writing documents, emails and Slack messages. Clay, a San Francisco-based sales intelligence startup, has published its own AI Writing Policy, emphasizing that employees are responsible for ensuring their thoughts are reflected in any shared document.

Dutch startup Polarstep also released an AI writing policy, noting that team members were struggling to understand each other due to the excessive use of certain phrases and vague language generated by AI. Clear and concise communication is crucial in fast-moving startup environments, and AI can hinder this process by producing lengthy, convoluted, and often inaccurate text.

Large language models still struggle to provide accurate, well-sourced information, and they can easily muddle sentences, exaggerate facts, and sometimes generate entirely false content. This makes it essential for humans to double-check AI-generated material to maintain productivity and avoid wasting time. Furthermore, AI-generated writing can lead to a loss of individual personality and voice, which may erode trust among team members and impact the team's overall cohesion.

While AI can still be beneficial for presenting information in various formats tailored to individual needs, such as turning documents into podcasts or enabling voice notes, the regulation of AI usage in writing is expected to persist. Companies like Anthropic, LinkedIn, and Substack are exploring solutions like watermarking AI-generated text and developing tools to detect AI-sloppy content.

These measures aim to strike a balance between leveraging AI's potential benefits and maintaining high-quality communication in the workplace.