Latest / Talk Commerce / AI Customer Service That Turns Support Into Sales with Michelle Donnelly of Crescendo
Transcript
- Brent Peterson (00:03.539) Welcome to this episode of Talk Commerce. Today I have Michelle Donnelly. She is the CRO for... Let's start over. I already screwed up your name. We'll start over again. Donnelly, yes. Crescendo, yes. I even had it open. Okay, here we go. Start over. Welcome to this episode of Talk Commerce. Today I have Michelle Donnelly. She is the CRO for... Michelle Donnelly (00:15.342) Michelle Donnelly Crescendo Brent Peterson (00:31.773) for crescendo. Michelle, go ahead, do a much better introduction for yourself. Tell us your day-to-day role and give us one of your passions in life. Michelle Donnelly (00:40.268) my passions. Crescendo. So everybody knows Crescendo for music. It's peak performance. So Crescendo is an AI startup. We're two years old. We were started actually in a boardroom of General Catalyst, which is a VC. And they were hearing all these people pitching these really cool digital agents and how it was going to transform customer support. And one of our founders felt like there was a missing piece of the equation. And the missing piece of the equation was the human in the loop. So we believe that customer experience is a combination of incredibly smart, autonomous digital agents, truly autonomous, and I'll talk about that in a little bit, that seamlessly interact with the human being. And those human beings are powered by AI. And we also believe that speed to value is incredibly important. So we can deploy in 30 days and we're doing all of our deployment with AI. We have a team of AI agents, human agents. that do everything from quality assurance to constantly tweaking the dial to make sure that we don't set it and forget it. We're not selling seats. We're selling outcomes. We're here to make sure that your customers' customers have the very best customer experience. And that's what Crescendo does. You want to know about passion? Okay, so when I was 28, no longer 28, I got a big commission check. Brent Peterson (01:56.583) That's awesome and passion. Michelle Donnelly (02:05.766) And I decided to travel the world and I thought I'd go for six months and I ended up going for close to 18 months. I traveled mostly by myself and I was mostly in developing countries. So I've done everything from work with yogis to study with the Dalai Lama to riding elephants, camels, you anything you can imagine, boats, trains, you know, big boats, have visited the gorillas up in the mountains of Uganda. have lost my passport, have had drugs planted on me. It's been like, I really had, you know, the sort of time of my life. And I think what that taught me was just to always be bold and always be fearless, which, you know, I've always integrated into my personal, my professional life as well. Brent Peterson (02:47.099) Yeah, it's awesome. that was two years ago. That's, that's. Michelle Donnelly (02:49.582) Oh my God, we are going to have the best conversation today. Brent Peterson (02:52.723) All right, Michelle, I love that passion that's such fun. I'm in a group called Entrepreneurship Organization. I just had a couple of my friends that did a tuk-tuk journey where you start in Goa and you end up in Rajasthan. basically they give you a tuk-tuk, so like a little three-wheeled cart in India, and you drive on your own. So it's kind of like the... Michelle Donnelly (03:20.271) my god, that must be terrifying. Brent Peterson (03:21.787) It's kind of like the amazing race where, know, that you're not, it's unsupported, but in you're with a group. So it's kind of like who can get there the fastest, but it's more about the adventure. so that's on my bucket list to do as well at some point. Michelle Donnelly (03:34.414) You got to do it. India is an assault on the senses. It's really a magical, magical place. Do it. Brent Peterson (03:39.771) Yes. All right. Before we get into that though, before we actually, we're to talk about customer experience. And I have a lot of opinions about this as well. So I'm excited to talk to you. I'm going to tell you a joke. We're doing the free joke project. I'm going to tell you a joke. You just give me a rating eight through 13. So here we go. After the airline lost my suitcase, I took them to court. Unfortunately, I lost the case. Michelle Donnelly (04:06.826) Well, since I just lost my luggage coming back from Milan, I'm going to give you an 11 because it's very relevant for where I'm at right now. Fortunately, you can track all that stuff in the app, but that was good. Brent Peterson (04:14.436) Ha Yeah, it's much better than 10 years ago even. All right, let's, mean, let me just tell you my experience with bots and my, I really want to tie this into customer experience because I think people are doing it well and I think people are doing it horribly. And I won't name the service, but I just had a quick question. And in the past, the service had been so great with their agents online using the chat bot. And I got caught and they realized they had now employed AI in their chat bot. And I got caught into a loop where I asked them the exact same question four times and they gave me the same answer. And I was so mad, I canceled the service. And that's after being a really loyal customer for at least five years. that's my entry point into how AI can go wrong. And I know it can go right. So let's kind of open it up right there. Michelle Donnelly (05:17.794) Yeah, I'm going to go backwards a little bit. You know, I started at Salesforce selling, you know, all the clouds, sales service, marketing, commerce. In our panacea at the time was how do we enable the service cloud agent to really understand the next best action for the customer? You know, I then went to an AI chip company, which created the fastest chip in the world, really based on inference, which is to really seek to understand what's next. And that led me to Crescendo, which is this beautiful marriage between digital agents and human agents. And so to answer your question about your experience, we knew years ago that once AI got to a place where the voice, like talking between the agent and me would be just like a human interaction, we knew that would transform customer support fundamentally for the rest of time. But you have to have. the right kind of agent. So if you are a company and you have a digital agent and you don't empower it or teach it or design it to be AI native and to really understand what the customer is saying and then connect everything about that customer to the last conversation, to the last product they bought, to things that you might know about their family, to their preferences and let the agent do 98 % of the work to give you the best outcome, it's a hashtag fail. The other piece of this is that we fundamentally believe that AI is smarter than humans when it looks at all those things I just mentioned, including the knowledge base. But there's always going to be a complex problem that that AI agent cannot solve. And I'm sure you've been on phone calls where you've been screaming human, human, human, or pressing one, one, one or zero, zero, zero, right? So we also fundamentally believe that we have to serve the customer how and when they want to be served. So I personally love chat. don't have time. I like to multi multitask. If I'm in my car, I want to move from chatting with an agent to talking to an agent. And then when I want a human being, I want a human being. And so fundamentally our solution will connect that digital agent to the human agent. Michelle Donnelly (07:43.415) And when I talk to the human agent, they're going to have all of the information that I've just shared right at their fingertips because we empower our agents with AI. so I believe, well, the one of the reasons why I'm at Crescendo is because I believe we're doing this totally differently than anybody else out in the market. Nobody is talking about the seamless integration between the digital agent and the human agent. So you had a bad experience because you didn't have an autonomous AI. You didn't have an AI agent. that has all that memory and those integrations to all those systems. And you didn't have the ability to go to a human being when you wanted it. So we're doing things very differently at Crescendo. Brent Peterson (08:26.131) Yeah, my other experience was with a founder who built a a little tool that does it via SMS. think about enabling sales via SMS. And he found one of his AI agents had an SMS conversation for an entire weekend with a customer. I don't know what it cost him, but. There was no out, like, you know, there was no like, let's stop now. The customer just kept answering and they never, mean, the good part is they never realized it was not a real person, but they were talking about football scores and whose team do you support and all this other crazy stuff that had nothing to do with, had nothing to do with the service and everything to do with that agents AI is so wanting to please you. and respond that it never got a chance to say, there anything else you'd like before we shut down? know, like, I think there's always an out that's needed, right? And I think one thing you alluded to was that AI agents have to be smart enough to understand the frustration from the user when they're saying, I need to talk to somebody real. Michelle Donnelly (09:39.661) Yeah, I totally agree. like, here's the thing. know, Crescendo purchased a BPO. Okay. So we, again, because we fundamentally believe that the digital agent and the human agent are critical. Right. So we have 4,000 agents, you know, all over the world. And it used to be when you have your customer support agents, your human beings, that it's all about deflection. How quick can I get someone off a call to resolve the issue? Right? It wasn't about sales. was about service. I need a refund. I want to return my product. My product is damaged. My bill's wrong. Right? And those agents just want to get you off the phone. We don't actually care, about how long that customer stays on with the AI agent. We don't care. I mean, SMS might be different, but we're doing everything through chat, voice, WhatsApp, email. So we don't care. If they want to be on that call, or that chat. interaction, we don't care, it doesn't cost us anything. But that's not a great customer experience. And so what we really want to do is not just support that customer, we want to create upsell opportunities for the customer. So I'm going to tell you a story. I was just at the market this weekend. I live in San Francisco and we've got this great market on Sundays. And, you know, everybody's carrying these like they've got these like wagons, you know, to throw all their their produce into them and their kids and their dogs. And they're like pulling these wagons all over the place. And so one of the companies we work with is called Veer and their, their sort of tagline is it's family adventure gear. And so they use our service and our AI is integrated to their ticketing system, to their CRM, to their knowledge base. But it also creates these graphs about everything that that customer cares about his bought questions that's asked over time, right? We call it crescendo influence. So I'm going to give you a scenario that I actually found fascinating. So I'm Michelle and I have a two year old and a four year old. I actually have a 21 year old, but she won't fit in the stroller. And I call the year and say, Hey, I I bought a stroller last year and I need a new canopy. Michelle Donnelly (12:01.07) for my stroller and I saw the order number or the item number and it's $79. So this is what I want to buy. And the AI looks and says, oh, Michelle, you bought another model that $79 unit that uses said you wanted is actually the wrong one. So Michelle, here's the right canopy that's going to fit on that stroller, which is great because now I'm not returning the wrong item and I'm having a better experience, but it doesn't stop there. So my AI knows. that Michelle mentioned right when she got on the phone that I was doing a family adventure, which is why I needed a new canopy. So now it's saying, well, why don't you get an all-terrain weather cover as well because you're going to be out doing this adventure and I'll go ahead and put that into your order basket. It also knows that I have a kid and that my kid in six months has gotten older and now my one-year-old needs more of a comfort seat versus what they had before. So it recommends this comfort seat. And so, this goes on. And then I say, like, well, you know, I have a pet and I don't know that if I'm supposed to put my pet in your, your, in the wagon, is that okay? And then the bot comes back and says, look, like these aren't designed for pets, but we can handle up to 300 pounds. And, you know, if you're going to put your pet in it, you might want to think about this accessory to make sure that your pet is safe. I'll put that into your bag as well. So This is what's incredible about our AI, right? It is actually thinking. And then when I hang up the phone, it's going to have all this data around me, about me, so that the next time I call in, it can continue the conversation. And it never goes away. So to me, and by the way, never needed to talk to an agent. Now we have agents, but never needed to talk to an agent. Now if at some point I said, I want an agent, You know, our customers decide on the rules. Some say, as soon as you say, want an agent, you get connected. Some say, hey, the wait time is 10 minutes. Can I ask you some more questions? So we do follow like the customer's brand and the tone of voice and how they want to talk to their customers. But I want you to think about the power of what I just described. We have merged support and sales. We have driven a $79 order for that one item into over $300. Michelle Donnelly (14:21.112) That is 3X. That to me is powerful. And the customer's really happy. Brent Peterson (14:28.903) Yeah, that's awesome. I think I like the other part about talk to an agent. it's going to be, you like you're going to be in the queue and I can think of myself being in the queue and in XYZ Corp and it's, you know, your 10th on it's kind of like waiting for an Uber. Your Uber is going to be five minutes away, which really means 15, right? A lot of customers would be open to, answer some, do you mind just help? can help you while we're waiting. You know, I think that's a really good approach. And some customers like, no, I'm so sick of talking to you. I just want to talk to a human. I'm happy to just sit here and drink my coffee and watch the little ticker go by until a real person comes along. That's such a great opportunity for AI to assist people. And then also I think the other important part to remember is making sure that your agents learn from what customers are doing. So like if customers are saying, no, you are super annoying, I need to talk to a person or I'm so frustrated. Or if they start swearing like, okay, this is a really great place to now just get the agent out of the picture and into a real person's hand. Michelle Donnelly (15:43.138) Yeah, for sure. For sure. Brent Peterson (15:45.555) What are you seeing in terms of that sort of crossover, helping a move from the agent to the live person? I really like, tell me a little bit about moving into a voice. I think that's another great way to hand off and would they move to a voice in a chat and on a phone call? Talk a little bit about that. Michelle Donnelly (16:02.69) Yeah. Yeah. Michelle Donnelly (16:08.462) Yeah. Yeah. Yeah. It's really powerful. It's called multimodal, which means that you, the customer can decide how you want to integrate with our AI agent. Right. So you can text, you can talk to it. And again, I was talking to you early about the power of these new LLMs, right? How fast they are. So you and I can talk like if you're my AI agent, like almost as if it's a human agent. So the power of that is, let me give you another example. We work with a sprinkler company and we have an app and it authenticates you in so it knows everything about you and how you're using the sprinkler and the sprinkler is a smart sprinkler. It knows when it's going to rain. It knows if your soil is too damp. It knows what time of year it is, et cetera, et cetera. There may be a situation where you're trying to do something within the app to fix something or to actually change your settings. And I don't know about you, but I can't text and change settings on my app at the same time. So I could be texting saying, hey, look, my sprinkler is, my soil is too moist. I want to change it from two days to three days. How do I actually do that? Right. And I'm typing that. And then the AI would say, hey, look, I can walk you through how to change it in the app. Let's go to voice. And then the agent just starts talking to you. And then you start talking to the agent. So now. I can talk back and forth with the agent as I'm on my app fixing the settings, right? So we don't care how you integrate. It's particularly important when we're giving you instructions, like how to go from A to B to C to D to set something up. We've done it with lots of different customers or like we might be able to send a video across to show you how to do it if you can't figure it out. Even we can send links on how to buy things too. So you can buy... You know, I remember when I worked at Ask G's, which was like a long time ago, it was a competitor, was a search company. And we used to do related search, right? So you bought a car, so therefore you might want tires, right? It would just do, it was a recommendation engine. Now our AI can actually send you links based on what you bought before that are relevant, right? So it's not just like, Hey, Michelle bought a car, like buy some tires. know, Michelle bought a car, she likes blue. Michelle Donnelly (18:33.238) She's got a five-year-old, she needs babyseat, babyseat We've got a partnership with Veer. I'm gonna send her a link that also has a 20 % discount so that she can connect that to her car. And then I'm gonna send her a video about how to connect that babyseat. I mean, it's like so powerful. And like a lot of people say, well, I don't want AI in my chats. Why not? It makes my life easier and I'm a busy human. So it went off a little bit, but this multimodal approach I think is critical to be able to serve your customers how and when they want to be served. Brent Peterson (19:08.807) Yeah, mean, think just having, yeah, I think you said it perfectly there. One thing that I've heard is the ability for AI to know as much about what you're doing. And a good way to do that now is to simply create more content on your website. So I think the mistake there is to create content with AI that just is generative, that doesn't say, is it doesn't... reflect anything unique, and just going back to the humans in the loop, creating content like as a blog that supports a product, and then having your AI read that to know more about it gives you an opportunity to even use some of the feedback you're hearing from customers who are asking questions that may not be answerable from the AI, then to create an opportunity to create content for the site. Talk a little bit about that. Michelle Donnelly (19:51.192) We see. Michelle Donnelly (20:06.318) Yeah, I want to talk about quality assurance. I want to talk about patterns and trends. And then I want to talk about content. All let me talk about quality first. Because what you're trying to do is create a great customer experience, Whether they're on your website or they're using your chat powered by the digital concierge. So you're trying to create a great experience. And your goal as a brand is to get your customer what they need when they want it in the right timeframe, right? Just in time service. So to me, it's not really about content. It's about being relevant. All right? So we have a whole team of people that are looking at all of the interactions that the brand has with their customers. So I'm going to give you another example. So we have a shoe company that we work with. It's a very well-known shoe company. And people love the brand because they do a lot of work, lot of humanitarian work. We noticed a pattern where customers were highly frustrated that if I ordered a size 10 in a sneaker, that same size 10 that I wear would not work with a boot. And so I didn't know how to order their products anymore. So instead of ordering one product, I'd order a nine and a half and a 10. or a 10 and a 10 and half, I didn't know how to buy from them. And it was super frustrating. And I actually stopped buying from them. It was a bug to me that a 10 wasn't a 10. And so they got all these patterns and all this information from us, and they were able to address the sizing issue, right? So that when you look across all of your customer base, it's not just what Michelle wants, it's the patterns that you see emerge around your customer experiences. So. The notion that the QA can then inform how to talk to your customers, how to build products for your customers is really important. And then you get into sort of this notion of content and you know, we're really different. you know, we built our AI with customer care and, and, and retention in sales in mind. And so we don't, our AIS, our AI is not built on Michelle Donnelly (22:27.502) rules and regulations, our AI gets better as you interact with more customers. So you have a knowledge base, right? So that's the content, right? But we're not fully focused on the knowledge base. We learn from your knowledge base. We make your knowledge base better, right? That's something that customers always say is, oh my God, my knowledge base is, we don't care. Give it to us as a starting base. Tell us like what you care about. And as we talk to more and more customers, we'll actually improve your knowledge base and we'll actually come up with better answers to your customers questions. And we don't send like paragraphs of, of, of content. It's like, I want to know if your stroller can carry my dog. Michelle, absolutely. We are strollers can carry up to 300 pounds. You know, they're not designed for pets. However, like if you, if you use these accessories, you can make sure your pet is safe. Done. So the content evolves and it becomes customized and personalized to that consumer. And I think that makes a much better experience versus just like creating content for your website that may or may not be relevant. So think about it as every time you hit my website, I can create customized content depending on what you need. Brent Peterson (23:50.705) Yeah, I mean, that's such a great way to look at it. think that, I mean, obviously, I think even five years ago, people were starting to talk about how can we make our AI bots or that I think they were just called bots at the time. How can we make them smarter? They weren't connected to an LLM, but they were autonomous to a degree, right? Talk a little bit about how this has, give us some impact if you have numbers that you can kind of reference that help us understand the importance of like that. That first example I gave you where it's just, you know, it's just the raw LLM that doesn't necessarily have a lot of ways to help the ways out. I don't know how, I don't know a nice way to say it, but the worst case and the best case, how do you, do you have some differentiations in that? Michelle Donnelly (24:48.946) so in terms of how we build our AI, we use the most powerful LLMs that are out there and we're super nimble, right? so we're always looking for the, the latest and greatest. do a lot of testing to make sure that the different LLMs that we use will actually be successful. and then we look at, what resonates with our customers. And then what we do is we look at things like how do you turn a conversation into a revenue or retention opportunity by combining that personalization I talked about and the memory that I talked about and your business context. So in terms of like some stats that I could share, we're finding right now with these early deployments where we use this, it's called. Crescendo influence, which is what I was talking about before, that in early deployments, 30 % of interactions influence a product selection. So I want you to think about that. In a typical contact center, it's a cost center. And now it becomes a profit center. This becomes really important too when you think about your traditional contact center. We're selling an AI native contact center, which is a digital plus the human agents, right? So when we go out and talk to CFOs, this is just another step that I think will be interesting. Our solution, like we can take over your whole call center, your agents, your BPO, right? So let's say again, you're a retailer and you've never implemented a digital concierge. Let's say you've got 10 customer service reps that are currently on staff or you're outsourcing them to Honduras or the Philippines. We can say to the CFO, listen, we're going to take all this risk off your hands. We will rebadge those agents or we can use some of our own. And then when you start to deploy these AI agents to your site, we will manage that end-to-end experience and we can save you anywhere between 10 to 30 % off your current spend. And we can guarantee you. Michelle Donnelly (27:08.93) consistent flat monthly pricing so that you're not dealing with your holiday spikes or maybe there's a problem with your product and there's a huge spike because there's a problem with your product. So what we really try to do at Crescendo is reimagine not only that end-to-end experience, but also how we charge and we don't charge based on seats. I came from Salesforce and Salesforce was all about seats. We don't charge based on that. We charge based on outcomes. We only want you, the customer, to pay if you get a good outcome. So we manage all of that. We save you money. We give you flat pricing. We deploy in less than four weeks. And we can get you speed to value very, very quickly. This is really important. I came from the world of SaaS. And you'd buy your CRM. And then it might take three, four, six, 12 months to implement. based on how big the deployment was, right? We don't need any of that. Like our deployment's included at no cost. So like we are truly transforming the way that customers interact and have a great customer experience. And then we're creating great experience for our customers because it's the speed to value, it's the cost savings, it's all the Q and A so that you understand. how to reimagine your product or reimagine your customer experience. like honestly, Brent, I'm having the time of my life because this solution is selling itself. So I'm actually scaling my go-to-market by 10X. it's been, and we're quickly going up market too. We started sort of more mid-market enterprise and we're quickly going into these large strat accounts as well because the speed to value and the cost and the customer experience are incredible. I wonder if I answered your question. Brent Peterson (29:04.019) Yeah, no, that's good. geez, there's so many questions I could ask and we're running out of time, but like one of the things that was talked about in the last couple of years on what AI is doing is democratizing. You'd said to your mid you've served the mid market. A lot of solutions like this in the past, like Salesforce, or, know, one of the other larger platforms are geared. only to the enterprise market. Talk a little bit about what your ideal customer is and whom you can serve that maybe wouldn't have been served five years ago because the technology was just too expensive. Michelle Donnelly (29:43.151) Yeah. All right. So we can serve anyone and everyone literally, as a CRO, you sort of want to put things into verticals or boxes. But the reality is, is that, you know, if you have customer support pain, we can solve it. interestingly, Salesforce started off as an S and B company, right? Remember like they went in front of the Oracle, their big conference and said no software. And they went up into the cloud and no one could imagine, you know, buying, you know, from the cloud and they completely transformed. you know, the way customers buy SAAS. And I am a huge fan of Salesforce. It was like 10 of the best years I ever had. They talked a lot about the customer journey from sales to service to market to commerce integration. All this happens now in a chat bot. I want you to think how incredibly amazing that is. This all happens on my desktop in a chat bot. Everything I want, marketing, sales, service, commerce. I mean, it's actually extraordinary. This is a game changing solution. And then I got a human being. when I want it. So Salesforce started as an SMB tiny, right? And it quickly moved up market. And so now everybody thinks of Salesforce is like, you know, they only do big gigantic companies, but that's not true. They started off an SMB. We started off small because we wanted to move really fast. Our sales cycles go anywhere from 14 days to 90 days. When you move into, you know, more of that sort of enterprise, a very large big box brands, you can't move as quickly, not because of us, but because of their own you know, internal, you know, reviews and how that works. But we are seeing a lot of very large customers come to us saying, Hey, I tried a platform that had an AI agent and it's six months and it's still not deployed. Can you do it? And my answer is I have four weeks and I can get this up and running. I can get it. can actually show you in two days, a custom solution for you. And customers love that. the big customers in particular, because what they're doing is they're with all this AI. There's a ton of noise, tons of noise. It's really, really, really hard for customers to figure out what to do. And we can come in and show them very quickly how powerful it is. And I'm a big believer that seeing is believing, showing is believing. So we're having huge success going after some really big competitors. Michelle Donnelly (32:06.904) that you wouldn't imagine we would be taking business away from. Kind of fun. Brent Peterson (32:11.835) Yeah, kind of like David and Goliath, Michelle Donnelly (32:14.126) We are, I love that position. It was like a position with Grok and it got acquired by Nvidia. So seen it, done it, been there. Brent Peterson (32:16.357) Yeah, that's great. Brent Peterson (32:23.003) Right. Michelle, we are running out of time, but as we close out the podcast, I give everybody a chance to do a shameless plug about anything they'd like. What would you like to plug today? Michelle Donnelly (32:34.328) Look, I just want you to come to our website, crescendo.ai. I'll send you a video Brent, so you can see some of our customers in action. I'll send you a video where someone, we tell the customer that it's AI and by the end of that conversation, they forget they're talking to AI and they're asking about the weather and the kids and everything else. Come to our website, check out our chat, talk to me. I'm happy to help you with any of the customer service problems you have. We're doing great work. Brent Peterson (32:59.399) That's perfect. Michelle Donnelly is with CRO with Crescendo. I'll make sure we get all those notes into the show notes and the video and everything else, all the supporting material. Michelle, it's been such a great conversation. We could have had another hour. Thank you so much for being here. Michelle Donnelly (33:13.976) Brent, thank you for having me. Good luck.