Macmillan Publishers introduced an internal artificial intelligence chatbot called NOA, or “No Ordinary Assistant,” in January 2024, presenting it as a secure workplace tool for tasks including email writing, copy drafting, marketing research and idea generation. Employees told WIRED that they were warned about hallucinations and instructed to fact-check its output, yet many found NOA unreliable and unhelpful. Macmillan later discontinued the system after its developer, German software company Chaptr, announced in October 2025 that it was ending the service.
Macmillan introduced NOA with a celebration
When employees arrived at Macmillan’s lower Manhattan headquarters one morning in January 2024, they found a flyer welcoming new team members called NOA. The cafeteria was decorated with purple balloons and banners, and illuminated marquee letters displayed NOA’s name, according to several employees who were present. A plate of doughnuts carried the message, “NOA IS HERE.”
In an email to staff, Macmillan described NOA as an AI chatbot developed by Chaptr, which is “backed” by Holtzbrinck, Macmillan’s parent company. The acronym stands for “No Ordinary Assistant.” Multiple employees told WIRED that many of Macmillan’s human assistants felt “uncomfortable” about the introduction.
Employees were told NOA was secure
The email presented NOA as a “secure environment where they can enter sensitive information without the risk of sharing data with third parties.” That description appeared to mean the system was a closed-loop enterprise tool that did not retain or share data for model training.
During training sessions, Macmillan staff said they were warned that NOA could produce hallucinations and that its responses always required human fact-checking. Employees were nevertheless encouraged to use it to write emails, draft copy, conduct marketing research and generate ideas.
Several employees said they were instructed to paste “only portions” of authors’ manuscripts into NOA. The chatbot also offered settings that allowed users to adjust its tone, including “lighthearted,” “stern,” “pessimistic,” “sarcastic” and “regretful.”
One Macmillan employee recalled that, during an early in-person training session, a Chaptr representative reassured staff that the company understood publishers were dealing with something very important: content. “And I thought, here we are publishing literature,” the employee said.
Staff questioned the chatbot’s usefulness
In late 2024, Macmillan employees were encouraged to use NOA to generate performance appraisal objectives. “We created a special prompt for NOAs to write down their goals,” one email stated.
Several Macmillan employees told WIRED that NOA was unhelpful and unreliable. “No one used it because it was so bad,” one person said.
WIRED spoke with several publishing employees for the article, including some who had since left their employers. The employees and authors interviewed were not identified because they were concerned about repercussions for their careers.
Macmillan later turned to Google Gemini
In October 2025, Macmillan employees learned through a company-wide email that Chaptr was discontinuing NOA. Since then, company leaders have been working to integrate Google Gemini into the publishing process, according to workers.
Some employees subsequently experimented with using Gemini to rate submission letters from literary agents. One editor said the ratings generated by Gemini were “useless.”
Source: www.wired.com