Macmillan staff say the AI chatbot NOA was unreliable and was stopped
Staff at Macmillan Publishers say that NOA was not useful. A company that the parent of Macmillan backs made it.
Employees of Macmillan Publishers in New York say that NOA, an AI chatbot, was unhelpful and unreliable. The company launched it with a party in January 2024. In October 2025, staff learned that Chaptr, the German software company that made NOA, would stop it. Since then, staff say, leaders try to use Google Gemini. The employees gave no names, to protect their careers.
What staff say
Staff were encouraged to use NOA for emails, copy, marketing research and goals for performance reviews. They were told to paste only a percentage of the manuscripts of authors into it. One employee says: “It was terrible, and no one used it.”
The proof
The account has no numbers for use, cost or test results. It rests on the words of staff. Staff say that they were warned that NOA could invent facts and that humans must check its work.
What is not known
No comment from Macmillan or Chaptr is given. The cause of the stop is not stated. Some staff tested Gemini on letters from literary agents. One editor called the results “useless”.
Insight: Big job cut guesses meet weak real rollouts
Eric Schmidt guessed, with no data, that AI can replace 2,000 of 10,000 employees. Staff at Macmillan say the first tool had no users. One editor calls the next tool useless for one task. This reads as a distance between what leaders claim and what staff do: a tool that comes from above, with no use data, is dropped. Measure use by task before you cut or buy.
Sources
Posted
