Unlocking the Future of Customer Service with AI

Samuel
In this new episode of Mastering AI with the Experts I'm thrilled to welcome Marcus Schmidtt a principal program manager on the Microsoft customer experience application team Marcus has been pivotal in launching Dynamics 365 contact center enabling EntERPrises to transform customer service with the power of AI outside of his work Marcus is based in Kansas City Missouri where he enjoys hiking cooking traveling and discovering the best barbecue spots with his family in the following conversation we'll explore out Copilot and Dynamics 365 customer service addresses the needs of large organizations by enhancing agent productivity reducing training time and delivering measurable ROI Marcus will also share his insights on the evolving role of AI in customer service and discuss some of the key feature that make Copilot Copilot a valuable tool for EntERPrises.
Samuel
Let's dive in so thank you so much Marcus for joining me today to discuss customer service contact center and more globally how AI will impact the the world of customer service agent working with those tools.
Marcus
Great it's great to be with you super happy to be here awesome will be really interesting so I know recently we announced our contact center product which people may be familiar with it in the name of Omnichannel for Customer Service I know there's difference we will probably cover them during the discussion today but first can you please start by giving us ey level overview of what is customer service what's contact center how do they complement each other so what's the difference between the two products certainly let me go ahead and just share my screen here hopefully you'll be able to see it I'll just start with the problem right like why are we doing what we're doing and I think we've all experienced this ourselves right that service experiences aren't very good as as a customer right you are on the left side of this slide and you are faced with this challenge of I want to engage a brand get some help but which channel do I choose like is it better for me to call should I maybe go to their app should I should I text them should I engage with them on my social channel should I email them there's all these different choices of Channel and depending on the channel you choose you may get a different experience because those channels typically have different self-service options in them they have different routing systems in them that ultimately deliver your conversation to an agent should a live agent need to be there right if the self-service solution doesn't work you got to go to talk to a live agent right so the same pain and confusion that we all have as customers is also on the agent side right the agents have multiple systems they have to deal with they have they don't have a big picture what's going on with the customer they themselves become as we call it human swivel teair middleware where they're moving data from this app to this app and this system to that system and so it's just super fragmented and super broken and then think about the people that run the contact centers they also have to deal with all these different systems and they don't get a big picture view of stuff either so it's every single customer that I talk to is in this boat right now like some to more or less right some are have better optimized their there their service environment some are not and they're still trying to get to the picture where we want to get them to which if I go to the next slide here is more of a complete solution more of a holistic approach to how we do service where you let the customer choose their Channel you have an integrated set of self-service options or we call kind of customer-facing Copilots those could be your traditional kind of chat Bots or voice Bots but Amplified and improved with Copilot and then we can route those conversations to live agents if we want to they have a single agent desktop they have a Copilot experience there to help them as agents and then supervisors and leaders get a big picture of view as well and as you said the intro right the thing that really is revolutionizing this generative AI it's the thing that's helping make all of this kind of come into this sort of a picture and one more slide and then I'll be quiet here is that you think about AI is really a way of accelerating that journey to that ultimate goal where it becomes this AI learning Loop where all the data that's generated through self-service transactions and through agent interactions and through surveys after a conversation all of that stuff becomes fuel for AI to make better decisions or to help agents make better decisions or customers make better decisions and ultimately take better take more action so it's that's the that's why I'm so excited about what generative AI and Copilot can do in contact centers and service is it helps everybody all along the spectrum of that journey and it as as data is generated it just gets better and better and better it learns basically based on the data that is generated so it's not only about AI right it's also about breaking the silos and making sure that the experience for the agent and for the customers are always the same and are consistent with what the brand want to.
Samuel
Yep and you can ultimately try to take some of the information you gather and the knowledge you get and as we call it kind of push it left so you try to push it out into the self-service channel so more customers can get the answers and the and take the actions that they need to in self-service before they ever need to get to an agent so it's kind of a win-win like customers a lot of customers don't want to talk to a live agent and a lot of lot of live agents are really burdened by a bunch of things that could be much better served with a self-service interaction I will assume by speaking with customer a lot of people are scared about using those tools and specifically maybe agent that can be scared of having to use complex AI tools yeah how do what you will you answer to a customer service agent that might be worried that it w be too complex to use yeah it's it's interesting the agents that I talk to there there's a little initial trepidation summary even worried that this going to replace them and they're going to lose their job take based on AI but then I think as they get to use it a little bit more they realize really this it's truly a Copilot it's truly truly there to help them do their job and be their everyday assistant their everyday AI companion as we say and it's as they start to use it they realize just how easy it is right it's basically just a chat interface if you use ChatGPT how to use Copilot right you can just type something in ask a question you get answers back but it's way more than that right it starts to become this thing that helps you get the right answer to your customers it helps you generate email ultimately it's going to start to take actions on your behalf as well that's some of the stuff we announced at at build earlier this year and then I know it was a big Topic at the Power Platform Community Conference just last week so there's a lot of evolution that's happening with co- pilot but I think all the agents that start to use it realize just it's just something there to help them it's another it's another productivity tool to make them that much more productive and basically is not theing the fun part of their job right exactly yeah most agents don't like to type up case notes and recap things and search around in a million different knowledge articles right they'd rather just have that interaction with a customer and solve problems so the more the Copilot can help them do that the better off they are.
Samuel
What what are the immediate or long-term business outcome that business executive out there looking to implement AI in their customer service department might expect from from customer service contact center or Copilot.
Marcus
Yeah ultimately it's going to make new agents more productive more quickly so it should shorten the amount of time it takes for agents to come up to speed and the amount of time it takes for them to get trained it should help boost productivity of the agents that they have today so basically the same number of Agents should be able to work with more customers more quickly and that gets into the last one which is the average handle time or AHT as you hear it in contact centers they want to want to drive that down so again not as much time is spent handling a conversation or handling a case because you don't have to search around to find the right information you have to keep your customers waiting hey stand by will I look that up sort of thing right Copilot it's there to help you do all of that and then even when you get into wrapup notes and those kind of things Copilot can help automate all of that as well so it's really get those new agents up and running more quickly make the existing agency have more productive and then help make the overall handling time less.
Samuel
You do have dataround the ROI from implementing those tools um.
Marcus
Yeah you can you can definitely quantify it I mean if you look at the average cost of an agent like average hourly cost right and just do the math of look if I' I've got these agents that cost me this much per hour and I can have that same set of Agents serve this many more customers this help helping me not necessarily expand my agent pool it's helping me serve do the do more with the same amount of people that I have you can also look at that on on a handle time perspective you can also look at all the what we in the slide you saw those kind of customer-facing Copilots so if you can have customer-facing Copilots chatbots IVR systems handle more things that's also a very quantifiable savings because those don't those don't cost nearly as much as a human being does sitting in front of the screen now you can as as we said Microsoft you can achieve more with less.
Samuel
Right exactly do more with less you got it.
Marcus
Now we'll see a world where people not embracing those technology will will lose their competitiveness com compared to their competitors right.
Samuel
That is also true I mean I went to a conference it was last year I think middle part of last year just when sort of all of the stuff was breaking into the mainstream and it was an Analyst conference and the head of the Analyst organization said this something we rarely tell customers but just get started now with something like normally an Analyst say well wait and see see what happens but they're like it's it's moving too fast and your competitors will do it and they'll take advantage of it so you need to at least start experience expen in and trying things now.
Marcus
Yes totally and actually I'm telling the same to to my customer and I think using AI is like you need you need to build this muscle right you need to build the bit of using AI for the agent but for the for the customers as well yep totally right and you're better to start somewhere so don't wait don't wait around because other people aren't waiting right.
Samuel
Then they're going to they're going to take advantage of it sooner than later.
Samuel
But how do you see that theyi will transform this the role of a customer service agent in the next three to five year but also how our organization interact with their customers and more generally yeah.
Marcus
I think in the next three to five years or so you're going to see this new kind of Copilot agent thing take off and this this gets to be kind of a complicated qu complicated discussion when you're talking about contact centers and human agents and Copilot agents so we have to parse it out a little bit but there's we announced back at build and we talked about it at platform conference you're going to see more on it as the months unfold here after the time we recorded this this but these agents are going to start to do work on behalf of a person or a team or a whole organization right you'll be able to ask have the agent go do something as opposed to you just asking questions and getting answers so a good example of that is something like order fulfiLLMent right so if I'm if I'm an agent and I'm helping you with you had something that was damaged in shipping and you need to return it and you need to replace my product ship to you once I validate that all that's accurate I could just have a Copilot agent go and do that on my behalf go and do the replacement shipment like go go send the new product to Samuel that corresponds to this thing in this quantity and then it just goes and do does the order entry on my behalf and comes back and tells me that it was successful so I don't have to do all that work anymore I mean I think that's transformational and in a huge way it's a it's an amplifier to the workforce you have having other other Copilot agents do the work that typically a human would have to do so so you'll be able to connect your customers service with any other system out there right other order entry systems fulfiLLMent systems even like other kinds of other systems of record or line-of-business applications or whatever like you or think you're you're trying to approve a customer for an increased credit line or a new credit card or something have it have the agent go out and get the customer's credit rating for you and return it right back into the Copilot experience as opposed to you having to go into a different system to look that up it's that kind of thing right I mean and that's just scratching the surface there'll be so many other things that Coop out agents will be able to do for people and for Teams and for whole organizations for sure.
Samuel
And how do Microsoft know plan to address those new trends and I think there are challenges that will come out of it as well yeah it it's it's an area where we're fortunate enough to have taken an early position and we're we're in a fairly good leadership position when it comes to a lot of this you're going to see this really start to roll out as we get into the next couple months I don't want to scoop our own news that probably would not be a good idea but if you watch what's going to happen over the months of October and November especially culminating with Ignite you're going to see a lot of interesting things that this showing kind of how we're going to continue our leadership and enhance our leadership but we've already got a pretty good track record now I was looking at the stats of the day I think we're up to like 2.4 Million monthly active users using Copiloting Dynamics and Power Platform so that's pretty impressive right I mean we got over almost two and a half million people using it now and that alone is a is a good thing in terms of it gives us great feedback from everybody using it like what what's working well what's actually helping them what's not working so well and how do we make it better so we have our own little feedback loop going to make sure that we can improve Copilot and make it that much better for future releases it's 2.4 Million only on Power Platform and Dynamics 365 Dynamics and Power Platform so it's a combination of Power Platform per Dynamics yeah wow now you got you got me curious.
Samuel
There's a lot that's going to be announced before the end of the year.
Samuel
Now I know you probably there's a lot of things you probably can't say but glob globally what are you the most excited about so that is publicly known yeah it's definitely that Copilot agents stuff I think that's really interesting to watch as to how that evolves where Copilot doesn't just be this chat companion where it answers questions for you and it's starts to do a lot more for you do a lot more on your behalf and it goes a little hand inand with Copilot agents is Copilot Studio if you haven't looked at Copilot Studio or what used to be called Power Virtual Agents way back when way back when like a year and a half ago or not even that maybe a year ago Copilot Studio is just so impressive now what in terms of the steps it's taken to make it really easy to build your own Copilots those customer-facing Copilots but also to go in and adjust and adapt and basically configure modify the Copilots that Microsoft has so either the Copilot in Microsoft 365 or the Copilot in Dynamics Copilot for Service Etc you can start using Copilot for Service or sorry Copilot Studio to go in and adapt and adjust the Microsoft Copilot experience so it's build your own but also go in and change ours as to basically better fit your needs you can literally extend the Copilot and giving access to different data source different system of Records exactly yeah the fact it can tap into all those different knowledge sources and it can do all the generative actions with plugins and everything like that it's really robust but you can started super easy right you can start started with natural language describe the copile you want to build and just type out a phrase and it starts walking through building the copil for you right and then after that you can adjust it and tweak it and make it more more know more intricately involved with your other line-of-business Systems whatever but it's it's a low learning curve and an easy way to get started and then you can go as deep or as complicated as you want to go as more people are using it as more data is pulled in and as more Copilot we're talking about extending it as more Copilot are created how does the platform will be able to endle this out the platform will be able to scale with with the with the customer it's I mean fortunately we're able to ride on top of the worldwide infrastructure that is Microsoft Azure so that gives us some sort of inherent scalability foundation and reliability Foundation based on Azure and things like Azure availability zones and all of that it's really something that we're super keyed in on to make sure that we can we can scale and meet the demands of the of the load as it comes in we actually been using it inside our own support Organization for the last oh a year or so now maybe a little bit over a year and we do 145 million customer interactions every year inside Microsoft's own sport organization and we've seen some really good results and really good performance so we think we're not only trying to ask our customers to try it out and give it a shot but we're we're using it internally and we think that's a pretty good testimony in case study as well oh we customer zero I can tell as well with the use of M365 Cal that yeah exactly customer zero is the key thing I everybody used to use that phrase eat your own dog food which I always thought was very disgusting so like customers Z much better now for people not knowing really the Microsoft infrastructure now can you can you explain what's what's the what are we using behind the scene to make those Copilot works yeah it's it's all again writing on top of azure but specifically Azure OpenAI if you look at Azure OpenAI it's the way of us basically having all of the AI capabilities inside Azure which is not even though it says Azure OpenAI it's not necessarily only tied to OpenAI models obviously we can take advantage of a lot of the OpenAI models that run in Azure OpenAI but we can also we have a lot of other LLLMs from other folks as well like metand what is some the other ones I think idia models in there now we have Microsoft small language models SLLMs that are in there as well and so we've got the I often times get questions from customers like what LLM are you using inside the Copilot for Copilot Dynamics experience I said well it's it's it's GPT-3.5 Moving to GPT-4o but really you shouldn't have to worry about that like let us manage that under the covers and make sure that we're we're taking the right LLM and using it for the right purpose and the fact that we're writing on top of Azure OpenAI helps us do that right we can we can move from this LLM to that LLM that's more tuned to that use case and not have to have the customer burden with that problem right we can do that on their behalf so that's the core component is azure OpenAI and then obviously we write on top of Power Platform as well so all the data that fuels the copil experience a lot of that data is inside Microsoft Dataverse which is the data layer inside our platform and then we use things like Copilot Studio and other things to make to make the whole experience happen that whole endend experience I showed at the beginning so it's really it's a combination at the highest level of the Azure foundation and then Power Platform on top of it so for people with experience with ch GPT which is pretty much everyone it's a really powerful tool but as well can sometime hallucinate so how do you make sure that Copilot inside of Dynamics 365 or Copilot in contact center will will give an accurate and relevant response to agent yeah this a huge question I get a lot as well so it's your your spot on in terms of asking this this point it's the key thing is the idea of grounding the copat experience in your data in your trusted data sources which is way different than the ChatGPT experience right the ChatGPT experience that everybody experienced Ed a lot of experience the ChatGPT thing that everybody tried a while back was basically acting over the entirety of the internet right so it's looking across the entire all the internet finding things right and sometimes there's stuff on the internet that isn't right it's like go go find that thing that somebody posted on a Reddit thread somewhere or somebody blogged about and that's not super accurate but it still got picked up by the LLM so in this case you're saying I just want my my model to be grounded in my data so my data could be the knowledge sources knowledge based inside Dynamics it could be knowledge sources and knowledge bases in other places it could be it could be my own publicly facing website it could be documents inside a SharePoint or OneDrive Library ultimately we're moving towards something that would allow you to connect to what we call a knowledge Hub you kind of see this in Copilot Studio today when you go and configure knowledge sources you can pick from a whole bunch of them that's going to be the same general approach that we take across Copilot across all Copilot experiences but being able to pick the the knowledge sources you trust is a key part of this right and that way you can make sure that you're not getting some spurious answer hallucination as we like to call them in the AI so what happen if I'm a customer service covering certain type of product in the system and we have this general knowledge base Capt that is captured inside of Dynamics 365 how do I make sure that the agent not going to access some irrelevant knowledge and or article that are stored in the in Dynamics for instance is it per role or yeah as the as the admin you're controlling first of all the what knowledge sources you're going to trust and then you even enforce at the mod at the user level like are certain users have access to certain knowledge do you even decide if you want to have some knowledge only used by certain agents and not by your self-service world right because there I talked to a customer that said well I might want my agents to know about pre-released product information but I don't want that published on my customer-facing Cod about on my website or anything like that so yeah you can not only control the a granular high level sense what knowledge sources you're going to trust but then you get more brand control meet that as well so I will assume as well that in term of GDPR and CCPA that's that's all something we're able to to support that following those gu those guidelines right exactly it's it's part of this May heard Microsoft talk about our responsible AI guidelines then privacy and trustworthiness and transparency and all of those things are a key part of that so we want to make sure that everybody understands that as you use Copilot the data your data is your data like we're not using your data to go and the models better or to co-mingle with other people's data to form a new set of knowledge that we can use in other places your data is your data it's in your own environment it's we're just we're going to make sure that we we adhere to the highest levels of compliance and we'll always post that up on the responsible AI website for people to take a look at I think it's one it's one AKA link I can remember it's aka. Ms if somebody wants to go out and take a look at the responsible AI stuff.
Samuel
Great you spoke about coop Copilot Studio to help extend extend the product can you tell us a bit more about Copilot Studio and how it can help tackle the knowledge that will be used to the different industry out there.
Marcus
Yeah it's this one of the most Innovative things I think the Copilot Studio has done this idea of saying I have a knowledge Hub where I can point it to a variety of different knowledge sources and if you don't have a knowledge Source that's supported like it supports Confluence and ServiceNow and Salesforce and OneDrive and SharePoint and Dynamics obviously and all these different places right but you can use any of the Power Platform connectors or you can build build your own connector right to connect into your own data source so it's you can cast as broad or as as a narrow of a net to try to capture the knowledge that you care about you can even upload a file or two to it like if you say well I want just's a special little chunk of knowledge I want to have that Word doc or that text file as a knowledge Source you can upload that as well so.
Samuel
How does Microsoft support EntERPrises and organization to the deployment and scaling contact center.
Marcus
Well the first thing is it's it's not just Microsoft supporting it we have this really awesome network of Partners all around the world that can come in and help customers not only first of all identify what they want to do and plan out their implementations their deployments actually help them through the implementation and deployment in the going live process and then supporting them after the fact so we have it's Microsoft plus all these partners that are out there so having a partner involved is super important I we think anyway the other thing though Microsoft has our own skin in the game so to speak so between Microsoft FastTrack which is a group that's funded out of our product team where I live those FastTrack solution Architects and Engineers can go in alongside partners and help make sure that customers are successful as they plan and Implement and go live the FastTrack team has been involved in thousands upon thousands of implementations all around the world so they've been able to learn from those and figure out the best practices what works well well what maybe doesn't work so well pitfalls you want to avoid those kind of things and then after the fact there's obviously Microsoft unified which servs our support organization where you can tap into our support team to keep you supported over time with customer success so it's it's a it's a multifaceted effort where we always want to make sure that customers are successful as they go through the Journey and then and make sure that at the flip side when it's all said and done they're willing to share their story with other customers and that's one of the things that I love to do the most which is to successful customers and then relay their stories to other customers.
Samuel
Do you have example of Fortune 500 companies that have implemented a product success story what what we've learned from those implementations yeah.
Marcus
Picking up on what I was just saying I love to tell the stories of other customers and what they've been doing so one of the ones that we actually featured in our business application launch event earlier this year is Lenovo the company that makes the I I think of them as the laptop company but I know they do a whole lot more than laptops a whole more than thinkpads I happen to have a Lenovo device right over here that I love really much and it works really well but they've seen a 20% Improvement in handle time that their agents are experiencing after they implemented Copilot so at the very beginning of our conversation we talked about improving average handle time well they've seen their average handle time improved by 20% so that's a huge huge benefit I also mentioned the Microsoft stats earlier like we're a Fortune 500 company we've implemented this internally I might actually share the slide I kind of mentioned some of the staff stat but let me go ahead and share it just so you can see the stats have a little bit of a visual aid if you want but here's just some some of the stats we've seen as we've done this inside Microsoft as Microsoft customer zero as we said before so first of all when even before we got into Copilot right just implementing our own Dynamics 365 customer service and contact center capability we saw a 20% reduction in misroutes where conversation was Landing with a live agent it shouldn't have landed with that person was routed wrong we've also seen a 31% increase in first call resolution which is another huge metric that most contact centers look at like how can I make sure that I solve this problem the first time could be a call could be a conversation through chat or whatever but basically first Contact resolution is probably a better way to phrase that and then with Copilot themselves we've seen three really interesting metrics we've seen the reduction in average handle time it ranges from 12 to 16% so Lenovo is actually better than we are there at 20% we're at 12 to 16% depending on the business we've also seen an increase in the number of cases that agents can actually manage somewhere between 9 and 12% again depending on the business and then the last one is really an interesting one about the decrease in agents needing to go get peer assistance like okay we don't want to stop Agents from talking to their peers but if the first place they can go is to Copilot and have Copilot give them an answer that they can use that means that agent doesn't have to bother another agent to get help right there's always going to be instances where an agent's going to have to seek help from maybe a more senior agent or someone who's more experien we're not going to stand in the way of that actually we have tools built into the product to help you do that with built-in Teams collaboration but if we can reduce that a little bit that just frees up more time for those senior agents to work on the harder problems and you see the size and scope of our our sport organization at the bottom I mentioned 145 million interactions every year and we're across all sort of countries across all sorts of types of customers as well everything from really large EntERPrise customers down to everyday consumers are using things like Xbox right so it's it's a pretty maybe not the largest sport organization in the in the world but it's definitely one of the more complicated ones so I think Lenovo is a good example Microsoft's a good example we've got a lot of other good examples up on customers. Microsoft. Com if you just go out and take a look at that website you'll be able to see a lot of other interesting customer stories.
Samuel
You you mentioned a lot reducing or automating some of the task reducing the workload I suppose that people are worried of losing their job right so what what will you ins search those people.
Marcus
Yeah I just think it's it's going to be on a case-by Case basis different customers may want to have different business goals and maybe some of their business goals are to reduce headcount a little bit I personally am a more optimistic person I think it's it's more about giving the agents more tools that they in their toolbox to do their job more effectively as you said earlier like eliminating some of the drudgery and some of the monotony out of their job so they don't have to do some of the work that they don't want to do and they can focus on the work they really want to do which is working with customers and solving problems it also helps those Junior agents become more senior agents because of the stat we just talked about I think it's it's actually going to help the same number of Agents do a better job at what they do and ultimately I saw some post somewhere on LinkedIn where it said you're not going to lose your job to AI you're going to lose your job to someone who is using AI like so it's not you versus AI it's you versus other people that have AI in their toolbox and so that's where I think Copilot really shines it's it's going to help the existing agents you have be that much better yeah.
Samuel
I think the same actually and I love that quote because it really represent it's a tool it's not there to replace you but if you if you don't use the tool you become less effective than the people that have the tools.
Marcus
It's like a car and one of my colleagues says it's not a replacement for your brain so like don't don't view Copilot as just I'm just ask up out a question what it says is exactly what I'm going to tell my customer like you still have to have some critical thinking here and say well is that is that the right answer is that the right answer for this customer in this situation so there's still some some human as we call human in the loop right where the human's going to have to look at some things and decide if that's accurate or not so I think it's just like you said it's a tool it's a tool just like any other but it can't replace it can't replace a person like any tools it will require time to adapt and start to use it properly making a bit of using it so what will you see best practice for organization looking to start using contact center yeah.
Marcus
We I mentioned one of them which is customers. Microsoft. Com looking out at what other customers have done right and that's a good way of understanding where customers are coming from and where they're going what their best practices are Etc another one is just basically taking advantage of some of the other resources we mentioned our partners our FastTrack team some of the success by Design guidance we have a lot of the learning information that we have up on learned up Microsoft. Com so those are the best places to get started but really as we said at the beginning one of the best things to do is just get started like get going try some things out maybe something just as simple as using it to help you generate email responses to customers if email as a channel right just try it out and get started and then you can learn and evolve and enhance your Copilot experience over time.
Samuel
Thank you thank you so much actually that's all the time we had together any final advice for CI CEO D that might listen to us.
Marcus
I think it's again just find a place to get started one of the places I also like to recommend people get started if they're little a little nervous about this think about their internally facing support organization internally facing contact centers things like your HR help desk or your it help desk or your facilities help desk or whatever right at Microsoft Samy we have this thing called ask HR where I can if I have an HR question I can engage the HR Group but I engage them kind of like a contact center right I go through a self-service bot if the self-service bot can't answer my question it can take me to a real life human and that real life human can answer it but being able to take these same Concepts and apply them inside your company is a good way to kind of get your feet wet and understand how it all works and then decide okay there's now that I'm comfortable now I can start using it outside the company thank you so much for the advice thank you for your time it's been a pleasure I learn I learned a lot I think people listening to us learned a lot as well how those tools are working what you can expect from it and what the future will will bring in term of evolution for those tools so thank you so much for joining us today Marcus thank you Samuel it's always great to see you always great to spend time with you thanks.