Editorial illustration for Macmillan Workers Say Publisher Celebrated an AI Assistant Before Killing It
Macmillan Workers Say Publisher Celebrated an AI...
Macmillan employees walked into the publisher's Lower Manhattan office one morning in January 2024 to find purple balloons, light-up marquee letters, and a plate of doughnuts arranged to spell "NOA IS HERE." The occasion was the launch of an AI chatbot called NOA, short for No Ordinary Assistant, built by the German firm Chaptr and backed by Holtzbrinck, Macmillan's parent company. Several human assistants at the company found the acronym "offensive," according to multiple employees who spoke to WIRED.
Macmillan told staff that NOA ran on a closed system, meaning no proprietary data would leak to outside parties during use. Workers say training sessions came with warnings that the tool could hallucinate and needed human fact-checking at every turn. Despite that caveat, employees were pushed to lean on NOA for emails, marketing copy, research, and brainstorming.
Some staffers recall being told to paste only a "percentage" of an author's manuscript into the system at a time. Users could also dial NOA's tone to settings like "lighthearted," "spartan," or "regretful." The celebration, and the tool itself, didn't last.
“Publishers have always seen authors as a bit of a nuisance, and now there’s a desire to make us obsolete,” says one bestselling Big Five author. “What’s fascinating to me is that in the process of trying to make authors obsolete by using AI, the publishers are rendering themselves obsolete.”
Why this matters
NOA's funeral is a useful data point for anyone pitching workplace AI right now. Macmillan didn't quietly retire a tool, it staged a launch party with balloons and light-up letters, then killed the thing and moved on to Gemini, which editors say produces "useless" evaluations of agent submissions. That's two failed attempts at the same basic task: using AI to help humans judge text.
For founders selling "AI assistants" into white-collar workflows, the lesson isn't that publishing is resistant to automation. It's that the gap between demo-ready and desk-ready remains wide, and companies are willing to burn real money on launch spectacle before they've confirmed the tool works. Researchers building evaluation or summarization models should take the Gemini complaint seriously.
If Gemini can't reliably assess a literary submission letter, that's a concrete failure mode worth studying, not dismissing as one bad use case. And for anyone tracking enterprise AI adoption, Macmillan's pattern, big rollout followed by quiet abandonment, is probably more common than the case studies suggest. The NOA balloons are gone.
The underlying problem they were meant to solve isn't.
Related Reading
AI Tools & Apps Claude gains shared context in Excel, PowerPoint; Microsoft adds Copilot Cowork
AI Tools & Apps Sarvam Signs 5-Year Defence AI Deal Covering Army, Navy and Air Force
AI Tools & Apps LG's recent webOS update adds Microsoft Copilot app, now removable
From Archives Judge Sees AI in Courts as Tool for Greater Access to Justice
From Archives Firefox’s Smart Window adds AI form filling and webpage monitoring
Further Reading
- Macmillan Staff Say the Publisher Threw a Party for a Workplace AI Assistant That Was Later Killed - WIRED
- The New Premium Product: Books Written by People - Financial Times
- Potential Impacts and Perceptions of GenAI in Trade Publishing: A Review and Agenda - Journal of Media Business Studies
- Book Publishers Face Rising Scrutiny Over AI Usage by Authors - The Christian Science Monitor
- AI-Assisted Writing: Is It Good for Authors and Book Publishing? - MIT Initiative on the Digital Economy
