Generative AI is becoming more and more prevalent in the workplace, but not all executives know how to embrace it and not all employees feel confident using it. That’s where startup Writer.ai sees an opportunity.
As CIOs explore how to most effectively use AI within their organizations, Writer provides ready-made language models and natural language processing technology to expedite the process, empowering employees to write blog posts and sales emails, create answers to frequently asked questions, and use AI-enabled apps to generate job or product descriptions, regardless of their technical level.
Ideally, this will help smooth the transition to AI-enabled workflows. According to a June survey by software company Freshworks, IT staff are the most comfortable with AI, followed by marketing and finance. Given that generative AI tools have emerged at a rapid clip since AI startup OpenAI announced its generative AI chatbot ChatGPT in 2022, it’s no surprise that not everyone feels at home with AI.
Companies like Adobe, Amazon, Anthropic and OpenAI are pushing their own enterprise chatbots with similar goals, but Writer wants to take chat out of its element.
The app is a new chat
While mainstream chatbots work well for individual users, Writer CEO May Habib said these chat-based interfaces aren’t suitable for businesses with many different employees who will inevitably enter different prompts and get different outputs for the same task.
The spokesperson referred to the broader “over-chatification of AI”.
“When people think about generative AI, they only think of chat, but there’s a lot more value that can be unlocked beyond chat.within corporate organisations,” she said.
But the Writer app still has something of a chat component.
When interacting with mainstream chatbots, you first have to “train” them by telling them what you want. The Writer app eliminates that step. For example, if you want to generate headlines for your blog posts, you can tell the Writer app that you want it to think like an editor, show examples of headlines you like, and ask it to create concise, attention-grabbing headlines. Once you’ve established the ground rules with the app, you can simply enter the details of your article and it will generate your headline without any further chat.
“There’s a lot more rigor in what end users need to feed into generative AI tools and what output they can get out of them,” Habib said.
Interview with the writer and AI studio
San Francisco-based Writer’s clients, which include consulting firm Accenture, financial software company Intuit, beauty brand L’Oreal, digital music service Spotify and ride-sharing platform Uber, can access the app in two ways.
A ready-to-use chatbot is available, Ask Writer, which a spokesperson describes as “similar to ChatGPT, but custom built for enterprises.”
It comes with around 25 pre-built apps that run the gamut from generating transcript summaries to creating sales emails and error messages to performing image analysis.
Customers can also use AI Studio, which allows them to build their own apps through no-code build tools or APIs, which allows developers to build new applications using Writer technology, and the open-source app development framework, which is a library of tools and code that developers can use to create apps.
According to Writer, this allows employees, regardless of their tech savvy, to collaborate on custom apps.
“At Lighter Inc. we say, ‘If you can write it, if you can explain it, you can create it,'” Habib said.
LLMs and RAGs
The apps are powered by Writer’s proprietary large-scale language model, an AI model trained on huge datasets (though a spokesperson declined to say what exactly) that allows the apps to understand and generate content. LLM is the basis for mainstream chatbots like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude.
Generational AI startup Vectara estimates that mainstream models like GPT-4o, Llama 3.1 405B, and Gemini 1.5 Flash hallucinate or generate false or misleading content between 1.3% and 7.4% of the time.
The authors argue that this is not enough for businesses that require consistent and accurate output.
The Leiter LLM program is ranked by Stanford’s comprehensive assessment of language models, which evaluates AI models on criteria such as accuracy, general information and bias, according to a spokesperson.
In July, Writer added an LLM specializing in healthcare and finance.
Healthcare models are trained on patients’ medical summaries so they can analyse them and inform clinical decision making, while financial models are trained on how to rebalance portfolios so they can do that work for their clients.
“If we can work backwards from the use cases that our customers will need, we can train models very effectively and efficiently,” Habib said.
Writer also has a general purpose model and a vision analysis model, which can analyze images such as charts and graphs, as well as sketches and handwritten notes.
In addition to these LLMs, Writer offers RAG (Search Augmentation Generation), a type of natural language processing for queries.
“Given a question, we pull contextually relevant data points from reams of uploaded data and pass them through LLM to generate an accurate answer,” Writer said in a blog post.
This allows Writer to pull all of a company’s data and link it to the LLM, according to a spokesperson. The startup allows users to download files of up to 10 million words (about 20,000 pages, according to the company) of their own data to ask questions, research, and generate output.
“The value is in your information and your data,” she says, “and much of the industry has struggled to create RAG solutions that are accurate enough for businesses to use.”
The spokesperson did not elaborate on how accurate Writer’s RAG offering was.
Writer also offers AI guardrails designed to prevent the app from violating legal or ethical rules.
From Cordoba to the Author
The basic plan starts at $18 per user per month for up to five users. Writer offers custom pricing for enterprise users.
Habib and Writer co-founder and CTO Waseem Alshiq began working together on natural language processing — the branch of AI that helps machines understand human language — and machine translation, which uses AI to translate text from one language to another, in 2013. The two founded Qordoba, an AI writing assistant, in 2015.
In 2017, researchers from Google and the University of Toronto published a research paper titled “Attention is All You Need,” which introduced the idea of Transformers in machine translation. Transformers are a type of neural network, a machine learning model that works like the human brain, transforming input into output by learning context and tracking relationships between words.
Habib and Alshikh decided to pivot from just language translation to translating business content into more usable content using this Transformer-based approach, and in 2020 they renamed Qordoba to Writer and began work on the LLM.
Writer has raised $126 million to date, including $100 million in a Series B round in September.
Investors include venture firms Iconiq Growth, Balderton Capital, Insight Partners and Aspect Ventures, investment firm WndrCo, and Writer clients Accenture and Vanguard.
Regarding future fundraising, Habib said, “There should always be more planning to do.”
