Transforming Customer Support with Conversational AI – Raghu Ravinutala, Yellow.ai

This is Techstrong tv. Hey everyone, welcome back here to Techstrong tv. I'm really happy to have our next guest on, 'cause we're gonna be talking some ai.

I know you haven't heard anything about AI out in the audience lately, but nevertheless, I think this is gonna be one that's worth your while. Let me introduce you to Ragu, Robin. Robin Nutella.

If I mispronounce, I apologize, but Ragu, am I close? You're good. Absolute Excellent.

So Ragu is the c e o of a company called Yellow ai. ai. First, um, if you wouldn't mind, give our audience a little bit of background.

Yellow AI is born with the mission to help enterprises automate customer and, uh, employee experience using, uh, ai. The genesis of the company is my own experience contacting a brand, not getting a right response or taking a lot of time for responding. And we believe that, uh, the world would be great when all the transactional and, uh, the regular elements of support are automated and humans are freed up for forming more meaningful relationship with end customers.

And that's our mission, uh, for the company is, uh, we have been in existence for over the last seven years with, uh, 1,500 plus enterprises customers across 14 different, uh, countries. Uh, these are companies, our customers are companies ranging from Fortune 500 companies to some of the large mid-market, um, uh, companies. We are backed by some of the best venture capital firms in the world, including Salesforce, Sapphire, Lightspeed, WestBridge, um, and we have about 700 people, uh, wow.

Across, uh, multiple different companies enabling enterprises to, uh, to drive great customer and employee experiences using ai. Love it. I was just, uh, actually in San Francisco at Sapphire's offices a couple weeks ago, so that's great.

And you know what, what's interesting is in a, in a, in a field that seems to be less than a year old, but we all know, we've all been working on AI for years and years and years. Yellow AI is a company that, you know, is a substantial company and for at this stage of the market development, right? So congratulations to you and the team there.

You know, I, I half kitted around in the beginning, Ragu as telling the audience. I'm sure they haven't heard anything about ai. We've all been hearing everything about ai.

It seems right. It's sucking all the oxygen out of our new cycles of our venture cycles of our, in the tech world, not, and, and, and also this is something that transcends the tech world, right? I, I tell people this cloud was a very technical thing, right?

When your mom or your grandmom looked up and said, oh, yeah, I keep my pictures in the cloud. They didn't really know what the cloud was, and it really didn't make a difference to them. They just saw the commercials, right?

But AI has the promise to really change not just our tech world, but our world. Um, but like any new thing, there's hype and there's, there's, you know, overhype and over promising. And I think people, you know, AI isn't gonna change everything everywhere, all at once, right?

There's some things where definitely make sense and things where we could work today to make it work for us. And I'd like, if you don't mind, you're the expert. You're the c e o at, at yellow ai.

Tell us where, where is, where's the rubber meet the road? How do we determine what's the best use for what's available today with ai? And how do we use it to make tangible improvements to our business, to our business models, to our customer experience today?

Absolutely. So, uh, AI has been around, as I said, for several years, and it has been, uh, put to phenomenal business use also over the last, uh, several decades, uh, especially around, uh, predictive, uh, ai. So how do you, how do you predict weather?

How do you predict, you know, machine maintenance, et cetera. So all that been around the reason it's now become more popular because now AI has been applied to large language models, which can replicate how humans interact, uh, with other humans. And now this is tangible to a large, uh, part of the demographics because now it's come from, uh, backend to more frontend and tangible that everyone can relate to.

It's not something residing on a cloud or something behind a computer. Uh, so the biggest, uh, change that we have seen over the last one year is large flag with models, uh, you know, almost replicating, uh, human kind of, uh, interactions. And they're very expensive.

So the next big thing for the enterprises is how are we going to take this new capability and, uh, adopt it our enterprises that can drive up significant business value. So the question mark about business value still, uh, remains. And what we have seen is, um, high value problem areas where there is significant expertise of humans that is needed and that can be replaced or amplified with AI are the right areas that are making a big difference.

Uh, one of the first areas that has seen a tremendous growth is, um, writing, marketing copy, copies and blocks, et cetera, which is rightly suited for LLMs and which requires high-end expertise, where the cost of the problem is significantly higher. We've also seen this expand into legal documentation, um, uh, documentation around, uh, uh, high value, um, non-generic, uh, documents and, uh, you know, document writing, et cetera. Those are areas that we have seen.

And in our own world, uh, what we have seen, uh, the large language models expand is the ability to empathetically have the interactions. So, which means that the breadth of interactions that they're having for customer support is increased, which means that instead of automating 60% of our, of the customer interactions, now we are probably able to out automate about 90% of the interactions, which again, is a significant value for the end customers. Um, we are also seeing areas where these interactions can, uh, make decisions, uh, uh, instead of humans, for example, they can get a budget for a day.

Uh, they can, uh, get the budget for providing offers, discounts for the end customers with the goal of, uh, retaining, you know, churning customers. Uh, so that problem statement, ideally being handled by humans now can be handled by ai. So the next, um, one year, two year is about adding, identifying these key pockets of high value business problems that could not be solved without LLMs.

Now, that can be solved, that can scale is the key, um, uh, area where the rubber board meets the road for the enterprise. Agreed, agreed. You know, look, as a publisher of content, you know, obviously we're, we're, I want to say playing, experimenting with AI in terms of writing blogs and, you know, papers and, and similar kind of written things.

And, and yeah, it doesn't, I wanna just be honest, you can't just ask it to write something and, and you off, you go. It, it, it could, it in my mind, there's two real ways to use it. One is it gives you a base from which you can then edit, add polish, and perfect.

So it gives you that base. So it's 70% of what you need, but then you gotta add that 30% and you also gotta make sure it's not hallucinating and everything that's in there is real, right? Yeah.

Or you could do the 70% and input it in and see what edits it makes. And believe it or not, we find that that's more valuable. 'cause it actually does a nice job on that.

Absolutely. Right? 'cause the other way takes a, it takes us a long, it takes it almost as long to edit Yeah.

As it does to just write the thing. So, um, but you know, we were talking off camera, I said we had this hackathon last week or two weeks ago now with some very, very well-known DevOps people about operationalizing AI for operations, operations for ai. And, you know, the kinds of things that you've spoken about ragu, things like, you know, taking information and creating sort of a custom L l m right?

Feeding into a vector base, putting it into a custom l l m, run that on top of your chat, G P T or whatever you want to use, and, and allow that to help with code. Yes. Right?

I, I saw a stat, something like 42%. They, they claim 42% of all the code in GitHub has AI fingerprints on it somewhere. Yes.

Either they, they ran it through ai, or AI initially did something with it, but something AI had something to do. That's a staggering amount. Absolutely.

A crazy amount of code. Um, and it's only going to get better. I, you know, I think we all recognize that.

Yeah. So I, I guess my question for you is, you know, when we talk about it and our audience, is it people, how real, how real is this from helping them with their jobs today and let's say next year, right? Because three years from now, who the heck knows three years from now?

Absolutely. So, um, at the core of LLMs, uh, I think the core is about them being able to predict the next best word. Uh, yeah.

And, and lot of, uh, everything. I think that's what the l LNS are basically doing. Um, and if I come to, um, IT people, I can see a range of, I think coding is a phenomenal example where it is really expensive to get, uh, developers and you wanna amplify their productivity.

And, uh, the dev productivity is one of the significant areas where AI is impacting and, you know, IT teams can absolutely leverage to, um, uh, to its best. And we have seen it working really well. In fact, converting huge amounts of applications into no code applications, because you can do a lot of things even without coding and using prompts, um, because you're able to, you know, uh, drive to the applications and drive to the outcomes with that.

So I see tremendous productivity gap productivity improvements, uh, from using LLMs for, uh, for coding. The second area for IT operations is about managing the documentation, uh, uh, managing, uh, with the interface with the end users, uh, managing the question and answers. And so, uh, I think these are, again, areas which are very well suited for, uh, leveraging the AI and large language models.

Uh, so I think that's an area where IT teams can absolutely leverage. Uh, the third area is about, uh, workflows. Um, one of the key things that, uh, AI and LLMs are able to do right now is look at the workflows, uh, that are happening in the system and actually codify them or, um, or, or automate them.

Uh, so a lot of it workflows are not well documented and not automated, and they are ad hoc, uh, leading to a lot of, um, you know, wasteful expenditures and wasteful workflows that, uh, potentially can be automated. So, uh, the third area, which I think could be of significant value is how does, uh, custom trained LLMs on enterprise knowledge and their enterprise workflows can be used, uh, to automate those, uh, workflows? I think these are the three areas where IT teams can, uh, leverage and, and potentially amplify their productivity 10 x uh, over the next one to two years.

I, I, I don't disagree with you. I don't disagree with you. What about, so here it, this, you know, again, coming outta my experience, uh, at this hackathon, how do we get developers and ops people, DevOps?

How do we get them smart about using custom l LLMs and, and conversational ai? I'll give you an example. Rather than asking the code to write, let's say, a a, a Python script, I saw one of the people here feed the Python script into the chat bot and say, explain it to me.

And you know what? I'm not a coder. I'm not, that's not what I, you know, I never, I never even stayed in a Holiday Inn last night, right?

But man, it explained every line by line, by line of this Python script about what it does, why it's there, what calls it makes, where even someone like me could learn to do Python using this thing in a week. For sure. I mean, how, how do we, how do we train people to use this, right?

Or how do we tell them, if not train them? Yes. Uh, um, I think this is a challenge that I think, uh, uh, in general companies are seeing and we are seeing with, you know, uh, our own company as well, is how do you kind of drive the adoption?

And the best way that we have seen, uh, uh, to drive that adoption is actually making users kind of fully lean into using this for every single, uh, uh, trying this for every single piece of work that they, um, that they do. So I think we are doing the right thing, be it, uh, conducting hackathons, um, uh, you know, identifying, recognizing the team members who are able to, you know, maximize the, uh, uh, adoption and usage and amplify their own work. So just kind of bringing this as a part of the narrative, uh, in the company.

And the worst thing that you can do is create specialized teams to use ai, because now, uh, AI is no more a specialized activity, and that's, uh, AI should kind of run through the several organizations of the company. And every single work has, uh, potential to be impacted by ai, which means that it has to be democratized and used across the company irrespective of the role, whether you are a Service desk agent or you are a c e O of the company, you know, uh, writing out the blocks. So I think those are the key, uh, you know, considerations, uh, to kind of get that option to, uh, very high level and, and enter up the maximum benefits.

Excellent. Agreed. Ragu, we only have a few minutes left.

I, I want to dedicate it, I want you to tell people about Yellow AI and how they can work with the company, you know, partake in your services. I, I, you know, I, I feel like we spoke a lot about AI over here, but let's, let's talk about yellow. Yeah.

Yellow AI takes the smartness of these large language models and brings in, uh, security, uh, controllability and, uh, enterprise readiness in enabling you to drive, uh, customer and employees support auto autonomously. And what we have seen at our customers is 90% plus automation rates on employee and customer support. Uh, we've seen, uh, customer satisfaction rates, uh, move up to 80% plus.

Uh, and this essentially is reasoning in, you know, multiple million dollars of savings while improving the customer experience. ai. Or the best is we have a great AI based assistant on our website.

Uh, you could interact there, um, to get more information or, uh, that can help us, uh, you to connect with our team as well. So, uh, absolutely. Excellent.

Thank you for being on Tech Trunk tv. Ragu, we invite you back anytime we, you know, look, we're gonna be talking about AI for, for a very long time, so anytime you want to come talk with us about it, please come back and, and, uh, keep, make, help make us smart about this. Absolutely fantastic, uh, talking to you.

Uh, All righty. Thank you. Ragu Alala, c e o of yellow AI here on Text Drug tv.

Go check them out, right? One of the companies that are doing real work with AI today. Uh, we're gonna take a break here on Text Strong.

We'll be back in a minute.