40% of our new engineering hires will be based in India: Notion CEO

In an interview, Ivan Zhao, founder and chief executive of Notion, said going forward, 40% of the company’s future global engineering hires will be based in Hyderabad.

“We’re focusing on hiring a lot more here. In fact, Hyderabad is our only engineering centre that is outside of the US,” Zhao said. The centre in Hyderabad was set up following Notion’s acquisition of workflow automation platform, Automate.io in a cash and stock deal in September 2021. The latter’s services and operations were fully integrated into Notion on 31 October, according to Automate.io’s website.

According to Zhao, Notion presently has 35 employees in India, a figure that he aims to increase in the near future.

As a platform, Notion offers document writing, database and knowledge management, and project management as three services unified under one platform. Zhao claimed that this is what differentiates the platform from other real-time workplace collaboration and management platforms such as Asana and Slack, or even Microsoft Teams.

But, unlike its peers, Notion has leveraged its popularity on TikTok amid the pandemic to create communities of consumers globally, including in India.

“Nearly 70-80% of our user-based communities are outside of the US. India is one of the sizable communities for us in this regard; while it is still not in the top five among the global communities, it is one of the fastest growing ones among our top 10 or 15 geographies,” Zhao said.

These communities have helped bolster Notion’s user base. Zhao said while the company presently reports “over 20 million monthly active users,” the actual figure is “much more.”

However, he declined to share the split between consumer and business users; only stating that Notion’s revenue is “predominantly from B2B operations.”

“A lot of our users are B2C, but our revenue stream is predominantly from businesses. It is in around 2019 that we made our ‘personal user’ plan almost completely free, and this includes limited storage as well. We serve startups, mid-sized companies and large enterprises. Our product is given for free, or for really cheap, to consumers. This helped us build a good community of users around the world, who then took our platform to their workplaces, which helped expand our business,” he said.

Describing the platform as being “made of Legos”, Zhao said that users are today leveraging the flexibility to build their own solutions such as document templates on Notion. “Last year, we had a number of users who earned over $1 million within a year, just by selling Notion templates or apps that would work on top of our platform. On websites such as Etsy, we have tons of people who sell Notion templates,” he said.

Zhao said Notion has leveraged this popularity to already become profitable even though he claims to have not spent “most” of the company’s venture capital funding. The latest round took place in October 2021, when it raised $275 million in a funding round led by Coatue Management and Sequoia Capital. The company was valued at $10 billion at the time.

The company is gradually using this funding to expand its implementation of generative AI — which it rolled out last month under ‘Notion AI’.

“We work with vendors to provide the AI tool, of which OpenAI is one. We use multiple off the shelf language models, and tap the existing data on Notion to feed into the AI models. By doing this, Notion AI will understand your previously written documents as well as ongoing projects, to give results that are very specifically tailored to what one particular user is writing. This is better in comparison to a generic platform like ChatGPT, which does not have context of one particular user’s prior work,” Zhao said.

To be sure, use cases and applications of generative AI gained prominence after the launch of Microsoft-backed OpenAI’s ChatGPT chatbot last November. The tool, which amassed over 100 million active users within the first two months of launching, has since seen a host of rivals and compatriots shape up, including Google’s own language model-based Bard, and Microsoft’s ChatGPT-based Prometheus Model powering Bing AI search.

The technology has seen widespread interest from all quarters — on 16 April, Mint reported that India’s top colleges are undertaking generative AI research to solve specific use cases within academia and research. Another report on 18 April detailed how India’s bellwether information technology (IT) services industry is racing to build generative AI capabilities for clients across financial services, healthcare, retail and other industries.

It is the latter aspect, coupled with the much-debated aspect of trust and authenticity in the source material of a generative AI tool, that Zhao said Notion is focusing on.

“Notion today is one of the first popular instances of AI embedded in existing workflows and within a writing software, made accessible in a platform where users are actually working. What works for us in terms of AI is that people already trust the information and knowledge they are feeding to the AI in Notion, which is an advantage in comparison to larger competitors,” he said.

This, however, comes with its own challenges. According to Zhao, these include the cost of AI engineers, as well as the cost of vendors that sell the large language models and AI algorithms behind generative tools. For Notion, OpenAI is one of its vendors, and while Zhao said that the company is also working with multiple other vendors, he didn’t disclose their names.

Beyond this, one key challenge for Notion also lies in finding the right way to implement AI.

“Integrating AI well takes a lot more, and depends on the quality that you want to create within a workflow. AI is not free, so this needed us to figure out a new pricing model where we cannot give all the features to all users. There is also a user experience problem in integrating AI well — if you want just AI answers, you can simply go to ChatGPT or Google Bard. But, if you want AI in a tool that you’d use everyday, you’d need to figure out clever user interface or architecture trade-offs, which is a lot harder,” he said.