Muddu Sudhakar, CEO, Aisera joins us to discuss the Aisera philosophy for AI mediated service provision, and some controversial recommendations.
Niche player in Gartner’s 2022 Magic Quadrant for enterprise conversational AI platforms, Aisera has been making moves to bring the best AI service experience to market.
We’re joined by Aisera CEO and investor, Muddu Sudhakar, to discuss the Aisera philosophy for AI mediated service provision and how it enables enterprises to get the best out of their human resources.
Muddu has such a refreshing take on how AI platforms should fit into business operations, and I liken Aisera to Bruce Lee’s analogy of water. “You put water into a cup, it becomes the cup. You put it into a bottle, it becomes the bottle”.
This is because Aisera’s platform is both a platform and a series of modular AI components that you can pick and choose which you need. For example, if you’re looking for an end-to-end chatbot platform with business process workflow automation, you can take the whole platform. However, if you just need an unsupervised NLU model, a knowledge graph, natural language search or any other component, you can take that component and integrate it into your stack. Therefore, Aisera is the water that can either fill the gaps in your enterprise AI tech stack, or fill the whole cup.
This conversation is full of lessons to be learned for businesses wanting to use AI services to automate customer experiences, including some controversial recommendations, such as:
- Don’t do a proof of concept: build a proper AI assistant, take a year to do it and then judge the value its bringing
- Don’t use experts or analysts: hire someone to work for your company who has the experience, skills and a vested interest in making it work.
All this and more in this great conversation.
In this episode
00:00 Introduction and VUX @ VOICE22. Find out more: https://www.voicesummit.ai/agenda
02:04 Welcome Muddu Sudhakar
09:55 What makes a good AI platform
16:06 Ways of organisations reaching requirements
19:25 Investing time in building
23:37 AI today
30:19 The aim growth of Aisera
40:05 Bring different assistants to Aisera
45:34 Robotic process automation
48:50 Examples of what Aisera is used for in companies
Check out www.aisera.com
And for more information on VUX@VOICE22, check out voicesummit.ai/agenda
Word for word
Yo, yo, yo, hello, hello. Hello ladies and gentlemen, boys and girls, welcome to vux world. I’m your host, as always Kane Simms and I’m gonna get a video done for that intro next time. It’s going to be nice and highlighted with lots of clips of our previous guests. So I’m looking forward to that. Good show and good tune. Happy Tuesday. And yeah, what before we kick off with an excellent conversation today with Muddu Sudhakar, who is the CEO of Aisera, one of Gartner’s Magic Quadrants, featuring conversational AI platforms and niche players in that. And yeah, we’re gonna get into the conversation with Muddu coming right up.
But before we do that, I want to tell you about the event that we have coming up in October in Arlington, Virginia, in the US of A, will be coming across the pond VUX world on tour, we’ll be doing a conference at the voice Summit, it’s gonna be VUX at voice 22, we’re going to be talking about the end to end automation journey for call centres. And so if you are interested in how to apply AI technology, in your call centre, conversational AI and natural language processing, then this day on the 11th of October is going to walk you through the step by step process of just how to do that it’s presented by Kore AI, and we’re gonna have a whole bunch of unbelievable speakers there. Lots of enterprises talking about case studies, lots of technology companies showing you the path forward, it’s going to be absolutely amazing. The agenda and all the details will be in the show notes. Please do check it out for VUX at voice 22. Looking forward to that.
And I’m also looking forward to this conversation today. With Muddu Sudhakar from aisera. Muddu is an absolute pleasure to have you on the show. Welcome my friend.
Muddu Sudhakar 1:40
Okay, thank you for having me. I like your show. You should call it AI world. User experience with AI. Excellent. I really like what you do. Thank you.
Thank you, thank you, I’ve been thinking about a rebrand because naming it VUX was a little bit short sighted. And it is a kind of start to call it Vux now because then it kind of passes off as its own sort of term, which doesn’t have to necessarily mean V U X. But yeah, I think something to do with AI CX actually predominantly, what we talk about these days is how the AI technology and natural language processing technology is enabling businesses to improve what they do operationally and provide better customer experiences. And sort of maybe a rebrand might be on the cards.
But one company that doesn’t have to rebrand because it’s making a hell of a lot of room in the market a name for itself is Aisera. You’ve been doing some very impressive stuff, you know, being on a mission as of late implement in this kind of stuff in the enterprise. So I’m excited to have you along and congrats on the good work you’ve been doing as well.
Muddu Sudhakar 2:48
Thank you. Okay, it’s actually a privilege to be part of your show. I always watch all the podcasts and what you’ve been doing to the community of AI contact centre, constantly. I think we need more of you to keep doing this. I love the shows, and keep doing that and educating the audience. Because AI is here to stay. It’s part of our world for the next 10-20-30 years.
Absolutely. Thank you. Yeah, I appreciate that. If only I could clone myself. That’s kind of what I’ve been trying to do. And I’m sure you can relate.
Yeah, that’s a good idea. We had Danny Tomsett from UneeQ digital humans yesterday. And we were talking about digital humans. So that’s maybe the next step is to create an AI avatar of myself, we’ll see. Who knows? Who knows. But no, definitely, definitely excited to get into the conversation.
So tell everybody who’s tuning in a little bit about yourself. Muddu, you’ve got a lot of experience in the startup world, you’ve got a lot of experience working with large organisations at a senior level in Splunk, Service Now, you know, you’ve invested in a bunch of companies as well, what led you to Aisera?
Muddu Sudhakar 3:51
Thank you for having me on the show. Look, it’s a journey, right. So as you said, I’ve been doing both working in large companies, startup companies, we started Aisera back in 2017. Me and my co-founder Christos, again, if you’re concerned when I was at ServiceNow, the when you’re talking about ServiceNow customers, you have a front row seat and what’s happening over this pre-pandemic.
Muddu Sudhakar 4:22
I mean, AI was there but it’s AI and machine learning was still nascent. It was going on I would say most of the AI things really do happen because of the cloud. Cloud is the reason how AI is happening right?
Muddu Sudhakar 4:32
So if you look at machine learning neural networks, are there back in the 80s 90s? What really changed the game is in much of the cloud with both Amazon and Microsoft leading the charge with the cloud. Without the cloud AI machine learning algorithms won’t be where we are today. And then come to 2017 as we meet with customers, but really being at ServiceNow you’re meeting coordinated contact centre people, right anybody who’s running your service desk, help desk customer support.
Muddu Sudhakar 4:32
And one thing that was obvious is, these people are really needing tools and solutions to automate mundane tasks. They don’t mind helping you if you have like a complex problem, like if you and me have some audio problems, setup issues, I have some basic things like my account is logged out, I can’t log in or I want to fix my network issues, Wi Fi issues, or it will be my customer, my subscription, I want to add my family member, I want my money back. I mean, most mundane tasks, I don’t think humans want to be involved in..
Muddu Sudhakar 5:30
Right after, well, in the first one month, two months, three months will be okay. Because as humans we are designed to grow and evolve. There’s a reason evolution in world history humans have evolved. So we can’t expect our other employees out of all the people who are in the contact center saying, Hey, you come to work, do the same mundane tasks, repeatedly like a robot for 40 times a day, or take 50 times a day. It’s just not fair.
Muddu Sudhakar 5:57
And that’s actually the reason how we started this whole journey and look, something has to change. And back then I talked to a gentleman by name Ram Shriram, as you know, Ram funded Google. And he sits on Google’s board and Ram had the same vision. And he said, we can’t do that, we have to take a chance in the world to make a bet on this type of technology. Suddenly it happens or may not happen. And you might have to write the technology code, and with that inception, we said look, is AI fully there, it may have its own technological problems.
Muddu Sudhakar 6:25
Back then when we started, there was no unsupervised. It was all guided flows. When people talk about conAI, even Gartner for them. The current AI is not AI. I mean, that’s why they call it corn platform. They call it NLP separately. as applied to me, you don’t want to differentiate like that. And the customers don’t care with NLU NLP AI machine learning is like, just something automated. Right?
Muddu Sudhakar 6:47
So this whole evolution of how we’re dissecting, etc, we thought, Look, we got to change everything. And we took a bold bet on how we can do a new AI solution that doesn’t require too much human intervention, it has to be simple to use, it has to be natural, but at the core, it has to have an unsupervised approach.
Muddu Sudhakar 7:06
It’s like thinking of your driverless cars, either from Google or Tesla. If you just keep on training the car, when you buy it, you’re not gonna use it too much. The car may learn from your habits that’s different that I call the behavioural that’s like personalization. Right? So there are aspects of where you need the training and learning on day one to use a car, I don’t need your data sets I don’t want to write it’s like when you buy an Apple phone, you don’t want Siri to say come and sit and train with me to use it to buy Apple phones, right?
Muddu Sudhakar 7:32
So I think this notion, some of these things, the industry still has to have to handle it. And particularly as we customers industry, but I think we are at a point where technology is already there, it’s going to get better. So that’s kind of the reason why we started. Can we do it? Can we change the lives of humans? Which is customer facing people and users? Can we change the life of the agents also, because they want to get out of poverty, like whether you like it or not, people in contact centre, they don’t get paid much. I mean, their lives matter, too, we need to move them up in the food chain. Right.
Muddu Sudhakar 8:06
So I think a whole lot went in our discussion process, how we want to do all this. And then I talked to Maynard Webb. Maynard, he sits on the Salesforce board. And he said the same thing. All these people, the industry executives like Ram Shriram Maynard Webb, Dan Owen, and Paul Martiz, Paul was my boss from VMware, all these years, look, let’s make a bet on this world, how we can change, right? Make this AI to be used by this world as best possible. And it’s not going to be easy, but it may take five to 10 years, right? So it’s along with me. I enjoy making bets. There’s no guarantee it will succeed. But at least we can control the destiny, if you all work hard together.
very good, very good. And I think the signs are showing that it’s happening, you know, the adoption of this technology is definitely growing. And there’s lots of competition, you know, I mean, the Gartner Magic Quadrant I referenced at the beginning there. You know, it was a lucky number of companies that were featured in it, not lucky because they’re all, you know, good in their own right. And that’s why they’re there, but for every company that was in there, lots of companies are not there.
There’s loads to unpack from what you’ve just said, the future of jobs, what are your mundane tasks? We’ll get to all that in a minute. But I wonder whether you can first, maybe share with the listeners, what you think makes for a good kind of platform, a good AI platform, because as I said, for every one that was included in that Magic Quadrant report, there’s probably 100 or 200 that weren’t. So what do you think differentiates a company like Aisera, from other technology platforms out there that are trying to do something similar?
Muddu Sudhakar 9:55
Yeah, absolutely. For me the way in which I see Gartner would obviously report I would say, they’re like other analysts form company share, they probably have some information. But again, the customers know more than the Gartner. Right? They have problems.
Muddu Sudhakar 10:12
First thing is you need to talk to the customers about their experience of the problem like you will probably know how to use a platform as you’re interviewing people, you know what your challenges are? To me. I always start with customers, the end users and the customers and their end users. What are their end users telling us? What are their problems? To me, that’s much more important than the new supporting organisation like Gartner, right?
Muddu Sudhakar 10:32
To me. They’re like, they’re just conveying certain things. But what really matters to me is customers and end users. For them. What they’re asking is, can you provide it in an engine, call it whatever an AI engine, or NLP engine NLU, I want a simple easy to use? If Kane asked me a question, can I understand Kane? What can you say? To the best possible, right in whatever language and that approach of understanding your thing? Requests? Right? That is the most important one I call the AI. it’s almost like the AI SQL layer. Right? That’s something we’ve been pushing out. Right? Remember how Oracle came into SQL they’re like, as equally like understanding what you do. Understanding that query request itself is an important one.
Muddu Sudhakar 11:14
The second thing, once you understand it, is how to serve that request, right? That’s where unsupervised natural language understanding or AI is important. The early days of most customers are all about supervision, right? Even though 99% of all vendors that I know, they do a supervised approach, like if you asked me this question, I’ll do this. If you know exactly what Kane is going to ask. And if you write all the steps of it, you’re probably a God, you won’t know all the combinations. There’s a reason why Google invented search, right?
Muddu Sudhakar 11:41
So this notion that I will create a dialogue flow, or I’d use a knowledge flow or a conversation flow and call it a convAI platform. That’s not a I don’t think people are getting that. So a lot of vendors may put it there. I tell customers, beware, I mean, this is where you have to be aware of as a buyer, what you’re looking for.
Muddu Sudhakar 12:00
Next is, do you have language models for multiple languages? Right. And you and I talked last time and we are doing your show, right? I need a language model for how English is different from Japanese and Chinese. Right? You need a specific language model, then depending on the request that you ask, your request may be fully loaded, it was even in English, you may ask me a question that has multiple meanings, multiple things. So how do I pass that out? And that’s where the AI, sequel layer kind of concept comes in. Right?
Muddu Sudhakar 12:34
And then, given all these things, you understand the domain that we’re asking, right? Is your request related to a casual request? Are you in the healthcare space or insurance space? Right? And you don’t want one bot for each domain? Like there are people who say, I’ll do a good job of it, I’ll do a good job for the contact centre. Or I’ll do something for the airline industry. Or I’ll do something for insurance like you’re going to, you’re not going to wake up and say I’ll use a different Apple phone, different Siri, will have one Siri to ask the question whether you like it or not, Siri will answer at that point, you have to live with it. Right?
Muddu Sudhakar 13:08
So this concept of the universal bot a bot, people are not still adopted, right? So to me, I look at a whole lot of things like this as a fundamental need in which you want to. And finally, you want to wrap this around with a simple concept of a lot of requests that you may not have an intent on, right? Like when you add something even adding meaning into it, I call it intent less. Well, that’s where you need to have things like cognitive search, neural search, right?
Muddu Sudhakar 13:34
So this is behind abandoning some of those things as well. So the concept of search, knowledge, graph, conversation, AI, right, all these things has to marry with intent, intent less. So the technology is changing so much. My only request to customers is first figure out what is the problem you want to solve? Who are your end users? What are your requirements or create some kind of a table and create a success material and then look for it? Because just looking at some Gartner report that won’t solve the problem.
Interesting, it’s really interesting. How, for a company who kind of like, isn’t, I suppose experienced in this field, who has never used conversational AI before but knows there’s a need for it. Part of the issue I’ve seen is that they don’t really know how to define their requirements, because they don’t really know what they need yet. And when they first start, whatever it is that they go with first, and I’m sure you’ve seen this happen as well. I would wager that you’ve gone into client organisations where Aisera is replacing an existing technology. And I would wager that the existing technology was put in place because the client didn’t really know what they needed. And they’ve just kind of gone with something. And they got stuck. It didn’t have what they needed in the end because as soon as they learn and develop,they back themselves into a corner sometimes.
So How would an organisation who is thinking about conversational AI think they want something for customer support or something like that? Or they don’t know anything about the platform? What are some of the key ingredients that they would need? You mentioned a couple of things there around, you know, not having a domain specific NLU making sure the platform is unsupervised uses unsupervised learning, things like that. But again, these are concepts that you need to know what NLU is, you need to know what unsupervised learning is, let’s say that I don’t know that stuff. And I just know that I think there’s a value in conversational AI. How does an organisation go about putting together its requirements based on your experience of working with clients?
Muddu Sudhakar 15:39
Yeah, no, I think it’s a very good question. I think there are two parts to this question. The main thing which I said to Kane first is, look, if you don’t know what you don’t know, you don’t know. Right? So obviously, you need to hire a person, right.
Muddu Sudhakar 15:50
So this is an area where if you’re working with, let’s say, a C suite person, whether it’s the CIO, or VP of Customer Service, or Chief Customer Officer, Chief Revenue Officer, whoever the person was chief contact centre officer, he or she, what they have to do is first hire the person who you trust, right. And that person had some domain expertise, and give them the full batton to do it.
Muddu Sudhakar 16:11
What the C level people should not do is, hey, I’m going to do a simple litmus test, I want to do a three month project and a one month project, but isn’t, this is technology you need to go all in? So all in a sense, you want to do at least one year, right? This is what I would like to see the C level people do, they don’t do it that well, in many areas they are not making enough bets. Not all bets are going to happen. I don’t mind if they select me or anybody’s similar to us, other vendors. But make one bet with somebody and go for one year.
Muddu Sudhakar 16:43
At a point, you’re not going to learn anything from a POC or proof of value, a proof of concept. Anybody who’s trying to show you a case around this is misleading. So I encourage or discourage people from doing proof of concept. First find an expert to run it, have a sponsor, buyer education. Maker, empower that person individually, then make a bet in some manner. Pick a vendor like they should watch this show what Kane is recommending, pick the top three vendors there and pick one of them. Like To your point, that criteria should be, I would say, Kane’s VUX World as a criteria. For example, listen to you. Listen to, take your requirements, I’m serious, if they let’s listen to your podcast, that’s good enough for them at this point that exists, and then make a bet.
Muddu Sudhakar 17:26
But the only thing is that the bet cannot be short. Some people are making bets. But then some are, are this illusion that a POC or a three month project? Six months? No, you will not, go on it for a year. a year is a good enough time. If you still don’t see the value, that might change the vendor. Right? It’s like we call it and when you buy a house, it’s location, location, location for us its ai, ai ai. That’s it. I think that should be your tagline under AI, three AI starts. If this AI doesn’t work, go to the second AI, do not, go to the third AI. Hopefully, by three times you found the right thing. If not, three strikes are out, maybe then you should find another job. But each year is the best to get this one year.
Yeah, interesting. That’s a really refreshing approach to that. Because of the amount of consultancies and all that kind of stuff that sell proof of concepts as a way of trying to get the foot in the door. And the amount of vendors that do this as a way of trying to land and expand gets their foot in the door. It’s rife, you know, which is, it’s really interesting to hear you say that that’s not a good way to go about it. Because the reality is that in a proof of concept, you don’t generate enough data to build a good enough model anyway. You don’t really refine the interactions to a level where they’re really good enough to be put in front of customers at scale. And so there’s no real other way to really find out other than really invested in building something properly. And that takes time.
Muddu Sudhakar 18:58
That’s the end of it. There’s one more to add, if I’m may Kane, I see the remedies. Again, I’m giving you the truth, honestly, not nothing to help Aisera. I’m not, but I see this. I mean, I’ve been doing this for more than 35 years and I see this mistake happen industry after industry. I tried to do the same thing at cybersecurity. We did the same thing in the past for the analytics industry. Right.
Muddu Sudhakar 19:19
So one other thing is don’t hire a consultant or don’t talk to an expert, like asking somebody for advice who’s not fully vested in your company is no good. Like, you get your for whatever is advice. So calling an analyst is calling somebody else. I mean, they’re not going to, they’re not in the project. Trust yourself. The best way you can if I were to give advice, I’m not authorised to do it. I can’t do it. I don’t know why, but most people won’t do that. Everybody’s ready to offer advice.
Muddu Sudhakar 19:44
So hiring 10 consultants or even one consultant or analysts is a wrong approach to this work. They’re not billing, they’re not operators. reading an article, reading on the side. So many CFOs made the mistake of letting go talk to some XYZ analysts or some expert, sorry That’s not the way to do it. That’s right. And then maybe use that to say, look, this report says that the members think that by doing that they’re doing the right thing for the company, they’re trying to not invest enough, because nobody wants to make a bet that costs their money, somebody or their job.
Muddu Sudhakar 20:15
But unfortunately, good technologies will need to make bets, like regarding whether Elon Musk or Jeff Bezos, they make bets, some bets will happen, some won’t happen. If you’re in the C suite, you are required to make bets, you make 10 bets out of which only one will survive. But if you cannot do that, you shouldn’t be in that role. You can’t say I want to play for safety, if you want to go to safety, give the CIO job to somebody else, or give your CFO job to somebody else. Right.
Muddu Sudhakar 20:41
I mean, you’re required to help your company to make bets. And you’re educated bets that you got to make, right, obviously, you want to do your analysis with your management team. But my advice to them is to hire people who you trust, and have them make some bets, or the bet should be long enough. But it’s like if Elon Musk comes and says I want to do SpaceX in three months, because the algorithm should take 10 years, if at the end, it may not come. Right. And that’s a beauty in which I see how companies like Microsoft and a great job looking Microsoft, it took a while for them to get Teams working. But now they’re rock and rolling, look at Zoom, they have done that better. Right.
Muddu Sudhakar 21:20
So all these companies are making bets in a space, whether like Zoom is getting into contact centre, Microsoft is getting the contact centre and they’re gonna change the life of contact centre, like you, in this video talking about so existing contact centre guys, if I’m there, right, I’ll be worried about them getting Microsoft and zoom getting the good experience. But that’s they’re not making bets for three months, six months, they’re going to go all in.
Interesting, it is true. You know, Microsoft, Microsoft don’t do things on a whim, do they and Zoom communication is all they do, essentially. And so it’s kind of really interesting in how this area would be a real kind of target for them. And if I look at Microsoft’s announcement, obviously, it’s been released in kind of like general availability now, but when it was first announced in the press release a couple of months back, if you look at the features on that list of things that this contact centre is going to do, because I counted something like nine out of 14 features had AI in it.
And so if you look at like three years ago, AI and chatbots, and voice systems, and all this kind of stuff was kinda like a layer that was wrapped over the top of contact centre, you know, maybe stick a chatbot against the live chat, and you’d kind of slowly tweak it and like, you know, patch AI onto the call centre. Whereas now, the call centres today, Google’s integration with ujet, and partnership with ujet, zoom, and, you know, Microsoft, these are contact centre capabilities that are AI from the absolute core.
And so I wonder whether you can shed a little bit of light on what you think in terms of enterprise technology architecture? Which do you see AI now still as being a layer that’s kind of over the top? Or do you see that, you know, in order for businesses to get real value, this technology needs to become so fundamental to the business that it is a real core technology?
Muddu Sudhakar 23:16
Absolutely. No doubt, I think there’s a famous world where Software is eating the world. AI is eating the world, if any CEO, any C level, any board, if they’re not in AI, we are wrong, the wrong thing. It’s going to be here for us. So it has to be fundamental to your business, whether you are in the meatpacking industry, whether an agriculture industry be it in the cosmetic industry, whether in high tech insurance, they’re using AI is a technology fully there today in many areas, the maybe not it won’t be there, right.
Muddu Sudhakar 23:45
But if you’re right away, it’s the same as how computing has changed for 40 years, AI is going to change. I don’t know how many years it’s going to be a new era, maybe its a boxing game, one of 10, boxing round matches. Right. But it’s good to have 10 more rounds to go or 9 more rounds to go. And each round may take five to 10 years. So it’s very early on.
Muddu Sudhakar 24:05
So if you’re doing software architecture, I always say, let’s create an architecture to your point, I think you asked the question, but most people in this scenario get really excited, you got to create an architecture that runs on any cloud. First of all, you do not want to run these services on your infrastructure on a local service. One thing that I wish when we talk about it, you don’t want to run a contact centre run on prem. Guess how many contracts are running on prem today? Right? Which person, any person, why would you run your current centre on prem today? I mean, you’re in 2022, right. I mean, that’s something like if I’m a contact centre leader, I would just change that, right to that point, I would run that in the cloud to some right.
Muddu Sudhakar 24:46
Then the question is, can you make that into the Kubernetes layer, I want that to be flexible. I want to be able to move things around. I want that resiliency. I want this to run in virtual cloud infrastructure. Then I want to make sure that my database to store the data is a polyglot Database. What do you mean by polyglot? means I don’t want to turn to one type of data? Like for storing voice requests, I may need a different database to throw chat requests, I might need something else. If I need the requests not coming from humans, what if tomorrow, you have to think the request would come from machines, right? If I’m running, if the machine is down, the mission will be sending me an alert. It’s not a chat, it’s time to chat with us.
Muddu Sudhakar 25:21
Muddu Sudhakar 25:53
So this platform should not be one dimensional today. And so the world will change, you have to create a new architecture, every business has to go look at incentives. What can I do to offer my business services? Where can I leverage AI? Is it for my end users or my employees? Is it used to grow my business? Kill my business? Cut my cost down? Right? Is it to engage my users more and my customers more, but is a fundamental aspect, it’s not on the top? Right.
Muddu Sudhakar 26:19
To your point, if Microsoft already put nine out of 14 features into AI, there’s a reason and Microsoft is 2 trillion $3 trillion company. They’re almost like a nation state. They’re probably bigger than the top 10 countries in the world right now. With GDP, right? So from that perspective, Microsoft is putting AI in just a contact centre, that gives you a sense as to where the world is going. Right.
Muddu Sudhakar 26:41
The other thing that you add, like, if you remember, on just a side note to the previous question. Look at Apple again, Siri, how many years did Siri take to get it? Correct, right? I mean, I’ve been using Siri, from day one, it took five years, six years. That’s like, they didn’t answer everything perfectly. And they will maybe maybe not. Right? But is it getting better? Absolutely.
Muddu Sudhakar 27:01
So this goes back to the earlier question, do you like these technologies to be fundamental and you have to go all in, and we go all in, you don’t stop the project and start the project. If some vendor doesn’t work out, find some other vendor, or maybe have multiple vendors to figure out do AB testing, do co-process. But you can go all in, right?
Muddu Sudhakar 27:21
Look at what Apple does, they build their own shops, they buy the chips from Qualcomm, they buy the chips from Nvidia, right. But he is building his own, he got there. So some customers say, hey, sure, I want to build my own. Beautiful, take a platform and build your own. But at the same time, also leverage what you got from the vendors out there, compare your notes. You don’t want to wake up, it’s not like a hobby and say, I want to wake up, I want to do that hobby. And that’s where today some other customers and thinking is it I wonder to three month project, I want to do like a hobby, I want to test, what No, don’t put your fingers in the water, let’s jump in the pool try to learn to swimming, swim, or we want to know when somebody will help you. Nobody let you try this.
Nice, nice, that’s a really good analogy that go to the fingers and just jump in. So you touched on something there, which is really interesting around the architecture side of things. So a lot of businesses these days, attempting to get themselves a more flexible architecture. And a lot of projects that I have, I’ve worked on some and I’ve also heard about a lot, which is that a company will begin their kind of AI journey, and then hit a roadblock. And that roadblock will happen because there isn’t sufficient integration into the business systems, there isn’t sufficient access to data, and all of those kinds of things. And so a project needs to be paused in order for that work to happen.
And many businesses are trying to get themselves into a situation where they have that flexible or more agile kind of architecture. And at the same time, what that looks like often is companies that have that kind of modular approach. And you mentioned in there that you don’t want to be trapped into one programming language or trapped into one database structure, you want to be able to have flexibility and have a modular approach. So you see companies like BBC, for example, where their conversational AI capabilities are drawing in individual capability, capabilities and components from all sorts of places. So they might use Microsoft NLU, for some reason they might actually bring in Google NLU, at some point, they might use different types of speech recognition for different devices, and surfaces.
And so there’s lots of these components that can be brought together to assemble a conversation with AI and AI architecture. Or you can opt for a full suite like aisera and other vendors and get the whole kind of package.
So I suppose from your perspective, you must be always thinking about this whole concept of vendor lock in, because what you don’t want to get into is a situation where somebody reaches the limits of Aisera and then begins to consider a different technology. You want to be in a position where you’re growing with the company as a partner, so I’m wondering if you can share Some light and how you’re thinking about the scalability, the extensibility, and the growth of Aisera, in line with this changing company technology philosophy of having modular component technology.
Muddu Sudhakar 30:13
Very good question. Kane I love it. So first, there are two parts of your questions, I’m going to pass it and try to answer. First is what a call is. If you want to use Aisera, you can actually rent components and offer a platform. So you can take my entire solution, if you want to do it. For those people who do not want to build anything or nicer, you can do it, that sort of thing. Or you can take our platform and build on top of our platform or next to my platform, right, if you don’t want to build, right. Third thing is you may want to take only certain competence from it. Like, for example, you may want only a knowledge graph from me, you may only want my neural search, right, we will expose everything as an API. We’ve already done that, right. So many of my customers actually think that you may not want to use it for my entire enchilada, you may want to buy only chips from me, or you may want to buy salsa from me, I’ll rent you whatever you want. Again, I mean, I’m here to help you. I’m here to work with you to decide what the component that you need to bring your journey at this point in time. Because if you have something working there, and you will only need that we will give you that.
Muddu Sudhakar 31:16
So that has been my philosophy from day one for everything I do. Or last 20, 30 years I have been. So I give you a platform, I give you individual components, but also give you a fully packaged solution if you choose to take it from us. So you have a spectrum of solutions. It’s not one not either, and that’s where I like to sit. And when you take this spectrum, it’s to your point that it should be broken over individual components. And those will be you can look at conversational AI as a wholy one thing or you might say I only want a conversation platform with only knowledge graphs, flows, I do not want any NLP NLP or I want in 10 phrases NLU understanding, I want to read only NLU from you. Or me, I want your unsupervised machine for my knowledge articles.
Muddu Sudhakar 31:58
We also give a conversation automation to your point, people talk about integration, etc. One thing that we did, they are very unique, right is, is don’t need really integrations, a reason to do integration is if you can’t do right your platform have a very good what you call a automation workflow platform, something coming from ServiceNow we did is we created what we call it conversation automation. Like you have a conversation AI Content Automation is to automate your workflows with a conversation engine, because there are many automation players too. But to automate those things, I don’t want you to write a connector to it. So I shouldn’t be able to write any software, because Software is eating the world. As long as you can write in the software, I can talk to you, assuming you have some kind of interfaces, some system interface API’s web hooks.
Muddu Sudhakar 32:39
So that’s the beauty of what we have created, you and us can talk to each other. Even though your inference and my inference may not match, I can get the AI layer or the software layer to do that. So all this competence, we’re providing some people to do it, I call them build, you can build your own solution using Aisera. Right?
Muddu Sudhakar 32:57
Maybe we should call it in the show called build AI or call it AI to build let’s come bring your own conversational AI a BYOD like BYOAI or something. Something that’s a concept of a thing, bring your own AI to us, build it on top of us next to us. That’s absolutely possible.
Muddu Sudhakar 33:17
What ideally I also tell people is don’t start from scratch. Like it’s like if I have to build my own cloud and I don’t want to go run from Amazon Azure or Google that I’m not being prudent I’m not being a citizen of my company. That’s something that you don’t want companies to do today. There’s enough AI in the world don’t start from scratch, you start from the base substrate with the substrate you want to use. It is up to you to use Kane substrate; you may use some other substrate but start with some substrate. And then build what are the unique competence that you want to do and by other competence also from other people.
Muddu Sudhakar 33:54
Look at how Nvidia started. It was like Nvidia was sitting next to Intel as a core processor in the graphics world. Today, they give you both GPU CPU and AI processes too but you don’t have to start from day one saying I want to build my own process. You don’t wake up and do that. I think that’s something you and your audience can help your audiences to realise that look, don’t start from the start from some substrate.
Muddu Sudhakar 34:16
Yeah, then on top of substrate next assumption, rent competence and build next to whatever you want to make a solution that fits your need. So then, when you buy a shirt or a good one, you want to wear something that fits your needs. You want to customise it to your audience, to users, and that’s where your customers will come in. But you don’t wake up and say I want to plant a cotton seed, I want to grow cotton, I want to fabric my arm and then you’re done and you’re not in a good season for your company.
There’s a really good model, which is exactly what you’ve just explained. I’m sure you’ve seen it, which is the grid, the different types of technology and how you should buy or build. So you’ve got the kind of first stage which is the commodity Use, which is what you’re getting at, which is, you know, databases, cloud storage, hosting, its all commoditized, you know, then you’ve got the kind of specific functionalities that have already been built. But you don’t need to rebuild, like a dialogue manager, or you know, from hearing you speak, your kind of like Knowledge Graph and stuff like that seems to fall into that category where it’s already been built, you can just use out the box. And then you’ve got these kinds of like, very specific company or industry specific things that no one’s providing for, which you would need to build. So for example, if you have a very particular line of business system, and there’s no integrations to it, you can’t just go and buy off the shelf, a piece of middleware that will just do it, that will be what you’d need to build. But the components that you could just buy off the shelf and combine that with commoditized, kind of like, you know, readily available services. And through that you can kind of stitch together your architecture, it sounds as though that’s kind of what you’re describing there.
Muddu Sudhakar 35:53
Yeah, absolutely. I mean, you can go build it. Now, let’s compare both of them build versus buy, always, when you try to build, you can’t build this with four or five people or 10 people. And that’s another mistake. If you really want to build, let’s make sure that you have enough money to build today. And also to maintain, you can’t build with X number of people, and then I’ll put it into a cold ice to continue, it’s not gonna happen. If this is a request, finish the software, you have to change yourself to requirements will change constantly.
Muddu Sudhakar 36:20
So the most people if you try to build, build versus buy still 10 to 15x difference. So if you really want to build, let’s make sure that you’re building it and you have a good business case and ROI, and you have enough dollars to build. Right?
Muddu Sudhakar 36:34
So an average build today will still cost five to $10 million, or even $20 million. If you do not have that much expense, don’t bother building it. Right? You may have a need to customise it, but do you have the budget to it? Do you really need to add my customers to it? If not, take some of the shelf components. Right?
Muddu Sudhakar 36:52
The same reason today, there’s a reason why the hyperscale vendors buy switches from Cisco and juniper directly. If they think the Cisco switches are good enough they’ll buy if they really need to customise for certain workloads, you may do it. So you have to keep your options open. Do I really need to buy today for an extra three years? Or do I need to build? And what if I build it? To your point? How am I going to build for the next two, three years, we have a budget for this year, and the next four years to come. But that budget is not going to be buy versus build is usually at least 10 to 20 X effort minimum, not care because by the time you have 10 employees, right? In one employee cost can pay for most of the solutions out there.
Muddu Sudhakar 37:34
That is nothing right to build a good solution left 10-20-30 people in it all. Like to your point, I need to understand the dialogue manager, who understands what a dialogue manager is right, who understands what’s a workflow, like it’s not easy to make those solutions. Right moment you start hiring those people. That’s an expensive solution. But really in this economy now that people are so tight. Look at the look at let’s realise a market that we are in the whole global economy is under turmoil. In this economy, usually build is the wrong source choice is not. Right.
Muddu Sudhakar 38:06
You build when you have a lot of resources, you have a lot of empty resources and you want to do something and money. Right now, pretty much every CIO I talk to, there’s no building. If somebody’s sitting there doing any special product, they’re all being canned, markets down 70% , are we entering a recession? Nobody knows.
Muddu Sudhakar 38:24
Is financial markets going to record to the same level nobody knows, is a SaaS multiples come back to the same level most people think it won’t come back. So the party’s over. So all these private equity companies are large vendors who put like 20 $30 billion, their funds are down they lost 40 billion already 30 billion, right? That part is over at this point.
Muddu Sudhakar 38:45
So it’s like the recession that happened in 2008, or 2001. So at this point, if you really want a good solution to your companies, I always say don’t be right. These are all going to build. You have to work but not really. Unless you have that much.
Yeah, there was no need to, you know, your companies, like I said, so we’ve built a lot already. And it’s all kind of good to go. You mentioned something really interesting there, which is BYBY, or BYO or AI is how he was. Yeah. Yeah, bring your own AI.
That’s a really interesting concept, because it kind of ties into what we’ve seen to be the next phase of conversational AI in the enterprise. I know. Companies like speakeasy have been doing this for some time. And our Cubus Greyling has been writing about this a little bit. We’ve had a lot of conversations about it, which is some of these enterprises who you’ll be working with.
As I mentioned, you’ve probably been to a customer who’s got a current platform. It’s maybes have reached the limit of it and you’re kind of being the one using Aisera right in its place. Or you’ve probably also got clients who already have a chatbot or something like that. And they want to use like a team’s bot or something like that. And so you may well have been brought in alongside other technologies within an organisation. And lots of organisations are in that position where they’ve got investments in different technologies, a chat bot, here a voice assistant there. And, you know, they don’t really want to scrap what they’ve done. But they do need to kind of level up and bring in some additional capabilities. And so I read that Sira has an orchestration layer. And the way I read it makes me think of similar things similar to speakeasy or like a one reach where you have the ability to bring assistants and AI assistants from different vendors and different places, and have AIC or almost front end those things. I’m understanding that right, if you can explain orchestration,
Muddu Sudhakar 40:50
No, it’s much beyond that. But good question. Very good question. So the way in which I see the world, I’ll describe, you tend to look first of all, to me when I remember I talked about you know, universal bot, and we do it by domain. Let’s say I do sales finance marketing. Tomorrow if you have to work at an earlier point of build. If you have a better bot in a certain domain, we don’t mind integrating with your bot because remember, I am the universal orchestrator. I am the endorser.
Muddu Sudhakar 41:13
But from that perspective, we can talk to any third party, but that’s no issue at all for us. And so that’s typically when people talk about integrating in the continuity, but when we talk about automation orchestration, I’m not talking about that.
Muddu Sudhakar 41:27
When I talk about automation, and orchestration, that’s I use conversational automation, I actually want to be the workflow engine orchestrator to take actions. So when you call and say, My account is locked, or I have to unlock the account and send you the new link to your password. Suddenly, when you say my, I want to be referred, I have to genuinely see, do you really need to be referred, do all the analysis and actually give you the money back?
Muddu Sudhakar 41:51
So I’m actually doing the performance, I will talk to your stripe, paper, look at your account, send the money back. So we’re actually doing the whole orchestration ourselves in the automation world. Right? Whether that is insurance payment, financial industry, we have FinTech companies saying, hey, I want to orchestrate moving my money from money transfer from UK to India, UK to China, UK to Australia, right, we will be able to do that. Right? So that’s what I mean by the orchestration.
Muddu Sudhakar 42:22
Today most call centres are still knowledgeable about serving FAQs, that’s always important. But the world that I see it is the most average request, it’s actually the other way around 70% of requests or requests, you have to do something telling you saying here’s an article go read versus doing the job for you is where the AI should come in. So AI should not go to AI plus automation. Right. So in a way, we should say, Bring Your Own ai plus automation, maybe we call it BYOAI-A, or something that sounds like that or something.
Nice. So you’re talking really about customer journey orchestration. And the platform is one its front end. But it’s also providing the cogs and engine inside to actually enact a process and deliver the actual end result of the service on on the backend.
Muddu Sudhakar 43:15
And the reason I do this is, at the end of the day, every company should focus on your business, not worry about I want to build it, I think that should come first you should do is when you wake up, what can I do to grow my business? Scale my business? How can I eliminate unnecessary costs? How can I improve the experience for my users? How can I get my users and employees productive? If you focus on those, then you will decide what you want. That’s to me, as a business buyer. Always look at the business requirements, the problem requirements and see what you are, if your business requirements are called request only knowledge or we just want to do that.
Muddu Sudhakar 43:50
But if your business requires you to do both AI, and automation, absolutely. Right. And if your requirement requires multiple domains, go for multiple domains, or multiple languages. Right? You don’t have to start everything and start small. Whenever you make a bet, you may want to start. I want to only do English, in one domain, the domain could be sales, I want to do best sales, conversational bot, or I want to do well in marketing, pick one domain, or write one language. And only if you think that it’s only knowledge serving just answered FAQs, and auditory or do the automation piece. Right, start I’m getting back to your point, I’m giving you the multiple touch points on how you want to engage with me twice over the years, but start one area grow. And along the way, if you want to build Absolutely, you’re the platform and API to build on top of us.
Nice. I touched on earlier. We both touched on earlier on how some companies don’t have the infrastructure to be able to enable some of these automation processes, don’t have systems with API access and, and those kinds of things maybe don’t have data in a good shape and stuff like that. So a lot of this stuff kind of needs to be in place to enable some of this automation. However, there are other ways of doing it with things like RPA and stuff like that, I wouldn’t know where you stand on robotic process automation and whether that’s something that you can consider.
Muddu Sudhakar 45:14
We actually check on automation. RPA is part of that, to me. RPA is a small piece of the overall automation, right? robotic process automation fits into an overall automation bucket. So the way in which I see the world is that automation is a much bigger umbrella underneath, I could do RPA, I could run books, I could do orchestration engine. So to me, robotic process, screen scraping, running process automation, all of that comes in your overall automation key.
Muddu Sudhakar 45:41
So today, we will do our automation or tomorrow, you want to use my partners like UiPath, automation anywhere happy to integrate with them or even ServiceNow. Right. And again, when you integrate with these when this thing, right, you already have an existing process, etc use them. You don’t have to elaborate, you don’t have to modify them, you don’t have to throw it, keep whatever that’s working. Use Isilon next to any of those. That’s how I see the word.
Nice, that’s a very mature way of looking at it. And setting up in a way that is, I suppose, it’s a non threatening and complimentary way of approaching things, which is that if an organisation has nothing, then they can take the whole thing. If an organisation has lots of stuff going on, but they have a very specific problem, then you can almost fit in a little bit like.
It’s almost like to use the kind of Bruce Lee analogy of water, water takes the shape of the glass and water kind of finds a way through the cracks to fill the gaps. And so it almost seems as though that’s the kind of strategy that if you look at a business from front to back, there are gaps in here where sophisticated organisations need to fill. And therefore Aisera sounds as from what you’re describing in your go to market is the water that can come in and fit wherever it needs to fit in order to enable a process, sitting alongside RPA, and all that kind of stuff. UiPath or where it needs to. But at the same time, if you want to fill the whole glass, you can and you can take the whole platform. And that’s that kind of thing, a very sophisticated way of approaching it.
Muddu Sudhakar 47:16
Beautiful, that’s exactly why I wish I use that analogy of water. But I used to be like an amoeba, I mean amenable to whatever shape and form you want me to take. I’m the single sample amoeba, I’m an entrepreneur, I’m a startup company, my job is to cater to customer needs. So we give you optionality in every shape and form, whatever you don’t want, we want to remove that. But I My goal is to as I said, water knowledge is the best way to describe it. So beautiful.
Nice, nice. And that’s really good that the two final things that I was wanting to kind of touch on, as I’ve seen on the website, you’ve got a number of good customers, high profile customers, you mentioned, Zoom already zooms down on the website has been a customer, McAfee very large company, you know, as well, I don’t know what the situation is with 8×8 as well, actually, which I found surprising.
I wonder if you could shed some light on the reason I was surprised because I thought eight by eight would be investing in their old tech in their own technology like Genesys and five9 and stuff like that. But I just wonder whether you might be able to shed any examples of what these companies are using Aisera for, whether that’s part of your NDA or whatever it might be, but any any examples of how it’s been used in those companies would be good?
Muddu Sudhakar 48:28
Absolutely. Look, most of my customers’ examples, which are already publicly available, repeat to read case revision. So look, even if you start with companies like you mentioned, start with McAfee. McAfee is using it for consumer business. They’re using it for enterprise business, they’re using it for what I call it like front office, customer facing side. They’re also using for back offices, right? They’re trying to use for all the businesses. So McAfee has hundreds of millions of users. It could be a teenager to a senior citizen, ask them questions about their cybersecurity and their issues with the antivirus not working on their device on their iPhone or their workstation. Right. So all of that, we have to handle it. So all the use cases will be pretty much all the things that we discussed in McAfee and we did the government, the amount of saving, we save them per year, it’s millions of dollars. They’ve been with us for multiple years. So it’s a great customer for us. They taught us a lot and we learned a lot through that process.
Muddu Sudhakar 49:26
Same thing with Zoom, zoom uses for both front office and back office customer service support it right. Many use cases, we learn a lot. I mean, Zoom is the poster child of this industry through COVID. The other one of the best ideas of 2020. Right? They define what the user experiences. They define the word X in the current CX UX UX right there the X experience right? My day starts with Zoom my day ends with Zoom. Right? I love zoom what they have done For the industry, right?
Muddu Sudhakar 50:01
So we are having somebody like that helping them across all the things, right an example of the thing about where that universal bot, multiple domains, everything so, right, high scale, volume, right, the amount of volume of requests and all those things. Right. So I like that.
Muddu Sudhakar 50:17
See, and then you talked about even vendors like 8×8 bed is a great company for the UCaaS space. We’ve been with them for years, right? Trying to help them and they may want to build their own. This is an example of where a customer, a partner can build their own, I want to give you that optionality to your point like at any point, you say, look, I want to build something more, I want to do XYZ, I will sit with you. I’m the water for a pipe, right? Whether the water is sparkling water or plain water, I don’t mind. I’m here to serve your needs, right.
Muddu Sudhakar 50:49
And like many customers like that, we have them chug. Chug is a very large company. They’re using us for multiple domains, right for HR request IT customer service right.
Muddu Sudhakar 50:59
Now, so I think that Grant Thornton is another one. It’s an auditing company. Right? It’s a world’s top number four. Number three auditing companies, the resources employees, people asking about taxes, audits, everything, right?
Muddu Sudhakar 51:13
So my job is to cater to different industries. I’m not locked into one vertical, right? As a startup, my job is to figure out where I can help and grow the company at the same time build a product, right? It’s a journey. We’re only four and a half years into the company. I’ve got x more years to go 5, 10 years to go, hasn’t this AI industry is gonna be under 20, 30 years. We are just like, the first one of the boxing match. There’s so many wonders because there’s really a problem to solve. There’s not too many people out there that problem is not to be solved. Right.
Nice. Nice. Perfect does a perfect thing to end on. We’re in round one. Ding ding. Let’s go. Fantastic Muddu. This has been absolutely amazing. I’ll put the links in here in the show notes, Aisera.com. For those who listen to our podcast is Aisera.com. And Siddharth just put a link in there as well in the show notes, which is Aisea.com/customers. If you want to have a bit more of an in depth look at some of the state things we’ve just been talking about there with Zoom and McAfee and those kinds of things. So do check those out.
Also go to voicesummit.ai/agenda if you want to find out about what’s going on in October in Arlington with VUX at voice 22 a full day of contact centre, automation, strategy, design, insights, technology, the whole nine yards, everything you need to automate your contact centre.
And if you haven’t subscribed already, you’ve heard Muddu in this conversation so far being very, you know, I’ve got to chop some of that out actually and stick that in some of our marketing material things. We’ve had some very kind words to say about us. So VUX.world/subscribe, if you want to get emailed about all these interviews we do with excellent minds like Muddu, and lots of others as well.
It’s been an absolute pleasure, thank you so much for joining us, I’ve learned a hell of a lot. As I said, Aisera sounds like the water that is moulding to the organisation, whatever their needs are being able to fit into the right place, taking the shape that it needs to, but at the same time, if you want to buy a glass full, then you can do that as well. And so I think it’s a lot of love the strategy, love the way you’ve been talking about it. And it’s been an absolute pleasure to spend the last hour with you. Thank you so much.
Muddu Sudhakar 53:26
Thank you for bringing the best out of me. So thanks for asking all these questions and making this. I’m a big fan of your show and what you do for the audience. We need you to keep doing this every day.