Zoho Corp on Friday launched a comprehensive enterprise resource planning (ERP) solution, marking its entry into the multi-billion-dollar ERP market dominated by domestic and global players such as SAP, Oracle, Ramco and Tally Solutions. Founder and Chief Scientist Sridhar Vembu tells Narayanan V about the rationale behind the move, plans to develop Kumbakonam as a talent hub, and the impact of AI on businesses. Edited excerpts:
What prompted your entry into ERP from here?
I was born about 15–20 km from Kumbakonam, and this region is special to us. We started our operations here in 2020 and currently have around 50 people across two centres. We want to develop Kumbakonam as a hub for our ERP talent. ERP is a large market with a massive global opportunity of over $300 billion. We also see this as an opportunity for the region to flourish. We are investing in building capability and a talent pool here. Our current plan is to scale this up to about 2,000 people, depending on how the market accepts our ERP offering. We are optimistic and confident. Once we scale up, the centre will focus exclusively on ERP, which itself is a very broad suite.
How is artificial intelligence (AI) impacting your business?
I would focus on one fundamental area where AI is being used most heavily—software development. In an interesting way, AI is helping automate programming itself and, by extension, software. There is this whole concept of “vibe coding” that is evolving. I don’t like the term. I call it AI-assisted Code Engineering (ACE), because you cannot pile up a bunch of bricks and call it a building. You need engineering, architecture, structural integrity, safety, appearance and usability. The same applies to software. AI is very good at generating code, but you still need engineering. You need to guide it properly and vet its output. That requires human judgement and domain expertise. Right now, large LLM players have a big advantage, but I believe smaller models will catch up. Currently, for experienced players, using these tools costs anywhere between $200 and $1,000 a month. Some companies are paying $1,000 per month per programmer. It’s insane.
The second issue is code quality. We have to vet everything and cannot blindly accept what AI produces. The foundational AI models themselves have become a FOMO game, with hundreds of billions being thrown at them. I don’t believe we can compete head-on that way. The Chinese have shown they can compete using cheaper technology, and we have to approach it similarly.
It’s not just about foundational models. In software development, there are other tools that, when combined with AI, can boost productivity significantly without being so costly. That is exactly the area we are investing heavily in India. The fruits of this will be visible in a productivity leap in software development. That is how we see ourselves launching a lot more software, much faster. We are going to make that leap this year.
You often highlight the risk of AI slops.
This is a serious issue, even in code. AI can make mistakes, including subtle ones that are hard to detect. That’s where the AI-assisted engineering part comes in. There has to be a human review layer, but there also needs to be automated tooling outside of AI to ensure this. So far, these tools are in a nascent stage and haven’t come in a big way. I believe they will become far more important over time.
You scrapped the semiconductor fab plan last year. Are you still looking at this space?
We are investing heavily in chip design. We have two companies invested to build chip design capability. We also have a nascent effort called Tenkasi Semiconductor, which is focused on chip design. This is not a fab. The chips we design will still need to be fabricated somewhere. We hope to use the Tata fab coming up in Dholera, Gujarat. Otherwise, today the option is to go to Taiwan. Those are the two choices. We do not have a current plan for setting up a fab ourselves.
That is a very different game. Even Intel, with its vast experience, is struggling to keep up with TSMC. That shows how difficult this space is and why deep, long-term R&D is required. ERP is not easy either, which is why only a handful of ERP players dominate globally. Email is also not easy—that’s why there are only three real global players, and we are one of them. All of this requires sustained R&D over decades. A similar long-term commitment is needed in semiconductors. We are doing chip design, but we are not doing a fab right now.
Arattai hasn’t taken off despite end-to-end encryption.
People write these one-month premature obituaries. That’s not how these things work. I don’t care about it. By that logic, Zoho should have been declared dead about 500 times. We started Zoho in 2003. Obituaries were written in 2007, 2010 and again in 2015. We have made a good beginning and there is still a lot of work ahead. Every week, updates are going out. You will see a surprising number of features every two to three months, and the adoption curve will steadily go up. There is also pent-up demand in India. There is a hunger for made-in-India software, which is why we want all this software to be built here.
(The writer was in Kumbakonam at the invitation of Zoho)