What once seemed straight out of a sci-fi movie is becoming mainstream for businesses of all sizes. Artificial intelligence (AI), particularly generative AI, has, as noted in the most recent annual McKinsey Global Survey on the current state of AI, “risen from a topic relegated to tech employees to a focus of company leaders.”
Ram gave us a fascinating behind-the-scenes look at how AI evolved at Zoho and how important AI has become for small businesses.
Rieva & Brian: How long have you been the director of AI research at Zoho?
Ram Prakash: Since 2016.
R & B: You’ve worked on AI research for seven years. Many people would think only a few weeks because ChatGPT suddenly showed up last fall, and the whole world got flipped on its end.
So, what were you doing with AI research at Zoho back in 2016, and how has that evolved to where you are today?
Ram: My interest in AI goes back to 2011. When I graduated from university, I was an intern at Zoho. That was when the notion of big data emerged, and the whole ecosystem changed. You had the big cloud, small apps, and the concept of APIs, and all of these blew up. This was when smartphone apps launched.
At that time, Zoho wanted to see if AI was just about fancy tech or if it would have any role in business. So we started with use cases in 2011. I was a one-member team for the first four years, doing many experiments to see if we could get some data.
This was the pivotal period when we set the base for an AI platform. We were trying to do sentiment analysis in Zoho Desk (back then, it was called Zoho Support), our customer support platform. And the training data that we usually got from sentiment analysis was mostly Amazon’s customer review data set or IMDB’s movie review data set—both heavily inclined towards the B2C world.
We started orchestrating all the data points so we could do analytics on top of them. Then we brought in machine learning (ML) which looks at past data, identifying trends and seasonality, etc. And we did a very soft launch—our customers didn’t even know we were using AI.
We saw how it evolved, and around 2016, we started telling our customers we were using advanced statistical techniques and advanced machine learning techniques to help them.
And that’s when we decided AI would make it big in business. We always have these [experimental] efforts going on, but it takes a lot of time for the technology to mature and become mainstream. Today we have AI across our product suite in almost every product and every significant touchpoint.
R & B: How long were you just moving along with AI, doing experiments, gathering data, testing products, when suddenly, everything changed with ChatGPT?
Ram: AI has this unique trait, dating back to 1950, where it always over-promises and under-delivers. But now we can generate data at a very minimal cost.
So we can collect, record, and store data in the cloud. And, thanks to the infinite computing power available there, we can process that data at scale. All this combined has made it the right time for AI to make its presence known.
Like any other technology, AI has been incubated well in the consumer world. And when it hits maturity, business catches up and has the late-move advantage. Five or six years ago, we didn’t have a lot of enterprise mobile apps, but today a lot of work gets done on mobile.
Six years ago, we had many AI use cases. For example, you could talk into your TV remote, speak to your smart speaker about the weather, or get shopping recommendations. Today, that ability has come to the enterprise.
And last fall, we saw generative AI make a mark. It started with people generating a lot of images. Then it rolled into a chat interface. And all of a sudden, it started writing beautiful summaries. And today, the over-promise, under-deliver cycle has ended. AI fits into mainstream technology.
AI is no longer going to be a differentiator. It will be the minimum you will have in any software product.
R & B: Which brings us to today.
Ram: Zoho started integrating Zia, our AI assistant, into almost all our products, ensuring a lower entry barrier into AI. But Zoho is not an AI company. We are retrofitting AI into products that are nearly a decade old.
[And getting it right is essential.] Think about business owners who’ve used a program like Zoho CRM for years and know their market and customers well. All of a sudden, we start giving them predictions. But they’re sure those predictions are wrong. So we began adding explanations, and our usage increased by around 60% to 70%. And that is pivotal.
In the consumer world, if I recommend your read the wrong book, it’s not a big deal. But if I recommend something that disrupts your business, that’s when things get really bad.
Even today, a good AI model is 80% accurate. And giving them explanations really helped our customers embrace AI. And now, with the advent of generative AI, our AI feature usage keeps growing because everybody’s trying to see how AI can help them.
If you think about it, AI has been the most hyped technology, but it’s actually the natural evolution of software, which keeps evolving. When someone has a problem and turns to AI, the solution automatically becomes part of the knowledge base, which becomes the benchmark.
R & B: You said, when you first launched AI, people said, “Turn it off, turn it off. What is this?” But now they know, so what do small business owners do to adapt even more parts of AI? And what are some of the biggest obstacles or pitfalls that business owners should be aware of?
Ram: There is a lot of confusion around the privacy implications of AI and how you put that in mission-critical systems.
Another area is deciding where AI is the right fit. You have to determine if you want to deploy AI at the core of your business or in the allied areas and see how it can add value and where it can make mistakes. And are those mistakes tolerable? How resilient is your business to those mistakes?
Start deploying AI in allied areas. For example, a customer support chatbot is a good place to begin. And run things parallelly. Have your non-AI system where humans make decisions, and your AI system run at the same time. See where the overlap between the AI’s decision and the human’s decision is. And where the agreement percentage is high, automate that part with AI. And where it doesn’t agree, keep running parallel systems to generate more data.
The entry barrier to AI is still very high today. Find a software provider/vendor with AI baked into their products that offers a unified software platform where all the data is available for the AI engine to make decisions with. That’s a better choice than picking a siloed software vendor.
All businesses should care about getting AI into their systems today.
R & B: What are the next best steps for small business owners to learn more about AI and how it can help make them smarter, more productive business owners?
Ram: It’s critical not to see AI as standalone tech. The focus should not be on AI but on which processes in your everyday work can be augmented by using AI. Even today, AI is still very human-driven, meaning whatever is generating AI, a person is giving the prompt. And to provide the best prompt, your business data, past experience, and the wealth of information you have about your customers and your domain will help.
Proactively find the ways AI can help you. It is not a threat to your jobs. AI is here to augment your capabilities. It’s s only when you combine with AI that makes it “superhuman.” AI is here to help you—embrace it with open arms. See where it can help, see where it can optimize.
For businesses, the broader goal of AI is to optimize processes and maximize revenues, and what business does not want that to happen? So think of AI as your magic wand.
This article was edited for space and clarity. The podcast with Ram will post here shortly.