December 3rd, 2018 / Published in: Interviews
Puneet Mehta is Founder / CEO of msg.ai, a VC-backed Artificial Intelligence company. msg.ai helps companies like WestJet, Target, Nestle and others automate high-quality resolutions to customer support issues across email, Web / mobile chat and messaging. msg.ai helps companies adopt AI into their workforce, working in tandem with human agents to provide superior support that meets customers increasingly high standards for quick, personal and accurate support.
Puneet is a former Wall Street algorithmic trading technology executive where he built predictive algorithmic platforms to power large-scale trading systems. These situation-aware automated trading platforms were amongst the first AI virtual assistants in any industry, providing traders with the most relevant insights, trends and analytics from an enormous amount of unstructured real-time data.
Advertising Age named Puneet to the Creativity 50 list. Business Insider magazine named Puneet to “The Silicon Alley 100: New York’s Coolest Tech People” and “35 Up-And-Coming Entrepreneurs You Need To Meet.” He was also featured in The Wall Street Journal’s startup documentary series.
Dulce: Hi everyone, welcome to another edition of DojoLive connecting tech experts like you, my name is Dulce and we are broadcasting from different parts of the world, I’m in Hermosillo, Tullio is with us from Los Angeles and Carlos if here from Mexico City, and this time we have a very special guest, this is Puneet Mehta he’s the founder and CEO of MSG.ai this is a pc back artificial intelligence company and we’re going to talk about AI and we’re gonna tell the topic for this DojoLive in a couple of minutes, but first of all we will like Puneet for him to introduce himself, talk about what he does, and his company, so welcome Puneet, thank you for joining us today, we’re very excited for you to be here so tell us about yourself.
Puneet: Thank you, a pleasure, I’m Puneet, I’m the founder and CEO of MSG.ai we are ventured back in San Francisco, and MSG.ai creates artificial intelligence software that is a work force multiplier for customer service, so the AI sits right alongside customer service agents teams and does a part of the work that only customer service agents would have to do, collaborating with them, solving some of the repeatable issues in helping them, having time to customers and having appeared customers.
Dulce: Awesome, so well today’s topic is cutting through the hype the real opportunity for AI to transfer customer support, why did you choose this topic?
Puneet: You know, if you think about the state of customer service today, we have all experience it as consumers, you know, it’s something that leaves a lot to be experienced it’s not really the most pleasant thing to pick up the phone and call a 1-800 number and stay on all four for a really long time, or you call and be bounced from agent to agent to agent, with the next agent forgetting or not really remembering what was the main reason that we even call in the first place, we’ve all experienced customer service where we felt, you know, can this to be better, we already ready driving cars but we still this level of customer service so it’s a problem that I’ve always wanted to solve and having build with the trading engineers on Wall Street and using AI for other purposes, it as, it was something that was really close to my heart something I really wanted to solve, you know, give consumers back their time agents, back their time and have AI be a work force multiplier, so that’s why, you know, I picked this topic which is how do we look at AI in a way that is not just a new shiny objetc but in real application of business technology that can give us happier customers but also significant business benefits.
Dulce: Great, thank you.
Tullio: Question Punnet, Puneet, sorry, I want to talk about AI today in your luggage [INAUDIBLE] but it sounds to me like yours is more around natural language processing, can you walk us through what’s unique about how are you using AI specifically our customer service because you have a lot of unstructured data right, so what can you walk us through how that works?
Puneet:Yeah, you know, here is solving this problem, you know, we are not just looking at it as one component as natural language processing, we are thinking about kind of finding the best way to help customer, so in some cases it might be natural language, in some cases it might be proactive and predictive resolution in other cases you know, you also have to, also has to be able to see differents images and visuals ins order to make or get to a resolution so it’s all these differents things exists within the AI domain that are stitched together in order to create that personal experience, you know, obviously natural language is a big part of it because as poor service as we know it today is normally based on a customer contacting an enterprise or a business and requesting help so that requires explicit expression of an intent a square natural language comes in but customer service ten years from now might not that might not be the same, you know, ten years from now customer service might be that AI predicts issues that might happen in the future and make sure they don’t even happen right, so that’s why the way we are looking at this customer service problem is a little different than just trying to create an NLP processor that answers question you’re looking at it more at a foundational level to say, you know, what could it take for a business to be more proactive to take care of the customers what would it take for them to predict certain things might happen in the future if the customers will be impacted and solve those even before the customers ask.
Tullio: Ok, so the end goal is to improve the customer experience I said, who are the targeted clients when you build this that you where thinking about, I know you came from banking, can you walk us through is this very furcal specific or agnostics, give us a little background and sort of, here’s what we think we can solve the problem.
Puneet: Yeah, I mean, we have customers or successful customers in a few different verticals, so retail e-commerce is a key one you know within travel we have Airlines, we have cruise companies and then we have Tech Calico utilities, so you know we have large telecom companies working with us large utility companies working with us, you know, we have also done some work in banking and insurance but those are obviously regulated industries so a lot of our customers come from retail e-commerce, travel, Tech Calico Utilities, and then obviously insurance and banking, those are the word those appear to us.
Tullio: So when does the software interact with the end user, let’s assume, I’m a bank and I’ve integrated this into what my call center I walk us through the bike.
Puneet: Yeah, exactly, so let’s say you’re an e-commerce company or a travel company or a bank when a customer messages you either on your website or on your app, or on a social channel, or sends you an email, those are the channel we address today, so we are focused on everything but the telephone right now, including you know, some of these smart home devices like Google Home, and Alexa being able to also make the AI brain work behind the scenes with them but most of our customers are using our technology on email chat or social.
Tullio: Ok, so I use a channel, I got an issue and it’s able to look up information and perhaps sometimes like you say make a prediction about why you might be calling, but how does it offer I mean, at what point I guess, let me get that right, is the end goal to limit the amount of human to human interaction or at least reduce that as best as possible and also improve the experience and how’s that working?
Puneet: Yeah, I mean anybody sim through the experience we asked consumers what where they looking for and they said they want immediate accurate respectful resolutions, to get companies the only limiting factor they have is the human bandwidth you know, to provide immediate accurate respectful resolutions and but they are that’s what we are trying you resolve, so part of the issues that a company gets contacted about our highly repeatable you know, they might also be low to medium business risk, and the medium exception management which mean everytime the resolution is the same kind of resolution, you know, these are problems that you could have hired really smart 15 year old interns to answer right, you might not have even needed so those are the kind of things that is helping with which are low to medium business risk, very high repeatability and low to medium accept management and then the remaining can go to human agent so that they can spend their time on really solving complex task that require human involvement rather than trying to solve beautiful things that don’t even require the minimum.
Tullio: That’s sound fantastic, so what are some of the things you’re measuring in terms of key performance indicator that are helping clients kind of gain me all right, are there any specific thing that you’re tracking and measuring an ongoing that say here’s what we’re doing, what are those things.
Puneet: Yeah, it’s interesting when we look at how do you measure success for an AI,if it is truly a work force multiplier, if it truly sits and collaborates with your team and that’s a part of the work, then you should measure success like you would measure success of a person right, so that’s what we do you know, AI is capable enough where you know, we have the confidence we go and ask each one of our customers to measure it really critically the way they would measure a human team, so it is measured based on the amount of work it has done which is how many issues did it resolve, it is measured based on the customer sentiment did they I make customers happy or did it make them actually more [INAUDIBLE] in our cases you know all customers have reported an increase in positive customer sentiment, and then sometime we look at other metrics like what did it do new customer experience overall, I did it reduce the average response time from eight hours to one hour, because what happen i if air starts to do a part of the work, you know, human agents can get back to everybody else also quicker right, so there are companies that you might contact and send an email to when you don’t get a response back for an entire day, we start getting response back for everything in less than a couple hours, AI would respond instantly but also human ages would have a shorter queue so those are some of the KPIs it’s you know, it’s a value in the form of customer sentiment and you have a loyalty but also the fact that you did not have to hire a lot of additional people in order to get the same amount of work done you know, it was a really long way.
Tullio: Well it’s really fascinating we are Nearsoft self-managed organizations so we do your peer review every six months, 360 degree peer reviews, I’m just imagining a bunch of AI robots having, giving each other review every six months the AI or how they perform, ok, so that’s important, I got a big question, it sounds to me like you’re disrupting the BPO space, is that what you consider you’re disrupting or I mean all this because that usually is stuff that was manually done right, but you’re tell me a little bit about sort of like what you’re noticing from your clients perspective that’s changing within their organization as a result of implementing your solution,within disrupting.
Puneet: I think they’re disrupting much more than the BPO component right, if you think about how customer service run today it’s mostly a silo within a large company or any size company you know, many times it’s an afterthought, first you build a business then you build customer service because your customers are about something’s not going right you know, most companies today don’t think that customer service is a channel where their customers are opening up with them, and they should have a conversation you know, that’s how they should understand which products to build next, that’s how that you understand where to double down and maximize so most of the large companies keep spending marketing dollars to understand their customers when and spamming them here in fact, whereas in case of a service the customers are willing to open up data so the way we look at this is you know, looking at it as an automation component is just low-hanging fruit that’s just the beginning, the real game hre is if you have a 24/7 open communication channel with your customers, what can you do with that you know, can you get real-time product feedback and you really understand them, can you increase loyalty, can you stop churn, you know, can you acquire customers by word-of-mouth in a better way most importantly you know, as more and more devices get connected there would also be an opportunity with to have your customers have a conversation just with the device that they have right, like one of our customers is one of the largest coffee machine companies in the world and when those coffee machines are cloud connected you know, a customer would not be calling a call center saying: Hey I see a machine lead will you help me?, and them the agent on the other side is not seeing the machine, I am seeing the machine and we are on a 30 minutes phone call that is really, you know, I’m annoyed because I have to be on the phone for thirty minutes to a customer service agent is like it’s taking me so long and the coffee machine companies like my goodness cost us so much, now when the coffee machine is cloud-connected I can have a conversation right with the machine right, that’s where AI being kind of that fabric in between that connecting medium really helps.
Tullio: Right, I mean, I’ve but I think with it’s unique what you’re doing right, because for the most part there’s two tracks right, you’re got work formation which is between machines avoiding problems right, there’s no human interaction there except for when something needs to be done and then you’ve got smart workforces which is where everything is going, but you’re kind of suddenly you’re married it to right, so if the machine is ready it can interact and then there’s a human component to this I assume at some point, how do you educate it to learn, do you have you built some that does that because so that like I said you know, if there’s a one off kind of thing how actually update it so it continuously learn.
Puneet: Yeah, that’s big right in you know, so we build this on a technology called deep reinforcement learning so the way we create this architecture is we’ve combined three main Ai techniques, supervised learning, unsupervised learning and reports we’re learning the supervises learning is mostly trained by instruction, unsupervised learning is mostly training by examples and reinforced for convenience [INAUDIBLE] is learning on the job, you know by experimenting different things by getting user feedback, by getting the agent in fact, the ongoing learning part is really significant like anytime any eye goes like just like a person like a human is performing on the job it is getting significant feedback, it is actually even devoted, so there are certain situation in which we might reward any eres could behavior and then it know as for more know that behavior in order to meet it’s quality goals or in order to meet that KPI.
Tullio: Interesting, hey Carlos let’s open this up, I have one more question after you open it up to the audience please, you’re a mute, hey we can’t hear you Carlos you’re on mute, but anyway if anybody wants to submit a question they can see the one, they can do so on Twitter on Dojo.
Carlos: Can you hear me now? I’m sorry, ok, cool I don’t know what happened you know, it’s technology sometimes unpredictable , ok well actually, yes, we first of all, two thing if anyone here in the audience has a question por Puneet you know how to use Twitter right, so this is the way to do it, go to Twitter this is our Twitter handle, go to DojoLive send your questions and then we’ll just ring him back to Puneet right, here on the spot and we do have a question actually and it’s coming in here from one of our guys from Misael Leon, and Misael is asking specifically about AI, there’s considering that AI is a relatively new thing in the tech spacem how do you go, how do you go about validating new features and functionalities in such a fast-paced market.
Puneet: You know, it’s all about the customer focus so the idea that you would only have AI do things that are required part of the business process and could you do it with the same level of quality that they where being manually done before you know, those are kind of the two ways of unpacking this problem, so if it is something that is not required to be helped would be your customers you know, don’t tempt that AI if there’s something you could have significant benefit and it’s low to medium businesses you know, that would be a perfect candidate to attempt to there, you could also look at AI as a method to make it more efficient for the human part of the team do their job quick, better, so if you think about it what is a human customer service engine do, they these three things, they talk you the customers and collect more information about the problem they different enterprise systems and they collect data from there they resolve the issue based on you know, whatever data they’ve collected from the customer and enterprise systems, so step one and two are things that AI can do really well, and it can be prepackaged that context and hand it to the human agent, so again, as you’re thinking about which problems you solve with Ai and in a fast moving technology component market, I would say things that are local medium businesses that they, I can solve completely and then where can I be the first line of defense and make it.
Carlos: Back to go, back to you, let’s go back to you Tullio.
Tullio: Coming other questions come up, I’m just fascinating because most companies are using it for inbound purposes rightm there they’re reactively responding to the query and such but you painted that your platform could be used, for album purposes like being able to actually develop in conversation with the end user to do all kinds of things, by the new products, new ideas and things of that nature, how’s that, how does it determine that, is there do people have to be hopped in how is that working.
Puneet: Yeah, I mean it’s it has to be done as naturally as it could get done if customer service was the central component today right, so if a customer is engaged and they’re already talking about a coffee machine that’s leaking, you know, you could potentially solve their issue and then ask them a couple of questions about in their daily coffee making experience right, and you know that you’re talking to a power, you know, that is somebody that appreciates and uses your product and that why on the phone with you, so it’s an automated filtering method where you known you are getting all of this like feedback from people who are only your customers and all the loyal to you, so we are, we are not thinking about this as a way to kind of reach out to ne people and ask them questions using AI, we are thinking about more like if you’re servicing customers anyways you know, would you attach certain components to it or could you identify them too if you have to feedback.
Tullio: Ok, fantastic, all right, so I’m very curious about how long is the company been around again?
Tullio: Ok, so you’re getting some traction, any key partnership that would be really valuable that could leverage this and embed this inside say some kind of OEM solution, are you thinking about that, what are your thoughts on that.
Puneet: Yeah, we get a few requests on that, you know, so customer service agent ask, call centers, BPOs, or as a matter of fact conversational interfaces would exist in different types ecology, even for answering employee questions right, within HR systems within IT service desks, all of those, so the partnerships we are exploring are essentially various companies that already are having conversations using humans on the net, and they want to automate a part of that, those are the kind of partnerships that be exploiting it.
Carlos: Tullio, we have a question from also from one of our teammates, this question is coming from actually, she’s in the people development team, from Anabel Montiel and she says, well she’s telling me here on Slack that even thought, that Puneet the question has not to do with AI directly, it’s more about because of her role at the company Nearsoft, she’s wandering and the following she’s asked: Can you tell us about your company culture, a little what are you guiding principles?
Puneet: The two or three main things that we look at one is obviously building a culture of intellectual curiosity you know, that is really key to us we’ll be building this technology for the first time takiing it to the world you know, there are certain things that our tech is doing for really large customers which you know, things that have never been done, and these large customers are able to utilize it in a way that is very high ROI so what that brings is that everybody in the company from you know, right from sales to customer success to the actual engineers and their scientist to the people doing sport, everybody has to have a certain amount of intellectual curiosity, because you know, they have to build things that have never been built before and they have to get other people comfortable with using them as well, you know, the second one is around transparency, so whether it’s around of people or for Ai you know, there’s a lot of out there that might look like a black box in our case we want to tell our customers out in air make a certain decision, what are the things that it has learned, what are the things that it can learn in the future, so whether it’s with building a technology company or building in, we believe in that component as well that, the transparency component it eky, and then you know, lastly I would say you know, there’s a certain amount of you know, from a certain point to be a certain mode of hustle required when it comes to early stage companies because it came you are, you know you’re building a strong foundation but at the same time, you know, you are significantly changing behavior right, you are going out there and asking large companies to try something they’ve never done before, so it does require you know disproportionate amount of effort, you kind of get that cannot ecology out there, so as far as our company is concerned you known, those are the three main things that pretty much everybody in the company would meet , you know, they would tell you and they pretty much live in velocity values.
Tullio: Sounds like you need a lot of grit, so I’m curious is this your first turn up venture.
Puneet: This is my second.
Tullio: So, you know, we have audience that’s always curious about making that transition from well established company to going out and actually building something, and we have an interesting audience next a lot of people or CTOs or heads of engineering and so on and so for both in startups and existing companies, those on the fence you know, thinking about going down the entrepreneur route what words of wisdom would you have to share, what’s your journey been like, in these two copies you work with.
Puneet: Yeah, you know, I would say, if you passionate about solving a specific problem you should just do it well, worth the experience that you would gain, you know, it’s there are thing that you would learn only by doing this, that probably is not available to experience anywhere else on the planet, in no school and no job in you know, no other situation, you know but do it for the right reasons you know, don’t try to do it or making a disproportionate amount of money or don’t try to do it for fame like none of that is guaranteed when you do a start up, you if you’re not passionate about the problem that we are solving and the value you create you know, then I would not recommend signing up for a journey like this, you’re really passionate about the problem you want to solve and what you want to do is you know, I guess see create value those are the right reasons to do this, making money end everything else is essentially a byproduct.
Tullio: Great advice, thank you so much, and as always when we’re starting to have a lot of fun times comes up so, I’ll pass back to Dulce you might have a question or two before we wrap it up, it’s been a pleasure speaking with you and I’m sure we’ll stay in touch very shortly.
Dulce: Before wrapping up we are three minutes from ending, this edition of DojoLive, but as a recruiter I’m always interested in knowing how do you find the best talent for your company?
Puneet: You know, its.
Tullio: I just wanted to, that’s a great question is like intellectual curiosity one of the sort of skill set I mean, I’m curious to hear the answer, how you go about doing that.
Puneet: You know it is a required skill set you know ,it’s interesting how we tested in different roles and functions you know, obviously you might test it slightly differently for engineer versus somebody who’s is sales, but is pretty much a requirement and hat we have noticed is even in the history of the company the people who have been really successful you know have had those three things you know, along with it will be like sometimes we do end up hiring people who are incredibly accomplished right, and who have done this a few times and have been very successful at doing it and don’t need to work for money but they do come work for us because of you know, they’re excited to bring me add to the word and they excited to do something significant and you know, that’s the fourth element I would add is the humility part that even though many of the people working for us have had fairly accomplished careers but they are very humble so how do we go about finding them you know, there is no one method I wish there was you know, a lot of them actually end up coming from networks or from the networks of the early employees you know, many times we also will look in other domains if is they might not have solved exactly the same problem, actually in engineering but they might have solved similar problems and watch them on deep learning you know, when you think about AI talents the word is a heavy shortage of deep learning or talent from the planet right,like you could they’re less than I would say, less than a thousand éolpe in deep learning who have made the largest contributions so with more talent pool you know, you’re essentially you have to bring in people and have to push them so that you know, they can start using the machinery in the different states and other early days of the industrial revolution, when nobody knew how to use machines but there where some machines and you pretty much had to bring in smart people and force them out of the way, on the deeper early set y ou know, that’s , that’s kind of what we are doing where we have hiring people from some of the best food on the planet and then we are poaching them on some of the AI methodologies.
Dulce: Thank you.
Carlos: Ok guys, listen, I know that we’re approaching our last minutes of today’s conversation but before we go, before we go we wrap up I’d like to mention something it’s just a quick announcement by the way, it’s about the next following interview we’re gonna be having here on DojoLive which is this week that’s going to be on Wednesday at the same time on Pacific and we’re gonna be having a conversation with Ravi Raj, the CEO and co-founder of a companny called Passage.ai which is also about artificial intelligence of course, and the topic is going to be, doesn’t get any cooler than this because it’s gonna be titled AI powered customer experiences, how leading brands are using AI middleware across voice and text channels so that’s gonna be interesting folks, remember that’s gonna be this Wednesday on DojoLive, right here Dojo.nearsoft.com and back to you Dulce.
Dulce: Thank you, yeah, well this was very interesting to know about AI in customer service, so thank you Puneet for your time for taking the time for joining us today, it was really fun, thank you Tullio, Thank you Carlos for being here today, we’ll see you on the next edition of DojoLive connecting tech experts like you and so thank you guys, and our audience for being here, so hope you have a great time with us and see you next time.
Carlos: See you next time.
Tullio: Thank you.
Puneet: Thank you.