Rytor raises $100 million to bring its "full suite" AI models to more businesses

2023-09-19

Writer is a San Francisco-based startup founded three years ago, focusing on building a secure generative AI platform to help businesses complete tasks. The company announced today that it has raised a Series B funding round totaling $100 million. The round was led by ICONIQ Growth, with participation from WndrCo, Balderton Capital, Insight Partners, and Aspect Ventures.

Notably, Writer's enterprise clients, including professional services company Accenture and investment giant Vanguard, also participated in this funding round, indicating their confidence in Writer's products.

This funding will help Writer continue to develop more new features and AI agents, and add its proprietary AI models to its "suite" of enterprise-friendly software applications.

The company currently offers 14 different AI models, including Palmyra Small (128M), Palmyra Base (5B), and Palmyra Large (20B), which have helped businesses complete various tasks such as content creation, email, audio transcription, and summarization. PalmyraMed is specifically designed for healthcare clients.

While the B2B software market has seen a proliferation of new generative AI tools and applications in the past six months, many of which use large language models (LLMs) like OpenAI's ChatGPT or Meta Platforms' Llama 2, Writer aims to stand out with its in-house AI models.

In a press release today, the company stated that its target audience is professionals working in fields such as support, operations, product, sales, human resources, and marketing.

Some of these models are tailored for specific industries, such as healthcare, and have fully interpretable characteristics, which is a unique selling point in an industry often criticized for its "black box" nature.

Trained on non-copyrighted, enterprise-friendly data sources

According to an email sent by Writer CEO and co-founder May Habib, Writer's AI models are trained on "trillions of words of formal and business writing vocabulary, excluding any copyrighted or proprietary content." This information was provided by a spokesperson specifically to VentureBeat.

So how does it perform? In this regard, Writer also has something to be proud of: the company's Palmyra model ranked first in three important tests in the Stanford HELM benchmark for AI models, scoring 0.609 in "massively multilingual multitask language understanding (MMLU)," 0.896 in BoolQ, and 0.413 in natural questions, surpassing models from companies such as OpenAI, Cohear, Anthropic, Google, and Microsoft.

"The final step, quality, is the hardest," Habib said in the press release.

Why is the leadership model named Palmyra? According to the author's explanation in a blog post in the model series, the name "comes from the ancient city in present-day Syria where Queen Zenobia resisted the Romans and established the Palmyrene Empire," as an example of respect for other female leaders, just like Habib.

More than just models

Writer also offers a "knowledge graph" to help models connect and reference the company's proprietary data (controlled by clients), as well as various chatbots and a suite of applications, including pre-set templates, professional features, and "composable" user interface options, meaning you don't have to be an AI expert to use Writer's LLM Palmyra.

The software layer does a lot of the work, so implementation is often very straightforward, such as Writer's "Recaps" application, which allows employees to drag and drop audio recordings of meetings, webinars, calls, etc., and automatically summarize the key points.

This means that Writer not only competes with other foundational model providers in the enterprise, but also offers applications that compete with professional software providers like Gong and Otter.AI.

Security and data privacy

The results are undoubtedly impressive, but why should potential enterprise customers trust Palmyra and other models over more well-known LLMs like ChatGPT (which recently launched an enterprise version), Kohre, Llama 2, or others? Habib claims that in addition to being trained on "clean" and enterprise-focused datasets, Palmyra also offers the appropriate balance of security and privacy that enterprises seek.

Habib wrote, "Authors protect the data privacy of LLM customers and never use it for model training." "Writer also offers customers the option to self-host their LLM."

In a post on X (formerly known as Twitter) earlier this year, OpenAI CEO and co-founder Sam Altman stated that OpenAI never trains on content submitted to the API or uses that data in any way to improve the model, and CNBC host Andrew Ross Sorkin conveyed to him that OpenAI has changed its terms of service to clarify this fact.

Palmyra has also achieved Soc-2 Type II, PCI, and HIPAA (Health Insurance Portability and Accountability Act) certifications, equivalent to Anthropic Cloud 2, also known as HIPAA certification.

Satisfied enterprise customers

In addition to announcing the funding, Writer also highlighted the benefits its clients have gained from its platform. Lingerie/underwear retailer Adore Me saw a 40% increase in non-branded SEO traffic by using Writer's Palmyra model, while process automation platform company UiPath said it has reduced content creation time by half.

Writer stated that its Palmyra enterprise platform has proven its value in multiple fields, with revenue growing tenfold in the past two years and a net revenue retention rate exceeding 150%.

With this new funding, Writer plans to further expand its product and accelerate the embedding of general AI into tasks in the enterprise domain. As AI continues to drive the adoption of advanced tools and technologies, Writer's comprehensive approach to flexibility and governance makes it an attractive choice for businesses seeking to fully leverage the potential of AI.

RECENT AI NEWS

RECENT AI TOOLS