Melissa Fersht — Executive Research Leader, HFS Research[00:21]
Hello and welcome to our new customer experience series. It’s 2026, the year of the how of AI, and I’m here to explore the impacts of AI on customer engagement and how that impacts our OneOffice. So today I’m joined by my friend and fellow Bostonian and the Global Head of Customer Experience at Genpact, Sachin Pai. So welcome, Sachin. Thanks for coming.
Sachin Pai — Global Head of CX, Genpact[00:46]
Thanks very much, Melissa. Very excited to be here.
Melissa Fersht — Executive Research Leader, HFS Research[00:49]
We are really excited to dive into the impacts of AI on customer experience. So let’s get right into it. I’d love to hear from your experience, Sachin, day-to-day working with customers, how you’re seeing agentic AI reshape the day-to-day experience for customer experience and contact centers, as well as where you see the biggest opportunities for customer engagement today and the impact of AI.
Sachin Pai — Global Head of CX, Genpact[01:15]
Great question. I think we’re at a watershed moment, actually, or we will be in one very soon. So I think when you ask about what we’re seeing with clients, with customers, I think there are three kind of pivots that we’re doing in the market, that we’re seeing in the market. One is, and I’ll get to the other two shortly, the most important one is I think a drive from self-serve to serve by AI. And let me elaborate a little bit more about what I mean here. You’ve heard a lot about self-service. We’ve all been talking about deflecting contacts and driving customers to self-service and customer service interactions. And we think that rubric is changing and being flipped on its head. We think self-service is dead. And we think the future is really for AI to serve the customer or do things on behalf of the customer. And this is very interesting, Melissa, because what it does and what it means is it means that you’re not trying to deflect contacts anymore. You’re not trying to reduce the number of contacts. You’re trying to get the AI to support the customer in doing what they’re trying to accomplish. So let me pick a real example, if I may. Today, it takes between 8 to 11 clicks to just do something as simple as transfer money from one account to another or create a one-time credit card number on your, let’s say, banking app. This applies across industries, of course. And the only way of doing this is driving people to self-service. That someone calls into the customer service organization, and we deflect the contact, and it goes into self-service. And guess what you’ve just done? You’ve just offloaded that entire 11-click, 8-click process to the customer. It increases friction. It actually makes the customer, you’re telling the customer, I don’t want to do the work. I’m going to punt that work over to you. And sure, it was beneficial up until a couple of years ago because you would reduce that cost to serve the customer and you’d put it back on the customer. But guess what? Now we’re in an era where the AI implemented appropriately with the agentic systems end to end can actually execute those clicks and execute those transactions on behalf of the customer for the customer. So if there are two clients, one that produces that friction through self-service and one where the AI does it for the customer, as a client or a customer, I’d want to pick the one that reduces the friction and does it on behalf of the customer. So let me pause there and see what your thoughts are. And we’re calling this moving from self-service to serve by AI.
Melissa Fersht — Executive Research Leader, HFS Research[03:48]
Yeah, I like that. And self-service is dead. You heard it here first. This is really interesting. But actually, it begs the question to me, where does the human element come in here now? And what role does empathy play? Because this has always been a big thing for me, especially in working in contact centers in my earlier career, which was there’s always going to be an escalation or there’d always be a moment where, you know, you can’t just click through. You need to talk to somebody. It’s an exception or say it’s a sensitive situation. It’s insurance. Somebody’s just been diagnosed with something or a loved one has passed. So, you know, is AI going to be able to replace those moments or will it complement them? Like, what’s the role here?
Sachin Pai — Global Head of CX, Genpact[04:29]
So really good question again. And I think the received wisdom is a little bit off track here. And let me try and explain what I mean by that. What you’re going to hear across in terms of trends as of last year, or maybe even earlier this year, the first two weeks, is that you need a human in the loop during some instances, especially during emotionally charged interactions. And I frankly think that era is also over. Let me pose with this, pose a question and then go on to answer it. If anyone in the audience is listening to this, think for a moment about the number one use case that consumer-facing apps, whether it’s ChatGPT or Mistral or Anthropic or your favorite, or Gemini, your favorite consumer-facing generative pre-trained transformer app, the number one use case, take five seconds, think about what that is. And okay, here’s the answer. It’s actually people asking for emotional support. It’s people asking for advice and relationship management. There’s a lot of research done in terms of how Gen Z and Gen Alpha are using it, teens especially. And while you might have a view on that right or wrong, the point is that generative pre-trained transformers are at a point where the emotional connection they can draw just using large data sets to predict what the most emotionally charged response is that’s going to either engage with the person on the other side is exceeding what people can do, frankly. Again, right or wrong, the point is that it causes those dopamine levels to rise and engage emotionally with customers. So we think actually that emotional engagement is a very good use case for actually getting AI to be injected into the interaction and do better than what a person could traditionally do. Now, of course, there are going to be humans in the loop, on the loop, but very, very strategically placed to actually improve the training behind the AI versus delivering that interaction itself.
Melissa Fersht — Executive Research Leader, HFS Research[06:48]
Yeah, I think this is such an interesting subject. And, you know, there’s definitely some discussions about guardrails around how people are using personal lives, I’m sure that there will be the same in customer engagement. And we can go down a whole other, we should do another videocast just on this topic. But I want to keep moving. I want to get your thoughts on how do we know that this stuff is successful once you implement it? So what kind of metrics do you put in place? And what kind of outcomes do you see to know, okay, the customer’s getting served better, we’re saving money, we’re making more money? How are you able to actually quantify the impact of this?
Sachin Pai — Global Head of CX, Genpact[07:29]
All right. So great question. So probably the most important question, actually, about business impact and about what the actual outcomes are. But one point about guardrails that you mentioned right before is also very interesting. We don’t actually think guardrails are hindrances to building AI. We think that the guardrails and the regulatory framework, et cetera, actually fed into the AI during inference stage, versus during training, actually help improve or helps improve the output that the AI produces. So we think of guardrails as a benefit and regulation as a huge benefit in those regulated industries where you feed in terabytes of data about those regulatory frameworks that actually gets the AI to produce an outcome that’s frankly more compliant than what people can do. And that dovetails right into the example, or actually the question you asked about, with what are the metrics? What metrics are now being brought to the fore to actually judge and determine the efficacy of these AI systems? And here’s where I’d say the third most important kind of topic really comes to the fore, which is it’s not about customer service, and it’s certainly not about contact center. It’s more about end-to-end customer experience. And here’s what we mean by this. It’s really about driving an end-to-end use case from start to finish that’s orchestrated by agentic systems and agentic AI. Now, this is not generative AI. This is agentic systems, which means that your AI needs to actually not just generate the next most probable word or next most probable thing to say in a conversation, but it needs to generate the next most probable action to take, and then the action needs to be taken to be able to do that. So very simple example, two examples, actually. And let’s say the financial services industry, let’s pick insurance. We can actually impact the claim loss ratio of clients. So you’re an insurance company, you charge premiums to clients and customers, and then you’ve got to pay out claims. And the agentic system is done right if you’re adjudicating claims correctly when you get your first notice of loss and then do the claims adjudication and then do the subrogation and then go down the value chain of actually paying out the claim, we’re going to be able to impact metrics like the claim loss ratio at an insurance company. Equivalent example in healthcare or healthcare insurance, you’re going to be able to impact the medical loss ratio that the payer pays out for charges that you as a patient have incurred from a provider where you’ve gotten services. And this kind of impact is fundamentally driven by agentic systems, which is why we’re calling it end-to-end, where the metric is not a handle-time metric anymore. It’s not even a customer satisfaction metric. It’s a metric that directly impacts the P&L statement of our clients. So we think this is the single most powerful reason to actually really double down on these agentic systems because what you’re driving is an outcome. And it’s also why we call it end-to-end customer experience, not customer service, certainly not contact center.
Melissa Fersht — Executive Research Leader, HFS Research[10:48]
Okay, this has been great, Sachin. Thank you so much. So I have a lightning round question for you, a pretty difficult one, so I’m going to put you on the spot. But what does customer engagement look like in 2030? And what role do we play in it? What roles do humans play in that?
Sachin Pai — Global Head of CX, Genpact[11:07]
I would say the next watershed moment is going to be one in which, you know, that proverbial zero and actually the actual zero, when you try to engage with customer service today for any company, I’ll tell you from personal experience and business experience, we see people clicking on or hitting the zero button to reach a human operator. And we think that’s going to be flipped on its head. I think sooner than 2030, we’re going to find situations where agentic systems are so sophisticated and the conversational ability and action ability of these systems to actually execute what a customer is asking for and requesting for and trying to do and trying to accomplish is going to get so frictionless that people will start hitting zero or the proverbial zero to reach an agentic AI saying, I really don’t want to talk to the human operator. I actually want the agentic system to take care of this for me. And I think we’re going to reach that watershed moment a lot sooner than 2030. There’s a race to superintelligence within certain domains. I think the domain that achieves superintelligence with foundational AI models is probably going to get here sooner than we think it is. And then our role will then pivot into actually building these agentic systems that are capable of doing things as sophisticated as building APIs on the fly. And getting those use cases that customers are asking for to be executed real time on the fly using GPT tech together with agentic technology. So we think that the future role is not going to be one where you get an escalation to human when the agentic AI doesn’t really kind of work the way you want it to. But we think that the role is going to be of us as agentic operators to actually fix those agentic systems, improve the algorithms, feed it with the right data, not just during training stage, but during the inference stage. And during the stage of actually collecting the data that’s going to make those agentic systems perform better and actually execute what the customer is asking for.
Melissa Fersht — Executive Research Leader, HFS Research[13:20]
Fantastic. This is going to be fascinating and interesting, that watershed moment coming even quicker than 2030. So things are happening quickly. Great. Thanks so much for your insights.
Sachin Pai — Global Head of CX, Genpact[13:32]
Melissa, thank you. Look forward to what the future holds for us.
