Start with the software you already own: companies on Microsoft 365 or Google Workspace should try Copilot or Gemini first, Notion and Grammarly users already have capable AI in plans they own, and ChatGPT Business or Claude Team add a general assistant for about $20 per person monthly on a yearly plan. Specialists earn a place only for specific jobs: Jasper for brand-consistent marketing content, Writer for enterprise governance, Copy.ai for sales workflows and Anyword for performance-scored ad copy. Three practical questions follow, on personal accounts, AI detectors and labeling.
Can employees use their personal ChatGPT or Claude accounts for work?
It is risky. OpenAI says it may use content from its individual Free, Plus and Pro plans to train its models unless the user switches off a setting, though it does not train on business plans by default, and since a change announced in August 2025 Anthropic asks consumer users to choose whether their chats may be used for training, keeping data for five years for those who agree instead of 30 days; neither rule applies to business plans. The cautionary tale is Samsung, which in May 2023 banned staff from using generative AI tools after employees uploaded sensitive source code to ChatGPT. Pay for business seats, turn off training where a plan offers the choice, and write a short policy on what staff may paste into any AI tool.
Can AI detectors tell whether text was written by AI?
Not reliably. OpenAI withdrew its own AI text classifier on July 20, 2023 because of its low accuracy; at launch it correctly identified only 26% of AI-written text and wrongly flagged human writing 9% of the time. A Stanford study published in the journal Patterns that July found that popular detectors labeled essays by non-native English speakers as AI-generated 61% of the time on average, while essays by US eighth-graders were judged almost perfectly. Do not use a detector score to discipline an employee or reject a freelancer’s work; judge drafts on accuracy, originality and sources instead.
Do we have to label AI-written content?
In the US, generally not: no federal rule requires a label on AI-assisted marketing text, though the FTC’s deception rules still apply to what the text claims, and its 2024 rule on fake reviews covers reviews written with AI. California’s AI Transparency Act puts its disclosure duties on the companies that provide generative AI systems rather than the businesses using them. Companies that publish to European audiences face more: since August 2, 2026, the EU AI Act requires disclosure when AI-generated text is published to inform the public on matters of public interest, unless a person has reviewed or edited it and someone holds editorial responsibility. Keeping a named human editor accountable for everything you publish fits that exception and is good practice anyway.