Llyod Mathais: what AI changes—and what it does not

That's 300 million people speaking twice a year, 600 million calls. AI can be pretty stupid as well. As they say, there's nothing to beat human stupidity. Empathy and a human level that as of now is far from being replaceable. Has to do a COVID RPCR and get a WhatsApp message and it gives you a report. That's perfectly good, that's perfectly acceptable, are not WhatsAppable. A lot of conversations are not AIable. So Lloyd, you had a very interesting session recently and it was on the same topic, right? Conversational marketing, but not in the age of AI. So now you've discussed a lot of conversational marketing. The net, what we call the best, the cream of the discussion, would you like? Wonderful to listen to all the interesting perspectives.

I'm no expert here at all, but thank you for being kind and calling me one. Just to get a quick sense, just briefly, I was part of an experts panel that I met up, pulled together on conversational marketing. So they push WhatsApp for business as a lot of us get messages. And I think just sharing two thoughts that came through in that session. All of us are customers of one telco or the other. And we know the two big monsters in the room are Airtel and Jio. Now each of them have more than 350 million customers. Just to get a sense, it's like the population of the United States. And the reality is about 300 million of each of their customers, which is about 80%, will never be able to speak to a human agent for obvious reasons, right?

If Airtel and Jio were to set up human agents for all their 350 million customers, assuming each of them have a complaint and want to speak twice a year, they'll possibly need the population of Chandigarh City, every man, woman, and child as a call center, right? So that's a scale. That's real, right? You'll need at least a million agents. One aspect of AI, as we all talked about, is the fact that's where the future is. The other is the reality of business economics, right? If you can't afford to have a million agents, you better find a way. And therefore, a lot of brands are also working at this purely on the economic angle. But having said which, AI cannot just be a machine speaking, right? I heard the example that Sachin gave earlier.

Uniquity Comic One was this about three years back, I think just before COVID, when there was a little tangle between, maybe costing a sense of the subject, the DDCM and the left-hand governor, as there's been forever. And someone from the Amadmi Party tweeted saying, the LG in Delhi needs a rework or whatever. And guess what? Chatbot of LG Electronics responded saying, yeah, we're working. Thank you for your complaint. Please direct message. So AI can be pretty stupid as well. But just back to the point, the second thing that I just want to tee off that came through is that in all this excitement about AI, and I think Grant touched upon it, there's an aspect of empathy and a human level that as of now is far from being replaceable.

So at this panel that I was part of on conversational messaging, there was a large diagnostics company, one of India's top three, that now does a lot of its reporting through WhatsApp messaging, right? So you do a black test, you do a COVID RT-PCR, and you get a WhatsApp message and it gives you a report. Now that's perfectly good, that's perfectly acceptable for a wide variety of things. But just remember that there are a lot of medical tests we do which are fairly sensitive, right? To understand that you are COVID positive through a WhatsApp message is perfectly acceptable. But hey, there are folks who do far more serious tests, right? So is this tumor malignant or benign? You can't have that on a WhatsApp message, right?

So where does the human interface come in? And can AI get down to a level of personal empathy that leads it? And we'll always, as humans, recognize, right? There was another instance of a top matrimonial site that said that a lot of conversations are not WhatsApp-able, and therefore what comes in. So I would say that the two aspects that all of us have to keep in mind is that there is a lot that I can do. A lot of it is something that we need to adapt to. A lot of it is something that's acceptable and can be driven by business economics. A lot of it is going to be very relevant to our business. But there are parts that will need human curation and sensitivity. And that is still some time away. That's just a point of view that I want to share.

Thanks, Vatsal. Thank you for that opportunity. And Lloyd, if it's on. Any questions for Lloyd? Anybody else? It is happening in terms of bringing that empathy in the interactions. So there are a lot of guys whom I interacted with and they are obviously keep on teaching and selling stuff across to me. Where, you know, they are, I mean, I heard about, even Rohit and some of the folks talking about analysing the interactions where you are trying to measure some of those soft settings in terms of whether some of the keywords are getting spoken, how is the tone picking up. So I think that's one part of it. The other part is that in a lot of our interactions, particularly on the KPO side, on the support side, the experience in general

hasn't been very nice any which ways because the agents on the other side are fairly, very ordinarily trained. And you will sense that whenever you get somebody pushing a loan or any other financial product. So what is actually happening is I find interactions with AI a little more transparent. And people declare that you are actually talking to a robot. I've seen that trend happening more and more. Which puts everything in perspective, black and white, right? Nobody is trying to, selling you something which is not there, just telling you that this is automated interaction. And at some stage, the human might step in as and when there's a need. So I think there is a lot of process play happening here where you probably will be much more transparent

and inspire more trust with some of your stakeholders. Great point. I think one important aspect is transparent and to let the person know that you are actually in contact with a robot. At the same time, being as in tune with what a person would want to listen to as talking to a real leader. I think transparency but at the same time, you know, recognize… I want to add something and great call out. So I think some of the things we've seen working with brands, not selling, is using AI to do analysis, especially on agent conversations. And can I get more summaries to analyze sentiment at large scale? Because typically, the analysis is done on 1-2% of the conversations. Maybe 3-4% because we'll have managers and managers to do. What is the adoption

we've seen is suggested replies. Which means, okay, we've analyzed a lot of conversations using AI and then giving the agent real-time suggested replies. Which still means the human is there but you're trying to reduce that friction and that poor inconsistence at least on mass use cases by saying, hey, these are the words or these are the sentiments. I'm able to recognize this by repeatable behavior and these are some of the suggested replies that a business should or an agent could give rather than purely pushing it out to a chatbot or doing that. So just to qualify that to make a great point but it's also the fact that AI's versatility can help immensely. Just to recognize that there's a big battle being fought between the two media networks

on IPM, right? So there's a television, B&M, Disney's Star Sports and then there's Network 18's Geo Cinema and they all try to innovate and for the first time right now, you have IPM in 14 languages, right? And the funny part is that Hindi as of now is just about 30%. So 70% of people are consuming the IPM in Tamil, Telugu, Malayalam, whatever and Bhojpuri, So this is the power and therefore one of the strong selling points that Network 18 is shouting about is that, you know, you can listen to the IPM in any language, Try Bhojpuri. I mean, I did get to great fun but… Absolutely, yeah.