As fears develop over AI instruments shifting past helping software-as-a-service (SaaS) firms to changing them solely, Salesforce representatives attending the AI Affect Summit this week, pushed again in opposition to speak of a looming ‘SaaSpocalypse’ and mentioned that even main mannequin builders depend on its enterprise platforms for mission crucial capabilities.
Earlier this month, the launch of AI startup Anthropic’s new suite of office automation instruments despatched shockwaves by means of world expertise markets, together with Indian IT shares, with massive names reminiscent of TCS, Infosys, Wipro, and HCL Applied sciences nosediving 4-7 per cent and the Nifty IT index plunging practically 6 per cent, the largest single-day fall since COVID-19.
The episode was termed ‘SaaSpocalypse’ in reference to SaaS firms dealing with potential obsolescence because of AI disruption. Nonetheless, the Slack father or mother has argued that it’s going to nonetheless see advantages and alternatives on this new panorama.
In an interview with The Indian Categorical on the sidelines of the AI Affect Summit 2026 in New Delhi, Arundhati Bhattacharya, the CEO of Salesforce South Asia, and Srini Tallapragada, the corporate’s chief engineering officer, make a case for the SaaS enterprise mannequin within the period of AI-driven automation. The dialog additionally lined shifting past AI pilots to full-scale manufacturing, sustaining AI governance and moral requirements, the rise of vibe-coding instruments, and the Indian authorities’s envisioned UPI-like sovereign AI stack for MSMEs and others.
Edited excerpts from the interview:
Q: One of many key points on numerous minds proper now could be whether or not massive SaaS firms are in danger as companies experiment with constructing inner AI instruments or deploying fashions instantly, in some circumstances bypassing conventional software program interfaces altogether. What’s Salesforce’s reply to this?
Tallapragada: The ‘SaaS is useless’ narrative is counterfactual. It’s very laborious to show or disprove. We work with numerous clients, together with mannequin distributors like Anthropic and OpenAI that not solely use our merchandise like Slack but additionally our CRM [customer relationship management] merchandise.
Software program firms work on what’s core to them. We use Workday for HR software program. I’ve numerous engineers however can I vibe code away this software program? No. To start with, it’s not our enterprise. Why would I waste my time on that? Secondly, it’s not sufficient to simply do model one. It’s a must to consider all of the regulatory facets and different particulars.
Story continues beneath this advert
Lots of our firm distributors are going to make use of mission crucial techniques like Salesforce’s choices. So, this narrative that ‘SaaS is useless’ is generalised. However SaaS goes to evolve. Like some other trade, in case you’re not including worth to the client, you’ll be disrupted.
Bhattacharya: The best way we do issues goes to vary. However does that imply we ourselves will now not be within the equation? I don’t assume so. As a result of when you find yourself working an organization, an LLM just isn’t the one factor that’s required. It requires context, reasoning, governance, and auditability.
We’re utilizing brokers ourselves, each internally and externally. We’ve got practically 20,000 clients in varied phases of implementation, 50 per cent of them are refilling their tanks with no matter they’ve already purchased. A variety of our clients are saying that they’re seeing precise worth. So, whether or not we’re actually going to be right here or not, after 10 years you possibly can come again and ask me that query.
Q: Can India’s $300 billion IT sector adapt by forming deeper partnerships with AI firms and constructing AI-native options? And who is healthier positioned to guide that shift?
Story continues beneath this advert
Tallapragada: It comes right down to management and willingness to vary. We’re working with numerous our System Integrators (SI) companions, they usually’re additionally altering how they ship providers. A few of our massive SIs in India are implementing AgentForce as a result of they wish to re-imagine how they ship providers. Most of them perceive that AI is foundational, and in case you don’t change the best way you’re delivering providers and outcomes for patrons, you’re not including enterprise worth and also you’ll get disrupted.
Q: Is there a deployment hole between the ability of AI and the capability of companies to make use of it? And may firms like Salesforce play a significant function in closing the hole?
Tallapragada: There’s a hole between the mannequin distributors and the hype machine speaking about AI, versus the actual enterprise worth clients are literally seeing of their revenue and loss (P&L) assertion. Lots of people have assumed that merely rolling out a mannequin is sufficient. “I purchase a mannequin, I give ChatGPT to all my workers, and declare victory.” They thought that was ample, however now they’re all realising that it isn’t sufficient.
What we’ve centered on is ignoring the hype and dealing carefully with our clients. We now have hundreds of shoppers utilizing our merchandise and demonstrating actual enterprise worth. We do that trade by trade so we will share greatest practices. For instance, in aviation, we work with Southwest Airways, Finnair in Europe, Singapore Airways, and Air India. Once we converse to the following airline, we will level to those leaders and say: listed here are one of the best practices, here’s what works.
Story continues beneath this advert
Firms can attempt to do it themselves, which many haven’t been profitable at, or they will use our platform and our vertical experience to generate measurable enterprise worth.
Q: Based mostly in your distinctive vantage level, how would you characterise the speed of enterprise AI adoption at present, each in India and globally?
Tallapragada: I name final yr pilot purgatory as a result of individuals did numerous pilots and demos of AI however nothing modified. I believe CEOs are drained. They’re saying, “Hey IT groups, we’ve invested quite a bit. Now present me worth.” That is excellent news as a result of enterprises are realising that they want good companions like Salesforce.
We’ve got one of many largest implementations in comparison with anybody. We acquired greater than 30,000 clients attempting to make use of AI instruments and brokers. Nevertheless it’s only a drop within the bucket. It’s nonetheless not subtle throughout sectors. This would be the yr of worth. It will likely be the yr AI goes from pilots to ROI.
Story continues beneath this advert
Q: Many firms cite points like hallucinations, reliability, and safety dangers as limitations to deployment. To what extent does enterprise adoption rely upon additional technical breakthroughs in AI analysis?
Tallapragada: There are numerous mitigating components. LLMs are commodities and there’s not a lot differentiation between one mannequin versus the opposite. They’re all going to proceed to enhance. However there’s nonetheless sufficient worth available, and that’s why enterprises ought to deal with what’s the worth you will get from AI at present.
Q: At a current press convention, IT Minister Ashwini Vaishnaw mentioned that the following part of India’s AI mission will deal with a UPI-like platform with a bouquet of ready-to-use AI options in key sectors for MSMEs. Do you assume it is a viable answer?
Bhattacharya: There was dialogue round constructing a public rails sort of platform, on prime of which personal gamers might construct their stacks. If it occurs, we are going to assess whether or not that’s one thing we wish to take part in. The concept is there, and I’m positive it’s possible. However we must wait and see the way it truly takes form.
Story continues beneath this advert
Q: Do you see the rise of vibe coding instruments as a aggressive risk?
Tallapragada: Traditionally, Salesforce has been a low-code platform as a result of when clients purchase one thing, they wish to get the quickest worth. I believe vibe coding instruments will take off in an even bigger manner as a result of it reduces time to worth. As a substitute of taking 30 days to see outcomes, firms can realise worth a lot quicker, after which construct much more on prime of it.
However on day three or day 4, at 2am in the course of the night time, when one thing goes improper and an engineer wakes as much as a pager alert, they can’t be one another and saying, “Who wrote this code? Some AI wrote this.” That isn’t what enterprise clients anticipate. So what we constructed is a device that follows the identical guardrails round safety, governance, and privateness, so firms can construct safely.
Q: Ought to enterprise AI be regulated in another way than shopper AI merchandise and instruments?
Story continues beneath this advert
Bhattacharya: Particular person international locations must determine their method. It’s nonetheless early within the cycle, so many try to take care of a light-touch framework. However with lighter regulation, it turns into much more vital for firms to apply robust self-regulation.
Since 2014, Salesforce has had what it calls an workplace of humane and moral use of expertise. All of our merchandise and processes undergo that workplace to make sure we aren’t doing something dangerous. We additionally make investments closely in monitoring cyber threats all over the world. Throughout all our processes and merchandise, we construct in accountability and auditability in order that choices will be traced. You possibly can return and see who made which determination, why it was taken, and which agent did what.