Enterprise AI startup Writer has outgrown its name. Its large language models that were initially popular for helping companies swiftly create blogs, summaries and product descriptions are now being put to work across departments for a broad spectrum of tasks like building applications, analyzing documents and ensuring companies stay compliant. But CEO May Habib doesn’t mind that her company does more than its name suggests.
“What we say now is that if you can write it, if you can describe it, you can build it,” Habib told Forbes.
Writer announced Tuesday that it has raised $200 million in a Series C funding round co-led by AI-focused venture capital firm Radical Ventures, Iconiq Capital and Premji Invest, a private equity firm owned by billionaire Azim Premji. Founded in 2020, the company is now valued at $1.9 billion with about $320 million in total funding. It expects to record about $50 million in annual recurring revenue by the end of this year from more than 300 customers.
Writer is a full-stack AI platform built on top of its family of proprietary large language models called Palmyra, which the company claims are able to provide more accurate (albeit less creative) responses thanks to a different architectural approach than its peers like ChatGPT.
The models learn each company’s lingo from documents like marketing copy and how-to articles, and pull details from financial documents.
Each customer gets its own private version of the Palmyra models, helping to mitigate any concerns about proprietary data, and then Writer’s team builds customized applications on top. For instance, Uber’s version of Writer generates responses to the most common customer support questions and helps its 40,000 human support agents. Qualcomm’s legal teams use it to manage hundreds of trademarks and find trends in the market. And Dropbox leans on Writer to generate SEO blogs and proofread its existing content.
Now, the company is planning to build AI agents, software that can access external tools and autonomously carry out actions. It’s why companies like Salesforce, Adobe and Workday, which provide tools that employees use everyday, are also investing in the startup, Habib said. “We see incredible demos on Twitter and the gulf between that and what you’re actually able to get done in your business is so huge,” she said. “We’re investing in the next generation of action AI …doing it in a way that helps customers actually get to this AI nirvana.”
A group of well-funded AI titans like OpenAI, Anthropic and Cohere are also in the race to sell their technology to enterprises, with their eyes on building AI agents. To compete, Writer has been training and releasing new and more capable models, while trimming down its own costs. In July, Writer launched Palmyra-Fin and Palmyra-Med, which cater to financial and medical use cases. The company said they are more accurate (based on multiple medical benchmarks) and cheaper to use than other existing models like GPT-4 and Claude 3.5. In October, Writer launched a foundational model that it trained using synthetic data (data generated by an AI model). The total training cost was just $700,000 — a fraction of the more than $100 million it cost OpenAI to train GPT-4. “We have to be scrappier,” Habib told Forbes.
Adoption of AI models within enterprises is still nascent, as many don’t have the resources to build AI products themselves, said Radical Ventures Partner Rob Toews. Writer’s decision to offer to build entire bespoke applications for its customers underpinned by its own models — what it calls “AI workflows” — has helped it achieve a “screaming product market fit,” he said. “I think the full stack offering wins in a lot of cases against the ‘take a frontier model, and then try to build something yourself around it’ approach,” Toews said.
As companies experiment with AI tools, executives are also concerned about systems making up incorrect information that could hurt their brand or inadvertently leaking sensitive company information to the wider public. “The gap between the LLM and the app is significant,” said Sandesh Patnam, a managing partner at Premji Invest. “But with Writer.. they control their own destiny because the underlying frontier model is controlled within Writer.”
Ultimately, enterprises are working with hundreds of different applications and in some cases, employees spend hours transferring the same content from one format to another. CEO Habib hopes Writer becomes a “super app” that can interact with different applications all by itself, and make workers more productive. “This next era is going to be software that writes software that uses software,” she said.
