The news that Grammarly has raised $200m at a $13bn valuation is proof (pun intended) that writers’ technology tools remains a hugely valuable niche.
Unusually, Grammarly says that it has been cash-flow positive for years now so the funding, partly from giant VC BlackRock, is an opportunity to build out its engineering and channels to market. It comes as the Ukrainian-American company launches Mac and Windows desktop versions to supplement its current free browser extensions. The overall Grammarly freemium service includes features such as grammar, spelling checking, tone analysis and plagiarism detection. Only available in English, Grammarly currently claims to serve 30 million people.
An SDK should also help Grammarly on its mission to be embedded wherever people draft words from emails, messages and Slack boards to presentations, Microsoft Word and Google Docs. And that’s some mission. Despite the rise of vlogging and social audio apps such as Clubhouse, our command of language remains out best chance to persuade, command and cajole. Embarrassment over an inability to create powerful (or even clear) prose remains common, hence the demand for copy at the behest of everyone from CEOs to social media managers.
Writing tools have a long history from the outlining functions and grammar and spelling checkers in word-processing, software-assisted translation (led by companies such as SDL), speech recognition programs like DragonDictate, predictive and auto-correcting smartphone text and Otter.ai, the popular add-in for Zoom transcripts. Still many non-professional writers (and some professionals too) continue to be concerned about their ability to write coherently and stay on the right side of the self-elected grammar police.
Grammarly says its differentiation lies in its AI-infused approach to help people “say what they really mean”. That’s a bold aim but prose can all too often be lost in half-remembered “rules” and vague notions of “correct grammar”. Automated tools rarely have an ear for good writing, preferring to let engines that rule out split infinitives have their way. By those lights we would never have Star Trek’s evocative and musical “to boldly go where no man has gone before” and instead we would suffer the clunk “to go boldly where no man has gone before”. But, to be fair to Grammarly, I like the way it nudges writers to actions rather than being prescriptive and the automatic suggestion facility probably suits a generation brought up with autocorrect.
"Brilliant writing awaits," trumpets Grammarly’s website. Maybe not, but functional prose and decent advice? Yes. A foundation in AI should see Grammarly set fair for long-term improvements but parsing the English language, with its roots dating back well over 1,000 years, remains a task for the ages.
Here is a short Q&A interview conducted over email with Rahul Roy-Chowdhury, Global Head of Product at Grammarly.
Q. How did the company start?
A. Alex Shevchenko, Dmytro Lider, and Max Lytvyn founded Grammarly in 2009 to help people write effectively, with the help of technology. Max and Alex had previously launched a plagiarism detection service that was acquired by education-technology company Blackboard. Brad Hoover joined Grammarly as CEO in 2011, when he came across the company as he was searching for a service to help him with his writing.
The first version of the product was a subscription-based service that focused primarily on grammatical error correction. Working at the forefront of natural language processing and new technological capabilities, we continued developing a product to offer more intelligent writing suggestions available in real-time. Recognising that effective communication goes far beyond proper grammar, we’ve long since expanded our product capabilities to offer guidance on further dimensions of effective communication, such as conciseness, consistency, vocabulary variety, tone, inclusive language suggestions, and much more.
Following the success of our consumer offering, we began seeing demand for enterprise-grade writing assistance among workplace teams, many of whom were individually using Grammarly at work already. This directly drove the launch of our enterprise offering, Grammarly Business, to help organisations accelerate business results through more effective communication. Over two-thirds of the Forbes Global 2000 have at least one Grammarly user!
Today, Grammarly is an indispensable communication partner for over 30,000 teams and 30 million people who use it daily, serving wide consumer and business use cases at any scale. Some additional stats include:
- Delivering 1.2 trillion writing suggestions to our user base worldwide in 2020 alone
- Supporting over 6,800 non-profits and NGOs from over 150 countries with a free offering
- Serving over 2,500 educational institutions through Grammarly @edu, with capabilities tailored to their needs
How do you differ from previous generations of grammar, tone and spelling checkers?
When it comes to grammar-checking in particular: In our 12-year history, our grammatical error correction remains unmatched on the market. But Grammarly goes much further than superior grammar-checking. Grammarly’s AI-powered writing assistance includes a plagiarism checker and real-time guidance for multiple dimensions of communication, including:
- Correctness—writing mechanics (spelling, grammar, punctuation, updated terminologies and usage)
- Clarity—clear writing, conciseness, and readability (helping people avoid overly complex writing and communicate more coherently)
- Engagement—vocabulary, variety, and engagement (helping people avoid bland word choice and repetitive sentences)
- Delivery—tone, inclusive language suggestions (helping people find the right tone when addressing their particular audience)
In addition, we know enterprise technology tools must meet the highest standards for data security, privacy, and confidentiality. That’s why we don’t just check the box on security — it remains our number one priority and most important feature. It starts with our business model: We make money when customers pay us for our product because they find value in it—not by selling user data to advertisers. Our writing assistance is backed by stringent, enterprise-grade security and privacy standards, including SOC 2 (Type 2) and SOC 3 reports, as well as certifications from the International Organization for Standardization and HIPAA compliance.
Grammarly also recognises the impact of the pandemic on individuals and organisations worldwide. As we continue to interact in more ways and places, effective and empathetic communication is so important. We recently invested in several new developments to solve more communication challenges:
- Advanced features for professionals, becoming the first platform to trigger full-sentence rewrites automatically
- For Grammarly Business: Snippets (preset response templates), brand tones (company tone profiles), multiple style guides (up to 50 in a single organisation), and enhanced analytics. These features make Grammarly Business the first and only writing assistant helping entire teams and organisations stay on-brand with consistent communication
- Helping people communicate more empathetically by expanding our support for inclusive language. For example, we released a suggestion that flags xenophobic terms for COVID-19. Appearance of the suggestion decreased by 90% year over year, which suggests people who were served these empathetic reminders were more cognizant of how they referred to the virus over time.
How has AI helped?
This blog post provides context around some of our investments in NLP, ML, and linguistics.
We’ve built and continue to invest in machine learning infrastructure to leverage the latest developments across techniques like Transformer-based sequence-to-sequence models, neural machine translation, and massive pre-trained language models. We’re designing solutions that address every aspect of communication, which means applying a combination of technologies—and always building with humans in the loop.
Our AI continues to learn based on analyses of billions of daily writing sessions. We checked 14 trillion words in the past year alone, deliver 100 billion writing suggestions each month, and improve 20 billion communications per week. All of this is powered by a system that relies on human knowledge—mainly through linguistic patterns—and large-scale machine learning approaches. Our NLP team has been engaging in applied research in this area for over a decade, and the unrivalled intelligence of our algorithms reflects our longevity in the field.
What are the biggest technical challenges you have faced in developing Grammarly?
Communication is incredibly complex and really important to get right. That’s why Grammarly’s one and only mission is to improve lives by improving communication. We think of communication as a lifecycle of thoughts and ideas to accomplish a shared goal. It starts with conceiving an idea that you express by composing your thoughts in writing. You make that writing better by revising your first draft before sharing it with your audience. On the receiving end, you read some writing that somebody shared with you, trying to comprehend the key points and takeaways. Ultimately you react to the writing by responding in some form, starting the lifecycle again.
We exist to provide communication assistance across all stages of the communication lifecycle. To achieve this, we need to explore a broad set of technologies. At the core, our technology needs to consider the entire context of a message, which is a huge technical opportunity— whether that’s helping you find the right tone, expressing yourself more clearly, or sounding more fluent as a non-native speaker.
As our technology evolves, we also have a responsibility to build a product that augments our users without perpetuating biases or stereotypes, and this is reflected in our commitment to developing our communication assistance responsibly. For example, we have internal teams and structured workflows that are focused on evaluating inclusivity and sensitivity in our product ecosystem, including our machine learning models. Some examples of actions we take to guide our product in being inclusive and respectful of our diverse user base include thoughtful approaches to dataset annotation, feature pre-release assessments, and post-release feedback. Respecting our users is at the heart of all these efforts—we listen to the needs and real-time input from them and are quick to action necessary changes within the product.
One criticism of such tools is that they treat grammar as a fixed entity, whereas many writing experts today view the development of English as a set of conventions that have been honed and modified over time. Do you agree that is fair and will Grammarly reflect that there are sometimes better ways we can express ourselves even if those are forbidden by certain rules?
We believe that technology should help people think better, not think for them. Responsible innovation begins with a simple question: Does our technology aim to empower people—or replace them? Grammarly aims to help our users learn to be more empathetic and effective communicators, rather than communicating for them. Our product’s suggestions are options for our users to consider; they are not automatically enforced.
That is a crucial difference and one that sets Grammarly apart. While our product is driven by AI—and uses data and analyses from our natural language processing team—we always build while keeping in mind the person typing at their keyboard, trying to say what they mean. We take very seriously the responsibility of making a tool providing support and assistance—enhancing each person’s own style, character, and technique.
Our writing assistant is an intuitive, real-time interface that suggests improvements to people for consideration. Grammarly includes explanations of the grammar rule or language practice influencing the suggestions that our users see. When reviewing our technology’s suggestions, people learn more about grammar and language, and in turn improve their own understanding over time.
Creating a product like ours necessitates extreme empathy for our users and those with whom they communicate. We are extremely mindful that our product works with language, which is constantly evolving. Over the years, we have attracted one of the largest teams of dedicated linguists and scientists who obsess over how language evolves. They collaborate with product managers and engineers to bring a variety of suggestions to life in our product that makes our users’ writing as impactful as possible.
All the while, we humbly recognise that the linguistic challenges are plentiful, and there is always room to improve and evolve our product to more effectively serve our growing user base. We look at these challenges as opportunities to consistently improve our product to meet the needs of our users now and in the future.
I think the new round takes total funding to $400m. Given that you’ve said you are cashflow positive already, what will you use the funding round for?
