AI startup writer raises $100 million to take on ChatGPT enterprise

Innovation

Firms like Spotify, Uber and Accenture use Writer’s generative AI tools to research, create and analyse content.
May Habib, cofounder and CEO at Writer (left) and Waseem Alshikh, cofounder and CTO at Writer (right) have raised $100 million in a Series B funding round led by Iconiq Capital.

May Habib, cofounder and CEO at Writer (left) and Waseem Alshikh, cofounder and CTO at Writer (right) have raised $100 million in a Series B funding round led by Iconiq Capital.

Writer


Generative AI startup Writer has raised $100 million in a series B funding round, valuing it between $500 million and $750 million, the company announced Monday. Writer’s large language models produce content ranging from incident reports and emails to product descriptions and executive summaries, placing it squarely in competition with OpenAI’s ChatGPT Enterprise, which was launched last month, and other fast-growing unicorns like Typeface.

But despite the crowded generative AI space, Writer CEO and cofounder May Habib told Forbes that enterprise customers are switching from Azure OpenAI over to Writer because the quality of outputs generated by ChatGPT has deteriorated. Her startup’s revenue has increased by 10 times in the last two years and four times since the start of this year alone.

“We are seeing a lot of folks reach out as they’re kind of stuck in proof-of-concept purgatory,” she said. “We’ve seen a few customers where they haven’t been able to take use cases to production because the generations weren’t high quality enough.”

Writer, which was founded in 2020 and featured on Forbes’ AI 50 list earlier this year, touts 150 enterprise customers including Uber, Spotify, Vanguard, Samsung, Accenture and L’Oreal. The round was led by Iconiq Growth with participation from WndrCo, Balderton Capital, Insight Partners and Aspect Ventures, and brings the startup’s total funding to $126 million.

Writer offers a group of 14 models of different sizes, from 128 million parameters to 43 billion parameters — far smaller than the size of OpenAI’s GPT-4, which reportedly has about 1 trillion parameters. The models, called Palmyra, were trained on public data from sources like web pages, books, Wikipedia, Github and transcribed video content from YouTube. The public data is filtered to remove copyrighted content, Habib said. Each company gets a separate fine-tuned version of the model trained on company-specific proprietary data such as financial reports and marketing copy. The models are also compliant with most privacy and security standards, including HIPAA and Europe’s privacy legislation GDPR.

Most recently, in July, Writer released PalmyraMed, which is trained on public medical datasets like PubMedQA that include questions and answers and articles, for healthcare applications. Writer’s smaller fine-tuned models are trained at a hundredth of the cost of other large language models and are catered to performing specific tasks at faster speeds, Habib said.

“The challenge enterprises face is fine tuning a good-at-all LLM on the ocean of data that they have and coming up with a model that can solve every problem,” she said. “It’s really the last mile of tuning, data preparation and integration into current workflows that enterprises need help with.”

Habib said Writer’s key selling point is its integration with the tools that employees use (Salesforce and Adobe), and that the app can be embedded into workspaces like Google Chrome, Figma, Google Docs, Canva, Microsoft Word and Outlook by installing Writer’s plugin or extension which automatically marks up the content with suggestions. “The novelty of chat has worn off,” she said.

This article was first published on forbes.com and all figures are in USD.

More from Forbes Australia

Avatar of Rashi Shrivastava
Topics: