|
Listen to this story
|
Zoho is one of India’s premier success stories. Founded in 1996, this privately held company recently crossed $1 billion in revenue during a global recession. It even grew its revenue in India by 77% over the last year, and has announced its plans to expand globally by opening 100 more points of presence all over the world.
The company is making strides in research, having filed 25 patents in the past two years. How did the company manage to turn a profit when the biggest tech giants are engaging in widespread layoffs?
Analytics India Magazine reached out to Ramprakash Ramamoorthy, director at AI Research at Zoho, to know more about how the company scripted this success story. Ramprakash has been leading the AI efforts at Zoho Corporation since 2011, which has now grown to become main USPs of the company, handling 100 million requests a day.
Future investments
Zoho aims to be at the cutting edge of machine learning and artificial intelligence in the workplace, and have expressed their plans to double their investment for blockchain and AI. Speaking on this move, Ramamoorthy stated, “Our newer investments will help focus on firming up and honing the current technology offering to stay relevant, given how fast AI evolves.”
The company aims to focus on research in statistical machine learning, computer vision, and natural language processing. These verticals provide a huge value add for businesses, as they allow them to derive unique insights which can be translated into better business.
Zoho has identified fresh use-cases to help companies capitalise on their data. Some of these include a sales coach that will help sales engineers sell better, a what-if analysis engine that would help users run simulations and experiment on digital twin platforms, and AI provisions in even more vernacular languages.
However, as seen with their privacy-first approach, all of these advancements will always keep the user’s privacy in mind. Ramamoorthy said, “As always, we will ensure none of these come at the cost of user privacy.”
AI in Zoho One
Zoho One, the company’s primary business product suite, incorporates AI in various facets of its services. Some use-cases include Zoho expense, which uses OCR to identify receipts, and Zia, their virtual assistant that utilises NLP for conversational AI.
Moreover, due to its nature as a unified product suite, Zoho One can go beyond the problems faced by data siloing. Ramamoorthy remarked, “Our AI engine can use data across departments to offer contextual next best action for a given role. For example, to help solve an issue reported by a customer on Zoho desk, our customer service help desk software, the support engineer will be equipped with the history of the customer as well as the potential solution for his/her issue pulled up from the knowledge base or similar tickets raised previously.”
These optimisations are present across Zoho’s product suite and offer exponential benefits for companies using it. What’s more, Zoho also plans to increase the use-cases and verticals it deploys AI in. Ramamoorthy stated, “This is just one of the examples. We will relentlessly use AI to ensure process optimisation and efficiency maximisation across our enterprise software stack.”
Privacy-first approach
Their AI models are built with a specific focus on data privacy, and are able to generate accurate results even without personally identifiable information. This allows larger organisations to use Zoho’s tech to derive insights from their customers while keeping in line with global data privacy regulations. Ramamoorthy said, “At Zoho, we strongly believe privacy is more than just a feature and ensure a privacy-first approach in whatever we do — from building our own data centres to a homegrown AI stack.”
Moreover, they also have a specific approach to their AI models that allows them to remove all personally identifiable information before the model is trained. “Any data that we use to train the baseline AI model runs through our PII (Personally Identifiable Information) detector and all PII is summarily removed before the data is fed to train the AI model,” said Ramamoorthy.
They follow a very clean approach towards using data in the model, as there are additional checks and balances at every step to ensure there is absolutely no PII involved. “Once the baseline model is established and goes live, the next step for the model is to be personalised based on individual user behaviour – at this stage we completely anonymise the process, thereby ensuring no personal information ever gets used to train or personalise the AI engine,” Ramamoorthy explained.
They also adopt industry-best practices when it comes to handling datasets and information. Ramamoorthy clarified, “We treat data the same way we treat our source code repositories. Our practices include proper and periodic reviews of access controls, maintenance of access logs and version control, and more. All data is processed only in the corresponding data centres, thereby ensuring our AI models are in sync with the local data storage requirements.”
This privacy-first approach also ensures a better model is trained, as it is largely free from bias due to the omission of personal identifiers. Considering the approach Zoho is taking, it is no wonder the company is thriving. From their India-first approach, to its privacy-focused AI algorithms, to their laser-sharp focus on customer dedication, the company is setting the bar high for its competitor.
