Source-backed interpretation.
Editorial interpretation is derived from the original DMAasia/WION video description and transcript. Statements by speakers remain attributed to the source; no unverified current-role or performance claim has been added.
The durable question is how AI and automation should improve AI without losing sight of data and measurement.
Source signal, edited only for readability.
Three takeaways worth carrying forward.
- 01
Read AI and automation as the primary operating question in this record.
- 02
Test its connection to data and measurement and attention against the speaker’s reasoning in the transcript.
- 03
Keep the 2023 source context visible before carrying the lesson into a current decision.
What the transcript puts on the table.
It's only when you start giving me explicit signals of, this is the content I like in this language at this time of the year, or this time of the day, that's when I start really starting having fun with it.
As an aggregator, I can give you the same news from three different sources.
Because now the two systems are going to start talking to each other.
Original context. Current reading.
Recorded in 2023, this captures the language, platforms and constraints of its time. Read it now by separating those period-specific details from the enduring operating question: how should AI and attention improve the decision?
Durable tests for the work.
- Use AI to improve a decision or outcome—not to decorate the plan.
- Earn attention with relevance before scaling reach.
What would you keep, change or measure differently if this event intelligence were put to work now?
IP09 · READABLE TRANSCRIPT1,314 words
For us, AI is not something new. It is the way we operate. We have data which is worth, I mean we have 7 petabytes of data. 100% AI, just to give you some numbers, AI is not something that we rationably use. That's our core business. AI helps me do that because it's manually not possible to do in any form and manner. We take a cue, Samir, for you, for customers want to hear what they want to hear. And as Indians or most, we are very opinionated. And news for one gets driven by that and we like politics, we like religion, sports, and all of them are treated as religion. How do you look at being a sports fanatic especially? I don't watch the EPL full but I want the news.
How are you using sentiment analysis or even an LP on understanding what each consumer wants to consume and how do you deliver that versus somebody pushing an agenda through? So thanks for this. It's a relevant question for somebody like us. So guys, I was looking at talking about the previous sessions granted I had a sidebar as well. But for us, AI is not something new. So just to give you a little bit of the lay of the land, at a CM of Daily Hunt, we have just to give you some numbers, just to have some fun with it. We have data which is worth, we have 7 petabytes of data. So that's what is used to train our algorithms and we are nothing but a 100% AI slash algorithmic content aggregator, which means anything
that you see on my app, which is a vernacular content is not driven by any editor anywhere in the world. So it's 100% AI. So just to give you some numbers, just to have some fun with it. We are there in 14 to 15 languages in Indian languages, such as Urdu, Nepali, Hindi, English, all of that. Around two and a half, three lakh pieces hit me every day. I metatagged them. I figure out what works, what doesn't work. And eventually there's a score assigned to each and every content piece that goes through my filter. So let's say, as an example, yesterday's match between CSK and MI, same reported by three players, which one do I pick, why and in what order. Then overlay that on the individual preferences.
So for us, simple thing, Grant and I were talking, the single biggest thing that we do, which is very difficult is when we power up the app for the first time, you just installed me, I know nothing about you. It's only when you start giving me explicit signals of, this is the content I like in this language at this time of the year, or this time of the day, that's when I start really starting having fun with it. But for us, AI is not something that we fashionably use. So that's one. Also, to Grant and a couple of other people, I just wanted to make a couple of comments. I'm 100% in Grant's camp that it's a tool. And just to have some fun with it, and just side-barring for a moment, I'm actually having an ongoing dispute with Citibank about a bill.
You know how they write, right? I'm just using ChatGPT to summarize and creating, using ChatGPT to reply. It's 30 seconds for me. I'm just messing with them to see when do they realize that somebody is really taking over the ride. But as we discussed, it's just a brainless system right now. It's how well you used to do. And hence to go back, honestly for us, it's the implicit, explicit signals. And how we use that AI, which actually is now separating the men from the boys. And there is no aggregator in the Indian content ecosystem of our magnitude or scale, but more importantly, political leaning. As an aggregator, I can give you the same news from three different sources. And nobody can tell me that I'm biased in any formal manner.
And honestly, that's the first question Grant asked me saying, how do you account for bias? So my 2019 campaign talked about how if you read a news and that type of raphael, if you remember, was a big topic in the Indian ecosystem. If you read a news about from NDTV, NDTV being a surrogate for X party, the next article automatically had to be a Z news because that would automatically counterbalance the part. So I've played the role of an aggregator. AI helps me do that because it's manually not possible to do in any form and manner. So AI as a tool is very powerful versus depending on the black box. I'm not assuming that I'm going to fall ill or I need to, yes, doctors tell us eat well and eat this. How does that communicate back to us?
Just add that to your fitness watch on an apple and see the magic happen. Because now the two systems are going to start talking to each other. So now I'm just pulling other stuff out of the air. Start thinking wearables, start thinking ordering of your medicines, start thinking diet. There are so many other systems who are currently not talking to each other. Start pulling them up together and a pen picture will emerge of when most likely are you going to fall in. And to your point, if I said seven petabytes of data, this is his seven petabytes of data. Because this is what he's playing with. And I have a similar journey because I can tell you that if this is Bombay, if it's May,
if it's two o'clock in the afternoon, if the connection is Reliance, most likely this is the guy. Very interesting. Because that's the level of data I'm looking back at. Because it's a remarkable degree of accuracy. Do you use email and if yes, how do you hyper personalize a subject line on content? Actually, my experience. Very targeted question. Honestly, not a lot of email. No, ours is a pure video business. So I think that the coolest exercise that I know about this is, is a product called Adgreets. If anybody from Flipkart, they will tell you. Adgreets is basically an ability to generate 50,000 versions of the same video. Now imagine selling an AC in this country across 50,000 pin codes with 50,000 different price points. So unbelievable.
When I saw it for the first time, it blew my mind. And I've seen Adgreets at play and I've seen Amazon use it. I've seen Flipkart use it. But yeah, for us, it's similar, but honestly, limited use of email. For us, it's hardcore video. Whatever we do has to be video forward. And for us, some of that play is actually in a very oblique location, which is notifications. So for us, that slice and dice, what to send, where and how is more a notification play and not an email play for lack of a better. Anybody from HDFC Bank here? So again, just, again, one of those benchmarking organizations to look at. Over there, customers, over base, they have around 60,000 cohorts that they deal with. That's a lot. That's large. And I don't know.
There are ex-bankers sitting in the room. So they could opine on it because it's a really large data set is what I believe. And they run multiple campaigns across the board at any given point in time, across channels. So there are the slice and dice on consumer to channel to product, to pin code to geography, to timing. So there are some other things that I can do. And I can do that. So I can do that.
