Dynamics 365 Contact Center Explained with Alpana Bajaj, at Microsoft

Alpana Bajaj, Partner Group Product Manager at Microsoft, discusses how Microsoft is reimagining the contact centre with AI-first design and agentic capabilities.

• How Dynamics 365 Contact Center evolved from Omnichannel to a standalone, CRM-agnostic platform born in the AI era

• The five out-of-the-box AI agents: Customer Assist, Quality Assurance, Case Management, Knowledge Management, and ServiceOps

• Real-world customer stories, including double-digit improvements in call deflection, wrap-up time, and CSAT scores

• How Microsoft uses its own contact centre as a testing ground, serving over a billion customers across 40+ languages

• Licensing simplified: per-seat for human agents, consumption-based Copilot Studio credits for AI capabilities

• A practical ROI framework covering customer experience, operational efficiency, and revenue generation

Link to the deck: Agentify Your Contact Center Alpana Bajaj

Link to the Commonwealth Bank case study

Thanks to Ribbon, this episode's sponsor, for their continued support of Empowering.Cloud

Transcript

Copy

Tom Arbuthnot: This week we are decoding the Dynamics 365 Contact Center. We're talking to Alpana who is Partner Group Product Manager in the Dynamics organization, and she takes us through what Dynamics 365 Contact Center is, how the commercials work, how the different agent scenarios work. She's really great at decoding how it all comes together. Thanks very much to Alpana for jumping on the show, and also many thanks to Ribbon, who are the sponsor of this podcast. Really appreciate all their support of the community. On with the show. Hi, everybody. Welcome back. really excited to have this podcast. Alpana and I were planning this quite a while ago, and then it was the, end of FY. We had the Teams Boot Camp, a whole lot going on. but really excited for this one to talk about Dynamics 365 Contact Center, I guess agentic Contact Center and, and where Microsoft are going with the product. so, Alpana, maybe you could introduce yourself and your role.

Alpana Bajaj: Absolutely. First of all, thank you so much for having me on your podcast, Tom. I've heard a lot about it, so super excited to be here with you today. for introduction, I actually lead the Growth and Product Charter for Dynamics 365 Contact Center and Conversation AI at Microsoft. my job sits at the intersection of product, customers, and business growth, and I often joke, that I'm a translator. I translate customer pain into product priorities and product innovation into customer value, and sometimes along the way, a lot of engineering acronyms into plain English. but at the end of the day, I'm super passionate about one thing, which is helping organizations transform millions of customer conversations into moments of trust by leading them through their business transformation journey. So that's my introduction.

Tom Arbuthnot: Awesome. Well, we're gonna lean on that decoding skill 'cause we know Microsoft is a world of, technology and acronyms, and, there's, there's lots to decode.

Tom Arbuthnot: Can you take us through a little bit of your background with D365 CC and your background at Microsoft? 'Cause the product's been around for a while now under different names and different guises.

Alpana Bajaj: Absolutely. So My journey with Microsoft actually started in 2014, wherein I joined the Microsoft CSS division for their IT tooling to transform their customer service engagements, their agentic experience for both human agents and as well as for their supervisors. So briefly, for two years into that role, I was pulled into the core product. I got this wonderful opportunity to drive larger impact with broader Microsoft customer base and to bring all that innovation right into the Dynamics 365 product portfolio. So roughly about 2016 is when I joined the Dynamics product family, product engineering family, and back then I was actually part of the customer service product. So that's where we were working on case management, transforming knowledge management, so on and so forth. So very quickly moving forward In 2019 is when we launched omnichannel with the digital live chat channel first, and then rapidly we added social channels such as Facebook, WhatsApp, so on and so forth. Quickly, by, I believe, end of 2021, beginning of 2022 is when we added voice channel into the mix, because voice is still a very strategic and important channel for all the Contact Centers. So gradually over a period of two to three years, we added a lot of channels to make it truly omnichannel experience. Then in 2023, 2024 is when we heard the feedback from a lot of customers that they loved what we were doing as part of the omnichannel product portfolio, and they wanted that product portfolio to work with any CRM of the world, not just the Dynamics stack. So that's when in 2024, what we did was we took the omnichannel product portfolio that we had, and we made it work with any CRM in the world. When I say any CRM, it includes both the third-party vendors that are available today in the market, or it could be a homegrown CRM that any enterprise may would have. Right? So that's when in 2024 we repackaged everything and we launched the Dynamics 365 Contact Center applications.

Tom Arbuthnot: that's really the point where it was like, okay, this is a serious independent play at Contact Center rather than a part of the Dynamics customer service kind of package.

Alpana Bajaj: Absolutely. So that was a very defining moment for us because it truly became a standalone Contact Center application driving customer engagements across a lot of CRMs. And we truly opened the platform. That is one of the USPs that got, enabled in the platform, wherein the openness of the Contact Center product that we have enabled lot of customers to ensure that they don't have to rip and replace their entire Contact Center platform in one shot, but it gave an opportunity to Microsoft to meet customers where they were in their transformation journey. So in the same spirit, then we added voice agents and lot of the AI era came about, and one of the best things that happened with Microsoft was that while we were designing this Dynamics 365 Contact Center We were designing it in the era of AI. So this entire product was born in the AI era, which gave us an opportunity to reimagine the entire product in a way that AI is infused at every step of the customer journey, and it doesn't appear as if it is a bolt-on. So when we did the split and launched Dynamics 365 Contact Center, we ensured that it not only works for any CRM in the world, but it is born in the AI era, so AI is infused at every step of the customer journey. So we made the best use of the opportunity that we had at Microsoft.

Tom Arbuthnot: Yeah, it, it's interesting, isn't it? Because you've got within Microsoft, you've got all these awesome building blocks. You've obviously got the Dynamics team and that set of knowledge. You've got Teams with hundreds of millions of users, and then therefore the IC3 team doing all the backend media stuff. and then you've obviously got huge investment in AI. so you've got fantastic kind of different pieces to kind of bring together, and I guess that's your opportunity and challenge is bringing all that technology together in an actual kind of customer facing product and service.

Alpana Bajaj: Absolutely, Tom. In fact, a lot of times customers ask me, "What is Microsoft's differentiation?" In fact, what you said just now in the last 30 seconds or so is the true differentiation for Microsoft. Because we have all these technology stacks that have been in the market for very many years, matured over time for enterprise grade trust, scale, such as Teams. We know Teams is taking care of billions of conversations every year. We have AI, where Microsoft has been investing a lot for many, many decades, right? And the other part is whenever Microsoft builds product, the trust, security, compliance, scalability, all these elements are baked into the product. So when that happens, then the AI especially, number one, is context aware because at the end of the day, models are going to become commodity, right? Everybody will have access to the same models. So what is Microsoft's true differentiation when it comes to AI? The AI that we have is context aware because we have all these business applications, customer service- sales, marketing. We have all these conversational, stacks such as Teams, right? So the AI that is trained on the customer's context, be it the structured data or be it unstructured data such as conversations, plus AI that is born in the, like Contact Center that is born in the AI era, plus AI that is available in the workflow in various different business applications, that becomes the true differentiation for Microsoft when it comes to AI.

Tom Arbuthnot: That's awesome. I wanna get into the, Teams and Teams phone extensibility and how it works alongside Teams, but maybe we could stick with that agentic piece first. Can you take us through how you're thinking about, agents in the Contact Center, as in AI agents rather than human agents?

Alpana Bajaj: Yeah. See, when we speak about agents, I would take us back to the day in the life of a Contact Center, because we should not be thinking about AI or agents just for the sake of doing something in AI, right? That AI or agents should meet a business problem and solve a business problem. So going back to the day in the life of a human agent, or the day in the life of a customer, or the day in the life of a supervisor inside Contact Center We go back to the basic things that they care about. Like for example, what do customers care about? Customers don't wake up hoping that they would have to contact a Contact Center, or if they contact a Contact Center, they're getting deflected by a conversational IVR. That's

Tom Arbuthnot: not what customers- Yeah, yeah. Like, like as a customer experience, you're already on the back foot, aren't you? 'Cause ideally they wouldn't be, so now you've got to get from, like negative to neutral to potentially positive.

Alpana Bajaj: Spot on. Spot on. So that's one of the biggest opportunities that we tapped into with our customer assist agent, which is proactive, meaning even before customers reach out to us stating their problem, the proactiveness of the customer assist agent tries to figure out where is the customer on their journey with the organization. And if there are any activities that are going on with the customer, proactively inform the customer, update the customer, and do things for the customer. So entire proactiveness of the customer assist agent is one of the true differentiators. And in situations when customers do reach out to the organization, how do we completely transform that engagement so that it is not the traditional DTMF IVR, but a new conversational agentic experience that we can give to the customer, which is consistent across all channels, be it if they reach the organization on voice or if they reach the organization on their digital channels. So proactiveness, truly omni-channel engagement. And one of the. this is a story that one of my customers told me, so I would love to tell that story out here when it comes- Yeah, yeah.. to the customer assist part. So when we looked at the data for this customer, they had spent roughly 10 to 12 months building a conversational IVR. And when they used, when they launched that conversational IVR in production, very quickly they realized that a significant percentage of the calls were getting escalated to human agents within the first 30 seconds. I'm sure you would have realized it and you would have encountered it too when you would have called a customer service call, phone number, and you would have, you know, screamed, "Connect me with a representative or an agent," in the, within the

Tom Arbuthnot: first 30 seconds. Yeah. Yeah, yeah. Just, just, just, just hammering zero until you get through to somebody.

Alpana Bajaj: Yeah. So when I was talking to this, CTO He came back with a statement that stuck with me, which is, "Alpana, we have spent millions building this IVR, and our customers are escaping this IVR within the first thirty seconds." That's not the place where any organization would want to be. So how can we transform their journey with the new AI and agentic tools that we are building? So a very short story, but what we did, long story short, fast-forward a few months ahead, was that we understood their business use cases, and within a span of three to four weeks, we built a conversational agentic voice agent for them. And with that conversational agentic voice agent, they saw double-digit improvement in the order status inquiry that they were facing in their previous stack, where within thirty seconds the customers were escalating to their human agents. That is the power of customer assist agent. First and foremost, you can build it very quickly. Second, it drives true business impact and for the KPIs that matter to business. So that's the customer assist agent. Then the next persona that we have in the Contact Center ecosystem is human agent. For decades and decades, Contact Centers and organizations have been working on optimizing for human agent productivity. So that's where, like in the past, we have been talking about agent scripts, macros, copy-paste tools here and there. But all that was very tactical investments that we made in the past. With AI, we had true once in a decade opportunity to transform the human agent experience and focus not just on the customer satisfaction aspect of it, but also your employee satisfaction aspect of it. So that's where our quality agent, provides coaching abilities to your human agent, provides the right nudges at the right time with the right help and context. So this is where we also have Copilot, which will give things like conversation summary, sentiment, completely summarizing the, entire transcript into few things that matter the most to the human agent, suggesting them with the right answers at the right time. And not just that, there are capabilities in there which will allow to completely offload some of the tasks that the human agents do to an AI agent. For example, this is again a customer story for you, Tom. when I was working with a customer, they told me that their human agents spend roughly somewhere between two to five minutes After every call, manually typing in and creating a case for after the call follow-up.

Tom Arbuthnot: Yeah, yeah. The kind of classic wrap-up time, and optimizing that has been a big conversation for years, hasn't it? Because how long do you need to give to do a good job versus that person is not taking calls and you've got calls coming in.

Alpana Bajaj: Exactly. So what we did was that we introduced our case management agent to this customer, and we did a quick, proof of concept pilot with a small line of business by routing it, like a 10% of their call volume to a subset of agents who had case management agent available to them. And with that, what they.. The experience that they got was that at the end of this call, the entire case form was pre-populated with all the fields that mattered to the business. The only thing that human agent need to do was review the pre-populated case and wherever edits were needed, make those small edits and hit the submit button. That brought down the after call wrap-up time from four and a half, five minutes to just under a minute. That was a huge- gain.

Tom Arbuthnot: and that feels like an area where I've seen lots of organizations happy to start with AI 'cause they're still like, "Am I ready to have AI be the front end customer interaction? Maybe, maybe not, but can I.. Can AI save my human agent's time? Can it make the summaries better, more accurate?" Because it's not just about time, it's about consistency and accuracy because you can train.. I assume in that scenario it was doing something along the lines of transcription summary, picking out key items, putting them in the right fields in the, the backend platform. Like the.. That's a hard job to do consistently when you're doing hundreds of calls a day, so you get not just speed improvements, but kind of reliability improvements as well.

Alpana Bajaj: Spot on. So the quality of cases that got created also improved. So as you said, it's not just speed, but the quality of the cases. The, the details in the case description, that improved drastically, so when the case got handed over to a back office human agent, the impact of the back office human agent also drastically improved. So that's the end-to-end play we are talking about. And, in the spirit of giving you more examples, I'll also talk about the quality assurance agent that we have. Many.. And I'm sure you would be able to relate to this. Many enterprise organizations have somewhere between 1 is to 20 to 1 is to 80. Kid you not, I've also heard 1 is to 80 number. Wherein this is the ratio of human agents to supervisors who are doing the quality audit on the different calls or- Yeah.

Tom Arbuthnot: What, what shot have you got 1 to 80? That's crazy.

Alpana Bajaj: Yeah. I mean, this was a big enterprise customer. Yeah. And for them, compliance is super important, so having the right levers for audit was super critical for them. And despite doing their best, this was, the best ratio that they could come up with. And the other problem is that with this ratio, all you can do is sampling.

Alpana Bajaj: It's really difficult to provide 100% coverage on quality audits, right?

Alpana Bajaj: And the second big problem that they were facing was that most of these quality audits were happening after the conversation, which means that this poor supervisor who is managing 20 agents, human agents, or 80 human representatives, has plenty of calls and conversations to go through. Imagine his or her life going through all those conversations, reading those conversations, auditing those conversations. It's a- Herculean task. Now, with quality assurance agent, you not only can do 100% coverage, but you can also do it real time, and that's a game changer because now when it, when the quality assurance agent is acting real time- It truly is an opportunity for the human reps to drive better customer satisfaction. You don't need long cycles of weeks and months for a supervisor to annotate the improvement opportunities, go back to business process improvements, then train your human agents, and then see the impact of those audits months or years down the lane.

Alpana Bajaj: All that cycle has been compressed into real-time coaching abilities for the human representative. That's a win-

Tom Arbuthnot: So the quality assurance agent is both, obviously kind of rating and understanding the call and the experience, but actually coaching the human agent in real time saying, "Hey, remember to talk about the extended warranty," or, "The customer said this, this is the relevant KB," that kind of thing?

Alpana Bajaj: Absolutely. Spot on. And talking about KB, we also have a knowledge management agent that silently listens to the entire conversation, and for the conversations, you know, Tom, you always have some human reps in your Contact Center who are like superstar human reps, who are able to get higher CSAT scores, who are able to resolve customers' issues in a very, effective and efficient way.

Alpana Bajaj: Yeah. So with this knowledge management agent, what happens is that this knowledge management agent silently listens to the entire conversation and sees how this human rep was able to drive those high-quality conversations. Based on their behavior, based on their patterns, what this does is that it automatically creates a knowledge article, which then not only helps- With other human representatives. But this knowledge article is also fed into our customer assist agent for self-service autonomous improvement as well. So it helps us close the loop and build a, a truly autonomous agentic capability.

Tom Arbuthnot: That's awesome. So we've, we've talked about customer assist agent, we've talked about quality assurance agent. Like how, how much is there.. I don't know what the number is, but how many agents are there kind of out of the box? And, and is it.. Do you then customize each of these scenarios? How does that work?

Alpana Bajaj: So at this moment, we roughly have five agents out of the box. customer assist agent, quality assurance agent, the case management agent, knowledge management agent, and the service ops agent. We have not spoken about the service ops agent, but we'll come to that very shortly. coming back to your question, all these agents are available out of the box. They are both configurable and customizable. I'll give you a few examples. Like for example, the customer assist agent, if you want this customer assist agent to work on your existing knowledge base, you can completely customize it, configure it. There's a very point-and.. easy to use point-and-click experience using which- you can give very natural language instructions. And with the help of natural language instructions, you can build a customer assist agent. Once you build, we also have out-of-the-box eval capabilities, evaluation capabilities that are available inside Microsoft Copilot Studio, which will help you assess the quality of the agents that you're building. Once you're satisfied based on the evals that you have, you can test it for from a human, perspective as well through the test panels that we have right within our product. And once you're satisfied, with one click you can publish it. So that's for the customer assist agent. For quality assist assurance agent, I would like to give you a different example. Now, lot of times various organizations have different parameters of quality audits and compliance aspects. For example, banking sector would have different questions, right? Tech sector would have different compliance aspects. So what you can do is you can literally create your own quality execution framework and plan. By that what we mean is that you can define the questions, you can define the weight for every question, you can define the plans, whether you want to execute for 100% of your conversations, whether you want to execute for a subset of the conversations. So all that is completely customizable and configurable, and you can tune it for your organization's needs.

Tom Arbuthnot: That's really good to understand. And like, so, like, we don't wanna get into the weeds of the commercials, but you've got those five agent scenarios. H- how does this work? Am I paying per human agent? Am I paying per virtual agent? Per session? Consumption? What does that look like?

Alpana Bajaj: I'm glad that you asked that question, Tom, because licensing and pricing has always been a very difficult topic for-

Tom Arbuthnot: We, we, we talked about you decoding complicated things.

Alpana Bajaj: Yeah. So here is my attempt to decoding that. Good. So when it comes to Contact Center, we have tried simplifying things. For human agents, it's today what we have is a per seat licensing model. When it comes to AI capabilities and AI agents, we have simplified the entire stuff. Everything is.. Because all our agents are built on top of Microsoft Copilot Studio as a stack, so it is consumption-based pricing, and the unit of that consumption is the credits that we have inside Copilot Studio. So it's human seat for human reps, and for AI capabilities, it's Microsoft Copilot Studio credits, which is consumption-based

Tom Arbuthnot: pricing That's great. Yeah, that's great to understand. Again, this is another part of this building blocks conversation of-

Tom Arbuthnot: Copilot Studio is obviously heritage is all in the Dynamics org, and like, I feel like that's been shot to, like, the most important thing in all of Microsoft at the moment 'cause it's doing so much with the AI agents. So-

Tom Arbuthnot: so per seat for the humans, and then consumption for the agents, depending on what you're doing, how much you're doing.

Alpana Bajaj: Yeah. So that's my simplification of the- That's great.. complexity.

Tom Arbuthnot: Great. And, and you talked about customizing these agents. How much of that is Microsoft partners? How much are people doing that in-house? like, what does that look like? 'Cause I guess it's not a one and done, it's a, there's probably continuous kind of work to optimize there.

Alpana Bajaj: That's true, and I believe this is the right point wherein I can talk about, remember the one agent that we did not speak about so far? Which is the-

Tom Arbuthnot: Oh, yeah, yeah. We've gotta do a service agent, right?

Alpana Bajaj: Service ops agent.

Tom Arbuthnot: Yes. You're gonna tell me the AI does it all for us now. Oh.

Alpana Bajaj: Dude, you are helping me answer all the questions. So this is what we realized. Humans are best at their natural language. So what we gave with service operations agent is how can humans express what they need in a very natural language, conversational way. And we take those instructions, decode those instructions programmatically using the service operations agent, and behind the scenes do all those configurations for you And reduce the complexity and the time it takes for you to do all those configurations. Like, Tom, I'll ask you a question. Typically, how many clicks or how many days do you think it takes for anybody to set up a Contact Center or create an agent?

Tom Arbuthnot: Well, I, I mean, historically, like I'm going back now, I used to do Cisco Contact Centers, and we would do, four to six weeks of like, work flowing and programming, and that was like, that was like a minimum for like setting up Contact Centers.

Alpana Bajaj: Now you can do

Tom Arbuthnot: it in days.. these days.. Yeah, yeah. These days it's got a lot faster, hasn't it?

Alpana Bajaj: Yeah. So with service operations agents, you can reduce those hundreds of clicks into couple of instructions.

Alpana Bajaj: And because you mentioned about SI partners, I'll give you another example, where this SI partner who tried our Service Ops agent, they came up with a very brilliant idea. So what they did was they kept having back-and-forth conversation with the Service Ops agent to build both their Contact Center and as well as some of the agents that we ship out of the box, right? And after that, they compiled the set of instructions into one big instruction per industry, and geo that they were targeting. And after that, any time a new customer came in in the same industry, all they had to do is copy-paste this one big instruction that they compiled for that industry segment for the new customer, and that's all. In very few minutes, they were able to set up the entire Contact Center.

Tom Arbuthnot: Oh, interesting. So they get to be like more of a specialist in a particular sub-vertical, and they're like, "We really understand, like, medical-"

Alpana Bajaj: Domain

Tom Arbuthnot: expertise ".. Contact Center," or, "We really understand, you know, insurance Contact Center." That's really interesting.

Alpana Bajaj: Spot on. So now they are.. We have simplified the technology aspects for them, so now they are totally focused on the domain expertise, the business process expertise, and that is the value that they are bringing in the engagement with different customers.

Tom Arbuthnot: Nice. And how, how are you seeing customers get to grips with the idea of kind of consumption and, and variable costs? 'Cause it's interesting, like the AI is, the more it's doing, the more it's costing, but the more it's potentially saving. Like, how do you map those with somebody who's used to a fixed cost per head?

Alpana Bajaj: that's a very interesting and a challenging question that Not just I'm facing, but I see a lot of CXOs facing. Like-

Tom Arbuthnot: Yeah,

Alpana Bajaj: yeah, yeah. O- How do you control costs and how do you prove the ROI of whatever cost you're, you're spending so that you can prepare a business case and get more investments on AI? Because at the end of the day, not making investments on AI is the biggest risk that you're taking. But at the same time, everybody has to be pragmatic for the real-world laws of physics and maths- Yeah,

Alpana Bajaj: in terms of they get continued investments on AI as well. So here's what I have, seen, Tom. First and foremost is the kind of framework that you can follow to prove, like, to find scenarios that drive value for your business use case. So don't do AI just for the sake of doing AI.

Alpana Bajaj: There are a lot of business cases where AI can really truly drive value for your business use case. So identification of the value-driving business scenarios, so that's one. Second part of the framework is defining the success metrics for that scenario. So what matters to you the most? So that's the second part of the framework.

Alpana Bajaj: The, the third part of framework is the business constraints that you need to operate in. So even though you have identified the scenario, even though you have identified the success metric for the scenario, but the business constraints may not allow you to do that experimentation in a full-blown way. So start with a very small use case, and then when you run, that experiment, go measure your metrics back again and do the ROI calculation. Like, for the amount that you spent, for the success metrics that you had defined with this controlled sample of pilot that you did, what is the ROI that you have? And I'm sure, like,

Tom Arbuthnot: and- Yeah, that, that's really useful because that, like, I think for partners listening in as well, like, 'cause it's, it's, in some ways, Contact Center I think is a really interesting opportunity 'cause it's, it's very measurable, it's very customer and business impacting. Right. So you're in an area already where you can add quantifiable value.

Tom Arbuthnot: But you talked about, like, agent coaching or monitoring. Like, if you're not doing 100% of that today, and you want to do 100% coverage with AI, that's net new spend. Now, there's a lot of value in that spend, but it's net new spend. Obviously, if you're using AI to deflect or handle customer calls, then there's a different thing there where that's some of the voice models are quite expensive, but compared to the equivalent of load of a human agent and, and all the flexibility and 24/7 and scalable and everything else. But I, I think I, I don't see enough conversation about that to help the business understand, like, there's a new model here.

Alpana Bajaj: Yeah. Very, very well said, Tom. So here are the three things that I have learnt based on the decades that I have spent in the Contact Center industry, and they are getting amplified in the AI era when it comes to, measuring ROI. And focusing on these three things will help you ensure that you're not just focused on efficiency, but overall effectiveness of the, organization. First and foremost is your customer experience. The second is your operational efficiency. And the third is whether you're converting your Contact Centers from contact from cost centers to revenue-generating engine. So when you focus on these three key pillars, not just one, so you have to focus on these three key pillars, then you will realize that AI will truly be the differentiating factor for your Contact Center. Let, let me take an example that you mentioned and, explain a little bit more deeper as to how, I have seen customers taking the best use of it, right? So when it comes to just efficiency, one of my customer was just measuring the average handle time aspect when they started with this, proof of concept and pilot, and they did not measure the CSAT part, and they did not measure whether the Contact Center is driving, more potential opportunities for revenue.

Tom Arbuthnot: So if I, if I hang up on one in three customers, then the average time goes right down.

Alpana Bajaj: Exactly. Exactly. And that's where AI completely transformed. The second place I'll give an example is this customer was running only 9:00 to 5:00 Contact Center because of the human, capital expenses, right? But now with AI, they moved to 24 by 7 Great option for them. So-

Alpana Bajaj: one, it helped them scale to twenty-four by seven, and yes, to some extent it was a slight increase in the cost, when it comes to AI. But when they measured it against the cost that they would have spent with human agents to achieve the same capabilities, it was just a fraction of the cost. And the CSAT, once they went from nine to five to twenty-four by seven, the CSAT percentage point improvement was more than double digit. That was a huge win. Yeah. Not just that, the upsell that they did with the AI agent, that brought in additional revenue. So initially, when they were just.. If you take a very myopic view and just look at, "Hey, what is the delta cost that I'm incurring on AI?" Maybe you may not be getting a very good answer. But if you take a broad lens and focus on all three pillars, whether the spend I'm doing on AI, how much relative is that to the spend that I'm doing on human reps? Or how much benefit I'm getting for that spend on the customer sat spot, and how much benefit am I getting on the revenue generation part? So when you take a look at all the three pillars, that's what gives you a true, meaningful ROI. If you just apply a very, siloed lens on one specific KPI, it'll not give you the right measure that you should be focused on.

Tom Arbuthnot: Yeah. That's really useful to understand, 'cause I think that is a problem, and I think quite often in Contact Center, certainly I've seen we measure what we could measure. So we can measure ho- hold time, we can measure, like you say, call wrap-up time, we can measure call time. We do sampling for CSAT, so the people that, you know, say bother to reply, we get some kind of sample of CSAT. We've got the potential to have much more visibility in AI as to what's really going on. But you do need that business buy-in to be like, "We're spending money for these outcomes." Like, just like if we took the Contact Center to twenty-four hours, we'd have to have multiple teams or round the sun or however we do it. So it's, it's making sure we understand we're comparing the, the cost with the value, not just it's a net new cost.

Alpana Bajaj: Absolutely.

Tom Arbuthnot: Awesome. Well, I, I, I wanna give you a chance, and you have got an awesome slide on customer stories. I don't know if you wanted to bring that up or anything else up to, talk through. But I'd love to hear a little bit about the Microsoft internal journey with, the 365 CC and a- anybody else you wanted to talk about.

Alpana Bajaj: Absolutely. So let me bring up the slide. I always mention this thing that Microsoft, I believe, is one of the most, complex and largest Contact Center providers in, in the world, I would say.

Alpana Bajaj: We, we have more than a billion customers, tens and thousands of service reps spread throughout the globe And taking care of customers in more than 40 languages. Language has always been a key focus area for multinational organizations, right? And we do millions and millions of interactions every year spread across different channels. So when we started our journey roughly around 10 to 12 years back, we had more than 17 to 20 different applications, and human reps had to.. There's a very popular term in Contact Center industry, swivel chair. Sometimes they're on this app, sometimes they're on that, Like-

Alpana Bajaj: lot of app switching, script following, so on and so forth. So we started this transformation, journey of providing a single pane of glass experience to our human reps. Then we added a lot of Copilot capabilities, and then we also transformed it with lot of agentic conversational bots. So with that, the good news is that we made the mistakes first Because whatever we produce in the Dynamics portfolio or the conversational AI portfolio, we dogfood it or we drink our own champagne. We try to leverage all the capabilities within our Contact Center first, and only when, those capabilities get mature to the extent of driving double-digit KPI improvements, that's when we expand it, right? So super happy to share some of the numbers that you see on the screen, wherein we have seen double-digit improvements based on our own transformation journey. Wherein we apply our product, we test our product, we try to break it with lot of complex scenarios. Yeah. And with that journey, whatever we learn, we try to bring it back into the product, and with that, we have seen double-digit improvements in the KPIs that matter the most, right? Not just us, but you will also see lot of other customers boasting on the internet, like if you go to LinkedIn or if you follow me on LinkedIn, you will see a lot of these customer success stories because this is something that I'm truly passionate about. Commonwealth Bank of Australia, one of the largest banks in Australia, again, serving tens and millions of customers. They have been on this journey of transformation with us, and, in the recent tour of, AI with Microsoft, they presented their story to Satya, and they shared their experience on LinkedIn as well. Again, like business transformation at scale. Not just this, we also have lot of other customer success stories, again, available on our public website, customers.microsoft.com, across different industries and geos seeing similar double-digit improvements on the KPIs that they benchmarked, us against some of the previous tools and apps that they had And this, not only testifies the level of impact, but it also signifies that the leaders who are taking the bet on the AI journey and going through that AI journey are seeing drastic improvements for the KPIs that matter to their business. So anyone who is, again, one of the customers shared this with me, that they feel they have FOMO, the fear of missing out-

Alpana Bajaj: they're not on the AI journey. So customers are no longer asking the question of, hey, whether the AI is mature enough that they, should, join, the bandwagon, but now the question has drastically changed. The, the question now that I get is, "Alpana, how fast can I get onto this AI journey and help me through the framework or the transformation journey that I need to focus on for my business use case?" So the conversation has drastically- Yeah, I

Tom Arbuthnot: think it's, I think it's the things you said earlier. It's like, how do I measure the ROI? How do I pilot in a scalable way? Like, what about the, the day two management? Those are the.. W- I think we're beyond the just like it's a flashy demo into like I, I get AI can add value, but how do I make this actually work for my organization as opposed to see a demo?

Alpana Bajaj: Absolutely.

Tom Arbuthnot: Awesome. So, if, if it's all right with you, we'll share some of these slides, and we'll, we'll link them below as well. And yeah, we'll definitely share your LinkedIn, 'cause I've seen you post a lot of the good stories there. thanks so much for decoding all of that. I know there's more we want to talk about that we can't talk about just yet, so, we'll have to have you back when the time's right, 'cause there's a lot of cool stuff coming as well. But that was a, a great decode, Alpana. Thanks so much.

Alpana Bajaj: Thank you so much, Tom, and, thank you so much again for this opportunity. Had an amazing conversation with you, and looking forward to coming again sometime, in the next few months.

Tom Arbuthnot: Awesome. Thanks so much.