AIMinds #046 | Oren Goldschmidt, Co-Founder and CEO at Mova AI

About this episode

Oren Goldschmidt is co-founder and CEO of Mova which is focused on helping companies understand what really goes on when they have conversations with their customers. Oren has been building software startups since high school but took a few years out to get a PhD before diving back into the startup scene. He’s launched and sold several startups and spent a few years as a VC before deciding that building great businesses is more fun than finding them.

Listen to the episode on Spotify, Apple Podcast, Podcast addicts, Castbox. You can also watch this episode on YouTube.

In this episode of AI Minds, Oren Goldschmidt, co-founder of Mova AI, shares his fascinating journey - from starting a high school startup to becoming a philosopher, serial entrepreneur, and a leader in conversational AI. Oren recounts his early days in the UK, his ventures in sunny Southern California, and his experiences navigating the complexities of startups and venture capital.

He later dives into the creation of Mova AI, a platform designed to transform customer service interactions by leveraging advanced analytics and AI. Oren offers insights on building relationships with large enterprises and explains how Mova AI helps companies enhance customer engagement through innovative AI-driven solutions. Tune in to learn how Mova AI is revolutionizing customer service and shaping the future of enterprise communication.

Fun Fact: Oren Goldschmidt began his entrepreneurial journey in high school, marking the start of his diverse career path which blends technology, philosophy, and business.

Show Notes:

00:00 Joined venture studio, gained accelerated startup experience.

04:46 Creating personalized experiences requires understanding conversations.

06:49 Personalized service through assigned dedicated bankers.

11:52 Inaccurate CRM causing agent inefficiencies, A/B tested solution.

16:19 Alerts on trends, issue reports, customer call analysis.

18:04 Monitor and address issues using threshold triggers.

20:56 Metrics and contextual insights improve performance evaluation.

23:32 Improve processes to enhance customer service.

More Quotes from Oren:

Transcript:

Demetrios:

Welcome back to the AI Minds podcast. This is a podcast where we explore the companies of tomorrow. Built AI first. I'm your host, Demetrios. As always, this episode is brought to you by Deepgram. The number one speech to text and text to speech API on the Internet today. Trusted by the world's top enterprises, conversational AI leaders and startups, some of whom you may have heard of like Spotify, Twilio, NASA and Citibank. In this episode we're joined by Oren, the co founder of Mova AI.

Demetrios:

How you doing today?

Oren Goldschmidt:

I'm doing great.

Demetrios:

Right, so you've got quite the journey in multiple startups that you founded and sold. You came from London or the UK and you moved over to the us. You're currently in a sunny part of the US and I am in a cloudy part of Germany. So I am very jealous as I speak to you. But that is not what we're here to chat about. I really want to know a bit more about you and let's start at the first startup that you created because. What it was in high school that you started your first startup?

Oren Goldschmidt:

Yeah. Towards the end of high school I was involved in a startup that I helped found which was fantastic. Maybe not perfect timing, but it was great. Really enjoyed that. And I originally went to university to do computer science. That was seen the obvious thing to do. That's what I'd been doing. But on arriving at this kind of amazing place, I quickly thought, hey, spending three years of my life being taught Java in a lecture hall with 200 people is probably not the best use at this time.

Oren Goldschmidt:

So I quickly switched and I studied philosophy and I had an amazing time doing that. But then after school, quickly went back to the world of entrepreneurship. I helped found a startup in the health behavior change space which we later exited, which was great. And actually took some time out after that to go back to university. I had an itch I wanted to scratch there and did a PhD, which was the least strategic but also the most fulfilling thing I've ever done in my life. It was fantastic.

Demetrios:

Those things you can do after you sell a startup, right?

Oren Goldschmidt:

Right. Yes, I think that was the right order. I'm glad I did it that way. And I've been in the US for 10 years in Southern California. Like you said, I originally came out here to be a vc, a very hands on vc. Came to join a venture studio out here which was just an amazing learning experience. I was head of incubation and a couple of other things and got to Watch a couple of dozen startups go from nothing to idea to real product and some exits that it was an amazing learning experience to get all that experience, decades of experience crammed into a few years, which was fantastic. But I also learned that.

Demetrios:

And you couldn't sit on the sidelines for too long. Right? Right, Yeah.

Oren Goldschmidt:

I mean one of the things I learned was like, I don't like watching other people operate. I prefer to do it myself. It's kind of frustrating, even though it was amazing. So after that I went back out as a founder, slightly different model, got together a bunch of consortium of banks and got them to work together with our team building a product called Engage, which is all about customer experience, customer engagement, something they were struggling with. And that was, that was a lot of fun, a lot of fun to get to build with large enterprises. And just from the start they know you have nothing and you're going to build something with them. And I was amazed.

Demetrios:

And they're okay with that? Yeah, they, they were willing to take the bet.

Oren Goldschmidt:

Yeah, absolutely. Yeah, it was great. We convinced them to sort of. And this is a time when everyone was trying to, you know, the corporate space innovation was huge. So, so it was, you know, they wanted to do something new and we helped them and, and that was great. And one of the, this is where kind of the genesis of, of Mova AI really comes from. From that. It was the work we were doing there was helping these, these large companies with hundreds of thousands of millions of customers figure out how do you build relationships with these customers, particularly when they call you or they're on web chat and they talk to you or send you an email.

Oren Goldschmidt:

How do you make that experience differentiated? How do you make them feel like they matter to you? And we build a product around that. But one of the things I realized in doing that, and that was good, that was successful, we extended that. But one of the things I realized in doing that was that, even when we had these amazing customer services people especially trained and we could create long term relationships with customers, we didn't really know what was going on in those conversations. So we had all these conversations going on, seemed to be going great. But what is actually happening in this conversation? What works, what doesn't? How do you scale this? We just didn't have that information. And that's where Mova comes in. And Mova was created to do exactly that.

Demetrios:

I don't want to jump into MOVA too quickly because I'm intrigued by Engage and starting from nothing Going to financial institutions, which in my naive view don't seem like the ones that are going to be open to working with startups. But I guess you prove me incorrect on that. And especially in these days, like you're saying of the innovation and digital transformation being hot. And so everyone wants to stay up to date or be on the edge of that innovation curve. In Engage, what exactly was it that you were selling?

Oren Goldschmidt:

Yeah, so imagine when you want to connect with someone at your bank if something goes wrong or you want to do something complicated, you want some advice? You can pick up the phone. Web chat is more and more popular. Chat and the app is popular, but you'll talk to a different person every time. Right. There's no consistency there. And for a lot of banks, some of the banks we were working with, they saw themselves as differentiated. They see themselves as differentiated on service. Like if you go to B of A, you get the scale, they're everywhere.

Oren Goldschmidt:

But service is not what they sell on. Whereas these small, these kind of mid sized banks, community banks, credit unions, they sell on service. so what Engage did was say, hey, you can pick the banker that you want to have a relationship with and they are going to be your banker for as long as you're with the bank. So you get a private banking experience. And we were able to scale that so it costs no more than running a traditional call center. That gave customers a much better experience. So that was the core idea and we had video and voice and some AI and other things there to make that more efficient and effective. But that was the core idea was we'll give you a personal banker, a real personal banker who will follow you and it's there when you need, do you need help?

Demetrios:

Oh, I love that. Then you don't have to repeat the same thing to 10 different people every time you call the bank and you say, oh yeah, but I told the last person this, it's not in your system. They didn't save that. And so hopefully this banker remembers you. But I could also see where this banker has like a thousand people they are responsible for and they don't necessarily remember every interaction with each one of those people.

Oren Goldschmidt:

Yeah, and that's where you have a conversation history that you can refer to and stuff like that. And it works great. It's a great customer experience. And we enjoyed building that and enjoyed the model. I mean, building that with customers from the start was great fun.

Demetrios:

And how would you do it if the banker wasn't available when I needed to talk to them?

Oren Goldschmidt:

So you still have the chat history. So if someone else jumps on the chat, that's fine, but they can see what you talked about yesterday, so you don't have to repeat yourself. So just the medium and the way we structured it takes care of a lot of the frustrating things about customer experience, just by default. But the problem, like I said, was how do you scale this? How do you get enough people who are good at this? How do you train people to be good at it? Because it's a new kind of skill, really, in customer service.

Demetrios:

And jumping into mova now, not only how do you scale it, but the other thing that I think of is how do you know that there's quality conversations happening? And how do you know which conversations are quality? So how do you set the metrics? What metrics are important to you? Is it time on a call or is it a satisfied customer afterwards? Did you look at all of those different metrics? And I'm sure with Mova now, you're really thinking through that type of stuff.

Oren Goldschmidt:

Yeah, I mean, for most call centers, most contact centers, it's. The key metrics are. Are handle time. Like, how quickly can you get a customer off the phone, just to be blunt about it. And to some extent, customer satisfaction, which is a really flawed metric, because the only people who answer customer satisfaction surveys are people who had a great, amazing experience. There's very few of those, but mostly people who are annoyed at you for something, and they never tell you why. They just tell you they're annoyed. So it's not a.

Oren Goldschmidt:

Neither of those are great metrics. So, yeah, with mova, it's a different world. And with what we can do with AI now, you can go into an incredible amount of depth to understand what went on in calls and understand it statistically across all your calls. What percentage of my calls did this happen on? How often does this happen? Not just anecdotes, but, hey, this happens 3% of the time, and that might matter. That might be a big deal. If you're losing those customers because you do something 3% of the time, you probably ought to stop doing it. So, yeah, there's a whole new world of insights that you can have. If you can actually listen to every call and understand what's going on.

Demetrios:

Okay, so we're dancing around what MOVA is. Maybe now is the right time for you to explain it, because I have so many product questions, and I am hoping that the next time that I interact with customer service, they are using Mova, because I've had some horrible customer service Calls, especially when it comes through my mobile phone provider and carrier. I want to really give a shout out to them for being really bad. But tell me, what is mova?

Oren Goldschmidt:

Yeah, so mova, it's what's called conversational analytics. That's the space. And what we do is we ingest audio recordings of phone calls. We can do it in near real time, but we can also do it for back data. So we ingest all these audio recordings of calls and use Deepgram of course to transcribe them, which been fantastic. And then we also take in web chats and texts and emails. So all the communications you might have with the customer and then what the MOVA platform lets you do is ask a bunch of questions across all those calls. So you might look at just identifying call purpose is a huge thing.

Oren Goldschmidt:

So what were they calling about? Often people have a sort of drop down in their, CRM to say what the call was about. That's we found that's really inaccurate. It's about 70% accurate, but it tends to be very inaccurate, particularly if something negative is happening and it's not very detailed. So being able to just understand, hey, why people are calling, you know, we have a customer and their primary core purpose, they didn't realize what it was. But most of their agents, and I won't go into too much detail, but most of the agents were spending time doing this really menial task that they didn't have to do. So we were able to say, hey, let's do an A B test, let's put a script in where they suggest that the customer does this in future in a different way. And suddenly their agents are freed up to have more meaningful conversation with their customers. But without actually analyzing the calls, they didn't know that's what was happening.

Oren Goldschmidt:

So you can really open your eyes as to what's going on. And MOVA is really about, it's about getting those insights and it's about getting to depth. So not just them as saying, hey, why were people calling, but what goes wrong in 3% of the calls, what goes wrong in 5% of the calls, what are the triggers, what happens and how do our best agents avoid that or turn calls around, which is really interesting if you can figure out, okay, if this happens, I'm going to train my agents that if this happens, you want to try this approach because it works 90% of the time, whereas what you might do just off the top of your head is ineffective. So yeah, it's great training material to make agents just perform at a much higher level than they would, you know, just with the basic training you get as a call center agent.

Demetrios:

And are you able to specifically ask questions like, what is a commonality between the calls that are our shortest duration?

Oren Goldschmidt:

Yeah. So, yeah, the way our, our platform works, there's a couple of stages. You imagine it. Like, There's a kind of cycle going on where you're curious as to, hey, I'm just wondering does this happen? Or how often does this happen, or does this happen regularly? And you want to kind of explore. So it's kind of an exploratory piece where we use AI to let you ask for general questions like that. And then you have a kind of monitoring piece where you're saying, okay, well, now I want to know. I want to see statistically how often this happens. I want to see if that's going up and down.

Oren Goldschmidt:

I want to see trends. I want to see which agents are doing this, which agents aren't. And then once you.

Demetrios:

And that's with you tagging some metric that you say, this is very important to me, let's start tracking it.

Oren Goldschmidt:

Yes. Right. Yeah. You find a thing that you care about that's meaningful, that's driving outcomes, you tag it and you say, I want to know. I want you to tell this over time. I want it by team, by agent. I want to essentially manage for this. Right.

Oren Goldschmidt:

I'm going to measure it, I'm going to manage for it. Then I'm going to go to the next thing that I can look to improve, and then just keep going through that cycle, and you can really ramp up the quality of what you're.

Demetrios:

Doing and does it. Do it retroactively? So this metric's important to me. Can I know how we've been doing in the past? And then also I want to be able to track it as we move forward.

Oren Goldschmidt:

Yeah, you can go, you can go back over, you know, as many calls as we retain. You can go back and you can look historically and say, oh, wow, I never knew that was happening. How often has it been happening the past six months? Oh, quite a lot. Okay, well, I better do something about that. Better go fix that right now. And then you can measure that it is fixed. Right. I went and I trained people today and tomorrow.

Oren Goldschmidt:

I know whether they're doing what they were trained to do, which is great. And it's an open platform as well. So the idea there is that it's not that we have A bunch of fixed insights. You can go and ask the platform whatever questions are relevant to your contact center, your people, your customers.

Demetrios:

And so that's me being able to almost customize the insights that I'm looking for. But you are surfacing insights as they arise. Like if all of a sudden the call duration is normally 15 minutes and the last hundred calls were all five seconds, maybe we need to throw a red flag or something. Is that you get an alert.

Oren Goldschmidt:

Yeah, So you can get alerts on trends, you can get alerts when specific things happen. We also do exception reports. So if you have a really bad call, and this happens all too often where something goes south with a customer relationship, a call doesn't go well. MOVA will actually produce a report and say, hey, here's the call, here's what happened, here's what went wrong, here's what the agent did, here's what they were supposed to do. Sometimes we know what they were supposed to do. Here's what you might want to do to fix it so you can, you know, something goes bad. First of all, you can call the customer and say, ah, sorry, look, let's fix this.

Oren Goldschmidt:

And secondly, you can go around to the agent and say, hey, this is how you should probably have handled that. This maybe didn't go so well. Here's a better technique for dealing with this. So we use those a lot. Those are great, you know, because it just, you know comes, you know, report comes straight off the system and tells you exactly what you're doing wrong.

Demetrios:

Yes, that is what I needed. And that when I talk to Vodafone, those guys. Oh my God. So the thing that I'm wondering too here, when you're talking about these reports is how do you flag, how do you know that a call has gone bad?

Oren Goldschmidt:

Yeah. So there are different ways to do it. So one is you can just, hey, ask the AI to hey, let me know when something has gone south. And that's okay. But really what you want to be doing is getting to the specific things that you care about as a company. So it might be in financial services, it might be a compliance issue. So you might have a super friendly call where it went. Great, customer satisfied.

Oren Goldschmidt:

But you made a slip. And in, for example, the mortgage industry, a slip can cost you tens of thousands of dollars. So you want to correct that slip immediately. So it maybe that or maybe a dissatisfied customer, it's a more typical scenario. So yeah, you want to set up those thresholds and so whenever you can do, you can find things, hey, if this happens, you want to trigger a report or you can say if this combination of things, you can create a score and say, hey, if this score goes above this threshold, that's a problem. Let me know when that happens. So yeah, you can do look at aggregate, you can look at individual events like compliance issues and make that generate an exception report. And then of course, if you've got scores, that's great because you can then use that to score agents and teams across a whole bunch of things you care about and say, hey, scores are down this month, let's look at why, let's do more training, let's figure out what's going on.

Demetrios:

And can you set up red lines so if agents all of a sudden start recommending to buy the next ICO of your favorite meme coin?

Oren Goldschmidt:

Yes, yes. Yeah, yeah, you can definitely set up red lines. And that's interesting. Like one of the things that, that has been great for us with, we use a whole bunch of AI to make this happen. It's not just one model. We use LLMs, we use intent based insights, we use sentiment, we use a whole bunch of models. But the LLMs have been great for us because you get context for the first time. So one of the things that we have customers who have really good relationships with their customers.

Oren Goldschmidt:

They're very comradely relationships. And one of the things we do is look for profanity in those calls. But they're peppered with profanity because they're just chatting and they've got these great relationships. So what you've got to do is look for context. It's like, okay, well, there are some times when you really don't want to use these words. Sometimes you can, it's fine. But the LLM can tell you in context what's going on. And it's amazing.

Oren Goldschmidt:

You can look for things like, hey, was there an opportunity to use empathy in this call and did the agent miss that opportunity? That's a pretty complex thing to ask, but you can do that. And that's really powerful where you can get context and really understand. In the context of this call, was this a problem or was this fine? Otherwise you're just red flags on everything and you want to avoid doing that.

Demetrios:

Which was my last question. How do you battle alert fatigue? Because I can imagine. I want to know everything about everything in my calls, but after a week of knowing everything about all of my calls across a thousand different agents, I can't handle it. And so I just tune out.

Demetrios:

Yeah. So I mean one thing is metrics, right? So if you can create a score and you can and over create a score that gives you a good proxy for how well an agent or a team is doing, then you can just look at the score. You don't have to look at individual calls. And then, yeah, in terms of alert fatigue, I think refining the insights and using those higher level context based models that we have to re understand, hey, in the context of the school, was this a problem or was this just not an issue in the context of what was going on? So yeah, we do those two things and that's been really good. And I think it comes down to, I think making this sort of program effective comes down to the issue of how deep your insights go, how contextual they are, how deep they are. And we had a client in healthcare and they were using us just to monitor their kind of scheduling calls and admin calls. And they said, hey, we want to try this on our provider calls where they're actually having the session with a patient. And we looked at them like, okay.

Oren Goldschmidt:

And what they're able to build was an actual clinical note taking system in Olay, which is not what it's designed for, but it worked, right? The models go deep enough that you can not just tell whether someone's using profanity on a call, but you can go all the way to, hey, translate this into a full clinical note and put it into a medical record system. So yeah, you can go really deep and that gives you information that is really actionable, really trainable. You can go and, you know, you can go and say, hey, I'm going to these five agents I'm going to talk to them today and I expect them to do better tomorrow at these things and see a change.

Demetrios:

You talked about how MOVA was being used in ways that you didn't realize. And as you said that, I was thinking, yeah, why doesn't every service provider that I interact with have something like mova? Because ideally it would mean that they're billing me less hours and so I'm saving money from them. I can imagine that if we can see where there are problems or if we can take notes on a call or we can have these different areas where you can plug in the benefits of mova, then it will hopefully save the client more money in the long run because the service provider isn't spending that time billing those hours to the client.

Oren Goldschmidt:

Yeah, yeah. I mean, I think things, yeah, note taking and summarization is huge we do that I think general quality of calls right Stop losing customers hemorrhaging customers because you give them terrible service and you don't even know you're doing it. You get your wires crossed you make mistakes. One of the things we also find a lot is people whose processes and procedures get in the way of customer experience so the agent is doing the absolute best they can they're great but there is some process that gets in their way and they just can't help.

Demetrios:

That's huge insight yeah wow.

Oren Goldschmidt:

That happens a lot particularly financial services you know but it happens a lot I'm sure it happens in telecom as well you get a conversation where the agent really wants to help that they just can't help because their systems or processes get in the way that happens a lot so yeah tons of places where you can improve the quality of the customer experience and improve the agent's quality of life as well.