Macmillan Publishers Launched an AI Workplace Chatbot With a Party. Staff Called It Useless. It Was Shut Down Within Two Years.
Macmillan launched NOA in January 2024 with balloons and doughnuts, but staff called the Chaptr-built chatbot unreliable and largely ignored it. The publisher shut it down in October 2025, less than two years later.
Macmillan Publishers rolled out an AI workplace chatbot called NOA in January 2024 with balloons, banners, and a doughnut display at its Lower Manhattan headquarters — only to shut it down in October 2025 after staff widely dismissed it as unhelpful.
NOA, an acronym for “No Ordinary Assistant,” was developed by German software company Chaptr, which is backed by Holtzbrinck, Macmillan’s parent company. The tool was introduced as a closed-loop enterprise chatbot, with the company assuring staff that proprietary information could be entered without risk of being shared with third parties. Employees were encouraged to use it for drafting emails, marketing research, generating ideas, and even writing their own performance review goals.
The reception among staff was largely negative. Multiple employees told WIRED the tool was unreliable and went largely unused. “It was terrible, and no one used it,” one staffer said. Workers also noted that the name itself caused friction — many of Macmillan’s human assistants found the “No Ordinary Assistant” branding offensive. During training sessions, employees were warned that NOA could hallucinate and that its outputs would always require human fact-checking.
A Chaptr representative’s comment during an early training session drew a pointed reaction from at least one employee. After the representative noted that Chaptr understood publishers were dealing with something important — “content” — the staffer recalled thinking: “And here I thought we were publishing literature.”
Following NOA’s shutdown, Macmillan leadership began attempting to integrate Google Gemini into publishing workflows. Early results were similarly underwhelming; one editor described Gemini-generated evaluations of literary agent submission letters as “useless.”
The episode reflects broader tensions between major publishers and AI tools, with staff accounts suggesting that enthusiasm from leadership has not translated into practical value on the ground.