We’ve become a people-first AI service software to enhance customer and employee experiences: Priya Subramani of Freshworks
In a conversation with YourStory, Priya Subramani, VP & GM of Customer Experience Products at Freshworks, decodes the SaaS giant’s AI strategy in India, advancements of Freddy AI, and what’s next on the horizon for AI agents.
Agentic AI is revolutionising how businesses operate by introducing advanced intelligent systems that can solve multi-step problems on behalf of users. These systems, powered by AI agents, are transforming enterprise workflows, prompting businesses to automate processes.
As companies strive to enhance both customer and employee experiences, making use of such solutions has become crucial for tech players to stay competitive.
SaaS major Freshworks, which provides customer relationship management (CRM) solutions to over 68,000 business clients, has embraced this trend with the introduction of Freddy AI, an easy-to-deploy autonomous service agent designed to enhance both customer experience (CX) and employee experience (EX).
The AI-driven solution is already being adopted by companies such as Bchex, Porsche eBike, Hobbycraft, and Live Oak Bank, integrating with Freshworks’ platforms like Freshdesk and Freshservice.
According to a report by Gartner, by 2028, 33% of enterprise software applications will incorporate agentic AI, enabling 15% of daily work decisions to be made autonomously.
In a conversation with YourStory, Priya Subramani, VP & GM of Customer Experience Products at Freshworks, decodes the SaaS giant’s AI strategy in India, advancements of Freddy AI, and what's next on the horizon for AI agents.
Edited excerpts from the interview:
YourStory (YS): From a product standpoint, could you share the developments at Freshworks in light of the current AI trends? What is the product strategy for Freddy AI agent?
Priya Subramani (PS): AI isn’t new to us—Freddy has been around for years. But generative AI has completely changed the game by creating a landscape that is worlds apart from our earlier capabilities.
We have adopted what we call “people-first” AI. While others might say “AI-first,” we know customer service is fundamentally people-centric by nature. That means giving our AI the ability to understand context and hold a genuine conversation—no more feeling like you’re clicking through menus or talking to a script.
Freddy AI is precisely that–it’s about being able to talk like a human, able to understand what you’re saying, ask clarifying questions, and engage in a natural back-and-forth conversation. It gives you the same warm feeling when you are talking to someone who’s listening.
YS: Can you share specific data on the ROI your customers have achieved with Freddy AI?
PS: Freddy AI agents can be set up in minutes. They’ve enabled customer support and IT teams to resolve about 40–45% of their service requests on an average. In many cases, we’ve seen a 50–80% improvement in resolution times, depending on how they’re implemented. The bigger challenge is getting people started with AI as customers and businesses often have apprehensions about data security—like “What is happening to my data?” or “If your model doesn’t use this data, how is it learning?”
However, the systems may require little adjustment before using Freddy. They primarily learn on their own, and if you already have strong knowledge sources in place, the benefits are even greater. Even the ability to draw data from multiple sources has proven not to be a limitation.
YS: Several enterprises and startups such as Automation Anywhere and Salesforce have continued to develop similar offerings in Agentic AI. What sets Freddy AI apart?
PS: Current technology makes it easy for companies to claim AI expertise—any firm with an ‘.ai’ domain can promise results. However, the real test of an AI solution’s value depends on the quality of its data. For instance, our platform isn’t just a model plugged into one environment—we have information from a wide range of support tickets and operational metrics, which allows our AI to continuously learn, and refine recommendations over time.
While companies like Salesforce have access to more extensive datasets, our strength lies in simplicity. That’s our philosophy—whether it’s our products or AI, how do we make it simple, secure, and trusted…that’s why we’ve adopted this whole people-first approach. Freddy AI gives the same feeling as talking to a person, but keeps the person in control. You can dial it up or down and decide how much autonomy to give. It’s really about how simple you can make it and how powerful it is with the learnings, the models, and the data the AI agent works with.
YS: Could you share some use cases of how Freddy Agent is being utilised in the Indian market and the kind of traction it’s gaining? Additionally, are there other geographies you consider equally prominent?
PS: We’ve gained strong traction in the fintech sector, particularly with long-standing customers such as Paytm, PhonePe, and Razorpay. Beyond fintech, the ecommerce sector also aligns naturally with AI-driven solutions. Looking ahead, we’re seeing a global expansion of opportunities well beyond these core markets.
Traditionally, information-heavy sectors such as legal services and healthcare are beginning to adopt Freddy AI. We see huge potential in the broader business services areas, as each industry has its way of looking at AI and the problems it can solve. India remains an important market for us, where major brands are already our customers, although their level of AI adoption still varies.
YS: What challenges do you see with Freddy from your customers’ perspective? Specifically, what do customers typically identify as a problem?
PS: Data privacy is a key concern that comes up with every customer. They want to understand where their data is being stored, how it’s being used, and how they can ensure its security.
We’ve realised that presenting overly complex architectures doesn’t help. Instead, we need to provide simple explanations that customers can share with their companies—showing exactly how the model learns and operates.
One challenge is the fear of not knowing and being scared about starting with AI. There’s also the fear of it giving incorrect information to customers. We make sure that we focus on addressing these concerns.
The focus now is on addressing two key hurdles—first, helping customers get started with AI, and second, building their confidence that it is performing correctly. Once those are resolved, the next step is finding ways to continuously improve the system for better outcomes.
YS: How do you plan to strategise in catering to your customers moving forward?
PS: Our focus has always been on our customers and driving growth. However, we found out that our internal processes had grown overly complex, making it difficult to get things done as we tried to tackle too many goals simultaneously. Now, we’ve refocused on what we call the “vital few”—the key areas that truly matter—and are putting all our efforts into them.
AI, for instance, will take 100% of our focus. We’ve streamlined our operations to prioritise three core areas—EX business, CX business, and AI. Strategically, it’s about streamlining our operations and making sure that we are focusing and having the right talent for the things that we want to do.
YS: In terms of product strategy, considering your experience at Walmart, how have you seen the progression of AI over time?
PS: Before the rise of GenAI, my experience at Walmart involved developing intelligent systems that could perform tasks like “smart substitutions”—automatically selecting alternative products when a requested item was out of stock, among other capabilities. Now, I think it’s going to be huge when it comes to ecommerce, supply chain, and retail sectors.
At Freshworks, we’ve changed our mission statement to where we’ve become more of a people-first AI service software to help businesses deliver exceptional customer and employee experiences. So, for us, it’s about what can we do to help various kinds of businesses.
YS: What’s next for Freddy AI over the next two years?
PS: In the coming years, we won’t need to manually program each step for an AI agent to complete a task. Early on, it might handle simple, individual actions independently. But as we evolve toward truly ‘agentic’ AI platforms, these systems will manage far more complex processes. They’ll be able to break down scenarios into smaller tasks, take actions, and delegate them to various other specialised agents, ensuring the entire operation runs smoothly from start to finish. That’s the next major frontier of AI, where everyone in the field is heading.
Edited by Megha Reddy
