Unleash endless possibilities with YellowG, the industry’s first generative AI-powered platform!

Blog

19 mins read

Conversational AI platform explained: Shubhi Saxena – enterprise automation expert

Updated: March 12, 2024
Conversational AI platform explained: Shubhi Saxena – enterprise automation expert
Conversational AI platform explained: Shubhi Saxena – enterprise automation expert

Conversational AI-powered chatbots and intelligent virtual assistants are the buzzwords for enterprises that want to improve their customer engagement, customer service, employee engagement, or automating their entire processes in the organization.

Today, Vartika Verma in conversation with Shubhi Saxena, Product Manager at Yellow.ai will try to clear the fog around the conversational ai platform, challenges, and solutions.

Vartika: Welcome to Yellow.ai lounge in its second edition. Today we’re in conversation with Shubhi Saxena who is in the product team of Yellow.ai. She wears multiple hats but let’s hear it directly from her. Shubhi, can you tell us a little bit about what you do?

Shubhi: Yeah sure, I am a Product Manager at Yellow.ai, and currently, I lead the Enterprise Automation vertical here. 

Vartika: Amazing! So when you talk about enterprise automation can you just define what that means? 

Shubhi: Yes sure. So Enterprise Automation essentially includes any kind of automation that you can do within the enterprise instead of outward-facing. For the productivity and better digital experience of your employees, maybe your vendor partners, and so on. So I would say anything but the customer-facing experience would come under Enterprise Automation.

Vartika: Awesome! So when we talk about enterprise automation can you describe what is Yellow.ai’s product vision? 

Shubhi: Absolutely! so what we envisage is or in fact what we say is just like organizations hire a lot of people to do a lot of tasks. Currently, many of those tasks do not really require human effort or creative effort, which can be automated and the upside of doing that is that then your resources are free to actually engage proactively with your employees, your customers, with their colleagues and so on. 

So we envisage an enterprise virtual assistant that organizations can hire to automate multiple operations across their entire enterprise landscape. So, everything from human resources management to IT supports to Procurement, Supply Chain, and Finance.

If you see the current enterprise systems that are available are outdated. And the digital experience in B2B is not at all at par with the B2C experience. So what do you want to do is we want to bring in the same or maybe better digital experience within the enterprise as well. 

Because ultimately the employees will create only what they consume, right. So if they are in a digitally healthy fast productive environment, definitely their productivity will improve way more at work as well. 

Vartika: So when we talk of Enterprise Automation for B2B do you not work with B2C kind of companies? 

Shubhi: We do. When I say B2B I think it’s essentially within the organization. It can be any kind of company. But when that company wants to automate its internal processes and streamline its communication, that is where we step in. 

Vartika: Awesome! So can you give us some examples of where you see the product vision is actualized today and you know where you see it in 2020? 

Shubhi: Absolutely! So, if you look at the situation today, we have a stronghold in automating the entire piece of conversation as well as automation. So if you look at maybe, either a customer or an employee experience we are able to automate everything right from their onboarding to their servicing, to their post-purchase or post actions completion, engagement, running data, and analytics on them across channels. Including WhatsApp, voice, web, apps, etc. So that is where we stand now.

Essentially, the infrastructure since we are saying that the world is moving more towards the conversation first mode of conversation you will not just need conversations and few bits and pieces you will actually need an entire platform and infrastructure that can drive those conversations and engagement. 

So right from understanding a piece of content to automating the set of actions that need to be taken to again providing a response that is one. But then a lot of other layers come in when you say that you want your digital experience to be on par with the human experience. You would want to have a lot of personalization there and you want to add a lot of contexts there, a lot of ability to understand and speak vernacular there. 

So those are some of the things that we are working on and we imagine or we envisage our product to essentially be highly modular. So that an organization can pick any part of the entire platform offering that they need in their workflows they do not really need the entire platform. You know maybe they just want the voice piece or they just want the NLP piece and that is what we offer. 

So yes, I think in short we essentially want to perfect the entire platform end-to-end which is needed for any kind of conversational automation. 

Vartika: So Shubhi, can you tell us a little bit about Yellow.ai’s product vision? 

Shubhi: Sure! So currently, as we stand today, we’re a Conversational AI platform, which means that end-to-end all the pieces that you need for any kind of conversational engagement, we provide that. However, moving forward we want to be a cognitive engagement cloud.

So I think the key difference between both of these is that a cloud is essentially modular and you can basically just pick one piece and plug it in into your enterprise landscape for example you might want just the NLP or just the voice and so on. 

So we want to perfect each and every piece and we want to provide it in a much more modular and decentralized manner. 

Vartika: So you actually believe that let’s say a large enterprise will need only one part of the automation and that works seamlessly with the current systems and that’s the future? 

Shubhi: Exactly! The point is that they may or may not. But, today there is no technology that they can go to if they need only a part of it. They need to buy the entire technology and the problem is that because of this they are not able to sort of absorbing new technologies into their current landscape. 

Vartika: So let’s break it down for our viewers, like what could you mean by let us take an example of Indigo and how would they use the Yellow.ai platform in a modular format?

Shubhi: Let say currently Indigo has a customer service bot, right. Maybe, let’s say two-three years down the line they want their crew members to talk to each other… Okay, I’ll give you one very good example. 

Let’s get into the shoes of the crew members on an airplane. There is a lot of noise in the cockpit and usually, there might not be for Indigo but for an international organization or international airways, they might have a lot of accent gaps. They might be speaking different languages and so on. 

In that scenario, it becomes very difficult for them to communicate with each other. So it becomes very difficult for the cockpit stuff to communicate with people on the field because of the cockpit noise because of the differences in the accents or maybe in the language that they speak. 

Now, if they just use our voice and NLP engine, what they can do is they can place a bot in between that takes the input from the cockpit staff or maybe the pilot removes all the noise from it, translates it into a common language, maybe English or the listeners’ language. Makes the accent neutral and then delivers it to the listener on the field and then the same vice versa.

So, here they are essentially only taking the voice piece of the bot. So that is what we mean by modular. Most of the platforms essentially give you a drag-and-drop builder but you do not have a lot of control over the NLP. 

For example, our machine learning model is our proprietary technology. It is not something that is built on top of another NLP solution. It’s a proprietary thing that gives us a lot of control and not just NLP, along with it, you get all the conversational analytics, and all the conversations happening on the bot in the same module. We have testing automation that comes for the maintenance and health checks of the bots and also a support desk. 

So what happens when a bot cannot answer something? 

It has to be handled by a human and that also comes in the platform itself. So we essentially provide you an end-to-end solution, which is easier for businesses to build on and it off and 

I think that is the biggest I would say a competitive fence that we have as a product. 

Vartika: So, it all looks great to hear but, are there really any examples of companies that have really used the power of platform because it’s a very nascent technology in itself. So do you have any examples of companies really using the platform power? 

Shubhi: Absolutely! So the one example that I can give you is that of the world’s biggest oilfield services company. So we started implementing solutions for them around five-six months back and at that time we had started with just one bot which was the HR bot. 

So the problem that they want you to solve was that they had 100,000 employees spread across 120 different countries and the policies were different for each of these employees. So they had a team of hundred and fifty people sitting out of KL which just answered people’s queries day in and day out.

So, they would search through multiple knowledge pieces and documents to get their answers. 

And that is something very repetitive. I mean every day someone asks, am I eligible for paternity leave or can I come earlier than I want it to back from paternity leave? So we built a knowledge management solution that essentially automated the whole process and that is the first thing that we’ve built.

And to our surprise when the different teams within the company saw the potential of that they started adopting this technology more and more for different use cases. One of them is field support automation, the second is assistance for travel-related queries, another one is a tool for essentially their drilling and measurement teams, their good services, etc

Vartika: And this thing, they’re building all this on their own? 

Shubhi: Yes! So currently there are 12 bots being built by their teams all across the world. I have people literally from every continent other than Antarctica which is uninhabitable and there are at least 10 more in the pipeline for next year. 

So, that is the kind of adoption that we are seeing from such large enterprises. 

Vartika: And is it really the business-driven to the level that somebody in the marketing team or HR team or anybody who is not very IT-focused, can they really build their own Chatbots

Shubhi: Absolutely! They can and in fact, that is why we have this concept of pre-built templates. So, that you can start with something you do not need to essentially start from scratch.

The entire conversation builder is completely UI driven. It’s very intuitive. I mean just to give you an example, we today only completed an HR chatbot which is essentially working very similar to oil field service company knowledge management bot where we have written zero new lines of code, no new line of code has been written.

Vartika: So Shubhi, automation has been talked about in the sense of impacting businesses in a very good way but do you have any examples where automation has actually impacted larger lives? 

Shubhi: Absolutely! I mean in fact what we are seeing is that India and Southeast Asia are essentially the home market for Conversational AI solutions.

What do you mean by the home market is that these are places which have adopted voice and chat solutions as their first mode of communication or their first channel of communication. A lot of people might not have studied a lot of stuff but they got access to a phone which had WhatsApp and Facebook and all of that.

So what automation does in this context and if you bring two to three things together if you bring conversations or chats together chat/voice together with the ability to do that in vernacular along with automation what that does is that gives you the ability to reach the larger masses. 

People who are not comfortable speaking in English, people who are not comfortable browsing websites, people who do not have access to very good internet, this gives you access to billions of those people to essentially the next billion or the Bharat as we call it and the impact that it can have is I cannot even imagine that. 

I will just give you one example of one of the customers who we are working with, so we are currently doing a POC with Syngenta. So Syngenta essentially supplies fertilizers, seeds, and so on, these things to the farmers.

What we are able to do now is that we are able to send real-time updates to farmers on what are the best crops to be grown this year or this month. 

What are the exact amount of fertilizers that they should put at what time and we can do this real-time based on the amount of rain that happened last week and so on and so forth? 

That is just one thing the farmers can ask their queries back when we have support agents who can just respond to them on WhatsApp just like how they talk to their friends just like how they send good morning messages they can actually get the message on what’s the best way to grow their income, what are the best jobs out there, what are the different ways in which they can use their vocational skills, and so on.

So I think the only thing that was stopping us till now was the ability to communicate with everyone on the same platform and that is something that Conversational AI has seen and that is something that I personally am very excited about. 

Vartika: So do you think messaging and the internet have a large play in actually bringing automation to the end customers? 

Shubhi: Absolutely! I believe that the one thing that was lacking in the Internet was messaging and vernacular. 

Now that we have that there is no stopping us from bringing everyone on board and you know basically that démocratising technology is something that I really see happening now. 

Vartika: And Yellow.ai has already forayed into this space. 

Shubhi: Absolutely! So we talked about Syngenta along with that, there are large public companies/Govt departments who we are working with or have already closed the deal with them to implement these things on a large scale for everyone. 

Just to give you an example, this is not necessarily coming from our customer list but imagine if you could order a passport or if you could file for a KYC or if you could file your income tax through a bot, through WhatsApp. That would be really convenient and those are the kind of things that we are working on as a company. 

Vartika: So that sounds interesting Shubhi. But, can you tell me how a chatbot is different from a website? I can do all of these things on the website as well? 

Shubhi: So there are multiple things and I get that question a lot from different kinds of people.  I think the key difference between a chatbot and a website is that a website is a non-cognitive channel. It is essentially static or fixed.

Vartika: It’s the same for everyone.

Shubhi: Yes. It is the same for everyone. Even if you personalize it based on the user profile it does not learn over time. So whereas, your bot is something that learns over time, based on your interaction with the bot that is one thing.

The second is, of course, the voice interactions. Voice interactions, voice-first is not something that you can easily conceive with websites whereas you can do that with bots and the third is just the overall accessibility.

I mean I think you would agree that WhatsApp any day is way more accessible to the masses as compared to websites. 

So these are some of the key things that make bots our preferred choice over websites as a channel. 

Vartika: What do you think are the most challenging phases that any employee in the product team here goes through? 

Shubhi: I think the first would definitely be onboarding. I mean even though we do have a process set up for onboarding, this is a start-up. And we really like to you know, since we are growing fast we try to learn and implement as fast as we can. 

As a product team, you will be at the core you will be interacting with each and every team in the organization so just being able to communicate the same thing to everyone just being able to meet their expectations and prioritize can sometimes be a challenge. I think that would be the first one. 

Vartika: So you will just feel that there’s too much to be done and too much to be communicated and you wish you had 24 more hours?

Shubhi: Yes, kind of like that. 

I think that there is no such thing as too much communication for a product person, right. I think all our currency is feedback whether we get it from sales, marketing, customers, etc. 

Our entire machinery runs on the fact that we have to sort of take all of that feedback and keep adding it in our product and how to meet as many things as possible for the company. So far as there is no such thing as too much communication and I think that is something that a lot of product folks may not understand early in the job but I think that is something that will really help them. 

Vartika: So Shubhi, have a lot of new joiners in the company. What do you think are the main challenges that new employee faces here and how do they best combat it? 

Shubhi: In the first challenge or I would say an opportunity in disguise is that we are working on something that’s a very nascent technology. No one has really done it before and there is no model to really replicate. 

So you might be coming across a lot of new situations previously unheard of and you will have to solve them on the spot. Right, so having them that problem-solving mindset and that innovative attitude towards things will really go a long way that would be the first thing. 

The second is I think as a product person there is essentially no such thing as over-communication. So customer problems and for us, everyone is a customer right so the product team, developers are our customers, marketing is our customer and so on. 

So different stakeholders’ problems are essentially our currency when you run on that. So I would really recommend people to communicate with each and every department and stakeholder of the company and outside the company and see how that conversation adds value to the product. How the problems get converted into features as I think essentially how we should operate. So I think if we have these we’re good to go and of course, On-The-Job learning is something that drums on. 

Vartika: So Shubhi, in your past work here, what do you think of the most challenging deployments and why? 

Shubhi: Sure! So, we talked about implementing Conversational AI solutions for the world’s largest oil services company. I think it was definitely challenging but it is something that I’m really proud of was the Indigo deployment. 

That was, I was not really a part of it, it was driven by some other stars of our company like Shweta and Surender. But, the reason that I am proud of it is just because of the scale and the kind of impact that it had on indigo. 

So the situation before the deployment was that people would call them for random things. I mean, I think you can really get how many times you lost a boarding pass in your mail chains, right. And people would call them just to get a boarding pass, just for web check-in, can I take 2Kgs of Aam ka Achaar in my baggage, right. 

So these are queries that are repetitive. And these do not really give agents an opportunity to proactively engage with the customers and that is the challenge that indigo wanted to solve. 

They wanted their agents to be their customer’s friends and proactively engage with them for better conversation. 

So what we did was, we did the entire requirement gathering. We saw these are the things where the top call driver said they’re the most queries coming from and we automated a lot of them and just took some sort of code and some numbers. The bot is less than three months old and right now we have more than one lakh users, new users are using the bot every month and more than a million messages are being exchanged and I think I just get goosebumps looking at those data in analytics and that is something that I’m really proud of. 

I believe it is the future of conversation. 

Vartika: Well, it is one of India’s largest private airlines and I have seen them carrying such a large fleet and personally speaking like I haven’t even been able to get to an agent because airlines are so inaccessible. 

Shubhi: Yes! 

Vartika: So it’s commendable that so many people actually have queries and it’s been resolved now through the chatbot. So where exactly is the chatbot resting? Is it an app? 

Shubhi: Yeah! So right now it is deployed on Indigo’s website itself. So now you do not need to login to download your boarding pass, blah blah blah. You can just do it through the bot itself directly and we have plans of launching it on many more channels in the days to come. 

Vartika: Interesting! Great, if I was to talk about any international deployment that you’ve done. Let’s talk about it not from the customer’s point before the team’s point. Like, what have been the most challenging parts like you know for an international deployment internally and Yellow.ai.

Shubhi: Sure so I think I would say again not necessarily challenging but things that we discovered as we did it was around from a technology point of view was around localization and globalization and managing different languages and stuff like that. 

So let’s say if you have an HR bot where you schedule meetings and let’s say using that bot you want to invite people from five different time zones. That is something you need to think about while building the solution and at the same time the experience has to be localized to each of the countries. 

You might want to talk in French in France and in German in Germany. Our bots are equal to search documents like I told you. In France, all your policies may be in French. So you will have to search in French documents. So those are the kind of challenges and because this is the language this is something that was never a key driver in any technology before this but when you talk about conversations, language is at the center of it. 

I think essentially a language, accents, voice are some of the things that you need to think about when you build scalable and global solutions. 

Vartika: So Shubhi, being one of the frontrunners at Yellow.ai, especially in the product team, what do you think is the future of conversational AI?

Shubhi: Well, I think all communication will move from the multiple channels that it is spread across to the single-channel which is conversations and it will transform because now the channel is cognitive. 

Conversations are getting smarter day in and day out. So I think yeah, that is the future of smarter conversations where technology is democratized. 

Vartika: What do you feel about women in technology? Do you think companies are doing enough to actually, you know, really build women leaders? 

Shubhi: Well, I think some companies are. A lot of them are not, right. But even for the companies that are,  I would really urge all women to sort of come together to help each other as much as possible and you know not to distinguish themselves at least I mean I don’t know they call it breaking the glass ceiling or whatever. But yeah, do not have that bias in your own minds. I think that is the first step that all of us need to take collectively as a society, not just men or women.

Vartika: Thank you so much for being a part of the Yellow.ai Lounge, Shubhi!  It was great knowing you and especially the product vision for the future. All the very best and we look forward to having another guest on this couch soon. Thank you so much!

See Yellow.ai in Action, Schedule a demo with conversational ai experts.

We've been called gamechangers for a reason.

The most trusted & award-winning AI platform out there.
This site is registered on wpml.org as a development site.