Tech firm Salesforce, which employs more than 9,000 people in India, is betting big on generative artificial intelligence (AI), and even doubled its total global investment corpus on generative AI startups through Salesforce Ventures. Arundhati Bhattacharya, chairperson and CEO of Salesforce India, believes that the pricing of generative AI tools will be the key to unlocking its business potential. In an interview, Bhattacharya, who was the first woman Chairman of the State Bank of India before joining Salesforce, also shared her thoughts on a range of issues including the impact of generative AI on jobs and a hybrid physical office structure. Edited excerpts:
Salesforce has made multiple announcements including Einstein GPT, Marketing GPT, AI Cloud and more. But are businesses actually willing to spend on generative AI solutions, or are they only exploring the field?
I think everyone’s only experimenting with it right now. Also, they don't even know what the pricing is going to be like, because the pricing of generative AI tools will not be as simple as OpenAI’s ChatGPT — they’re giving the service away at throwaway prices. Pricing is something that companies will have to understand before they really agree that generative AI is the way to go.
In India, we are basically a low-cost centre. If putting 20 people on the job will get you the same results as ChatGPT, would you opt for the people, or the tool? We don’t know. Hence, this is a phase where everybody is trying to discover what the way forward is.
We ourselves, at this point of time, haven’t yet come out with the pricing. When you’re using trillions of data points, and a large language model (LLM), the computing power that is required is also costly.
So, what’s your advice for CXOs who ask you for generative AI deployments?
I tell them that these technologies are getting developed, and that we should be ready for it by ensuring that we have the right kind of data cloud and structure, to ensure that these LLMs work.
Organisations also have to have a certain amount of digital maturity to be able to be successful while using generative AI unlike individuals, since you are trying to get intelligence out of data in a manner that will give you deep insights.
For that, you need to have your data in order. So, the first step there is to ensure that you have your data, and we invariably say that if you’re already on a data cloud, that will definitely be a big help. That’s a preliminary take on where you need to be in your digital journey.
Where do you sense the resistance to generative AI?
Generative AI gives people who speak plain language the power to do things that they never had before. But people are also realising that generative AI can also hallucinate (give wrong output), and therefore, they have to be very careful about the prompts that they give. Even then, one needs to check for accuracy.
Also, enterprises hold data of other companies based on trust. Will they be able to maintain that if they adopt a tool like this? Enterprises also have to be sure that their data remains private. This is important because companies don’t have mandates from their customers to allow their data to flow all over the place, in case of LLMs (large language models) training on it. Even regulators won’t allow this. So, they’ll also need to be reassured that their data will remain their own, while they use AI models. Privacy is thus of great concern.
The second question is pricing, and the third is liabilities. For instance, if a tool offers a wrong prediction, who will bear the liabilities? How will enterprises know that a model has not been trained by someone else’s intellectual property?
What impact, according to you, will generative AI have on jobs, and how should this issue be addressed?
Every time that humanity has taken a leap forward, there has been fear that there will be massive job losses. But every time that we have evolved, new types of businesses and new roles have come up, and we have still accommodated.
No matter how far generative AI progresses, I feel that in things that need creativity and emotional intelligence, these are areas where human intervention is still required.
Now, how do you upskill yourself for it? The only thing that I can tell you is that you need to stay with the times and keep looking at the ways that are evolving for keeping yourself up to date with the skills that are required. For instance, a field such as prompt engineering has come up. Will everyone be converting to that? Obviously not, but there will be people who will be required in this field, such as an AI ethicist.
And do you believe that machines will eventually become sentient? Does the thought scare you?
This is mostly what scares humanity more than anything else, and it scares me too, because I don’t know how these things are progressing, and they’re progressing fast. I know that there are small Indian companies that make small toy robots. They’re doing this to teach youngsters as well as people that have special learning needs. These companies are actively working on emotional intelligence as a part of AI. Now, while that may be very helpful, at the other end it is scary as well. I don’t know whether that (sentience) is going to happen or not.
Talking of India and generative AI, are you happy with the pace at what India is progressing in this regard — be it Niti Aayog’s framework, the new Digital India Act, etc.?
If you ask me about the IT sector and not only AI, I’d say that India has done a remarkable job in the India Stack. I don't think there’s another country in the world that has anything similar, which truly ensures inclusion.
When we started with Jan Dhan Yojana, that was the foundational layer to this Stack. What it has done is created a kind of inclusion that is rarely seen. With the way that small businesses have evolved today — there are people sitting across the country sourcing say exotic vegetables from cities such as Pune and Bangalore. These kinds of things were not thought of 10 years ago, they were considered almost impossible. Today, there is hardly a pin code in the country that is not covered by the kind of Stack that we have created. There, the government has done a great job.
If you talk about AI and generative AI, this is a space where we need to leapfrog. I have so far not seen too much usage of that in the Stack that we already have. This is mainly because the government isn't really dealing with data that much — they’re dealing with just creating the rails for the services.
They’re not looking at the data, they are just the facilitators for people who will ride on those rails and deliver the services. They (the government) have created the infrastructure, but they haven't created any way of getting insights, because they have respected the fact that data needs to be private. It is an individual citizen’s right.
On the government front, I have really not seen much usage of it, and rightfully so. I don't think they are wrong in that, and that's the way it should go. The government has done a lot to empower individuals and corporations — it has not done anything to look at the data themselves. In a way, it is good — though they can definitely anonymize the data, put it in a model, and create insights into it.
For instance, a model that offers insights into which part of the country is learning stuff better. And, if certain parts of the country are not, they can decide what needs to be done in order to get there.
In this context, do you believe that we can build an IndiaGPT?
The Prime Minister once commented that our Mars mission’s per-kilometer cost was less than the cost per kilometer of an auto ride in Ahmedabad. So, India has a reputation of being able to do enough jugaad of the right kind to do things at a far less cost. On top of that, we’re also a very data-rich country. To that extent, while I can understand where Sam (Altman) is coming from. (But) I don't think it (creating IndiaGPT) is beyond the realms of possibility.
There was a time when we felt that we didn't need to create semiconductor chips, and there were enough supplies. Today, we are encouraging the semiconductor industry actively to ensure that the chips are made here. It’s better that we at least start and fail, than not do it altogether.
You took charge as the first CEO of Salesforce India on April 1, 2020. How has moving from banking to technology shaped your leadership style and strategic thinking?
When I took on this role, I decided that this is a good segue because I’ve been interested in the information technology (IT) sector for long. The move has turned out better than I expected — with the transformative changes coming in AI right now, and the way it’s going to inform every activity that we do, is exciting. I have a ringside seat to it, and from here, you wonder what would be the future of work, and how things will pan out. Had I not been in this position, I’d have missed out on a lot.
In the banking sector, I struggled to get people to accept a work from home culture back in 2015. I wanted women, who needed to be at home, to be able to work from home. We even created a platform so that they could do it. But, there was an enormous pushback — people felt that if you allowed people to work from home, productivity would suffer, and nobody would want to work.
Since then, we’ve busted so many myths. Having been a big part, it’s exciting to see workplaces transform into people who are remote, and people who wish to come to office but with flexibility in timing and location. Now, with generative AI at work, we will see how it's going to impact our decisions and communications. We’ll get to see how we’ll add value to jobs, and look at efficiency and productivity at workplaces.
During the pandemic, every CEO seemed to adopt work from home as a norm. Now many are asking employees to return. What is your strategy?
For us, it is still a hybrid working structure. For our sales team, we’re not fussy about whether they work from home, or office. But there are two things that we’re definitely stressing upon — one, you need to be with your customer physically. This may not be possible every time, but definitely needs to be done whenever there are difficult conversations to be had.
We expect our sales teams to get together physically in office when they’re discussing strategy, rehearsing pitches, and conducting debriefing post customer calls. This is because there are a lot of insights that don't come through in structured remote meetings, since there are no side comments and responses drawn in the way you would see within an unstructured meeting in a physical space.
But what about those employees involved in making products?
On the product side, it’s important to be present physically when employees are brainstorming. However, when they are reviewing code and other similar tasks, they can just as well be at home. One of the reasons for this is because while coding, programmers sometimes need two screens, and it's not feasible to carry such setups to the office and back at all times. They can also have a similar setup in office, we’re open to that.
In our main development centres in India — Bangalore and Hyderabad, we have a lot of youngsters that don't have any families in the city. They are more than pleased to come into the office. Sometimes, even those with families prefer coming to office if they’re living in multi-generational homes. They may have a working spouse, children and parents, which is why having enough space to work freely from home is difficult.
People without families also prefer the social interactions at the office, plus that they don't have to cook when they come to the office. We don't insist on too many at-office days. But, during brainstorming, they're expected to be there. That is why we don’t expect them to be fully remote. For instance, there would be very few cases of, say, someone enrolled in our Hyderabad centre working from Kerala.
There have been multiple reports of mothers not willing to rejoin the workforce, if the clause of working from office is made mandatory. Is your strategy helping working mothers?
As I said, if employees want to be fully remote, which means that they are not within commutable distance from the office, those cases are exceptional. It's not that we don't have remote employees — we do. But they would really need to be very valued employees, to be allowed to be fully remote. But even the remote employees will need to come in at least once a quarter, because there are so many updates and learning opportunities. More than that, the team connection is also very important. To that extent, for working mothers, we’re fine with them staying in the city of our offices, and coming in every once in a while. But, if it has to be totally remote, we have to go through a process and make sure that we are comfortable with it.