Ep
3
November 1, 2024
42:19 Min

AI's Leverage - Timing, Data and Human Relationship: Insights frm Aravind Krishnaswamy

On the Podcast
Speaker

Aravind Krishnaswamy

Director, Product Management for Cisco’s Contact Center
Host

Thiyagarajan M (Rajan)

Partner, Upekkha

Joining me today is Aravind Krishnaswamy (Arvi), Director, Product Management for Cisco’s Contact Center & Customer Experience. We discuss the opportunities and challenges presented by the shift towards AI in product management and startups.

They cover topics such as the evolution of technology, optimizing for distribution and proximity to users, the role of probabilistic systems, and the importance of data and geopolitics in the AI space.

This conversation provides valuable insights for Indian founders and product managers looking to leverage AI for their startups.

Here are the key talking points from the episode:

  • The importance of timing, leverage, and solving real business problems with AI
  • Arvi suggests that Indian founders should optimize the distribution
  • Discussion on probabilistic systems and data-driven decision-making
  • Arvi highlights the potential of service-based models
  • Importance of considering the infrastructure stack
  • Challenges and opportunities for product managers in the new AI paradigm
  • Discussion on the significance of access to data and its strategic implications
  •  Importance of owning the relationship with the end consumer

Arvi's approach to keeping up with changes in AI

Aravind 'Arvi' Krishnaswamy is an entrepreneur and tech executive living in Bangalore, India, with his wife and daughter. He holds multiple software patents and is a published author on product management frameworks.

His passion lies in using technology to improve people's lives. Arvi has worked with several startups and public companies (Cisco, Intuit, SupportSoft), building products used by billions worldwide.

He has experienced successful NASDAQ IPOs and two startup exits, most recently Cisco's acquisition of CloudCherry in 2019. Arvi is currently the Director, Product Management for Cisco's Contact Center & Customer Experience.

He works on improving how frontline employees offer personalized experiences for consumers who connect with their business.

Transcript

Hey, Rajan. Nice hat. Love the colour. Thank you. Thank you. So welcome.

Uh, welcome to this episode of, uh, pivotal clarity Podcast. I'm so excited to have you here.

I always, uh, enjoy our conversation.

And, uh uh, I wanted to sort of bring you to the podcast because we've been talking about a I on and off on, uh, like, you know how things have changed.

You know what it means for product management and like what it means for start ups and founders from India, I thought, uh, the audience would, uh, greatly benefit from, uh, the insights that you have the conversations that we've had in the past. So welcome to the podcast.

Thanks, Raja. I'm super excited to be here. Thank you. Thanks for the opportunity.

I mean, you've got an incredible audience of followers who, uh, listen to insights that you share and, uh, just happy to be here. Yeah. So for the listeners, uh, let me do a quick introduction of RV and I We like for you to, uh, add in whatever I have, Uh, may have missed out.

So So, Arvind Krishna Swi, as RV is known, is the director of product and strategy for a I at, uh, Cisco's, uh, contact centre and Customer Experience Business unit. He joined Cisco through an acquisition of a start up, uh, customer experience start up called, uh, Cloud Cherry, where he led the product and engineering as part of the founding team.

Uh, like, uh, I mentioned we call him as RV. And he's been in both the start up as well as the public company sphere, Uh, including Intuit and Support.com. And he's seen at least two acquisitions and two IP O RV and I overlapped during into it around 2015.

Is that right?

Are we?

And, uh, as well as, uh, when both RV and I worked side by side during our mobile start up days, uh, around 2010 now are we is leading A I projects within Cisco and has a unique vantage point on a I as the incumbent, which has the distribution, the data and also a newfound resolve through its, uh, new fund to be, uh, quite ahead in this, uh, a I platform shift.

I mean, anything that you'd like to add that I might have missed.

No, that's a great summary. I think.

You know, uh uh, for me, the journey has been a a mix of different start ups as well as, uh, multiple large enterprises and seen, uh, multiple waves of technology that have driven things.

I think, uh, mid-nineties. I was at IASC working on forum and sort of saw, uh, evolution of, uh, what I'm gonna call old computer.

And, uh, from there, the way with the internet, Uh, the hype, uh, companies that emerged through it and companies that are the same ideas and how they progress literally a decade later, through down to now, And, uh, a I, I think represents the biggest shift that I think we've all seen enough people have spoken about it, but I think, uh, we're at an exciting point where I think it's There are specific things around new computer, old computer, where I think there are opportunities today which, irrespective whether you're a big company or a small company or a start up, I think there are things to take advantage of that could start to really shape the, uh, companies of the future.

Yeah. So let me start with, uh, this first question. It may be a little personal. One. Your website, where you write your thought is called the s.in. A zero dot N.

Uh, what is it about?

Like, uh, why?

What does it mean?

Yeah.

So, uh, Argos, uh, it it turns out Argos was, uh, uh, the dog of Ulysses from Homer's odyssey.

And, uh, there's a beautiful story of how, uh, Ulysses went off to the war, Came back after, uh, over a decade and, uh, how his dog was waiting for him and, um, recognised him, uh, when he came back just before the dog passed away.

Uh, there are different variants of that ending, but there's the A more beautiful one that I'd like to imagine really happened.

Uh, for me, it it was personal.

Uh, as, um, uh, our dog who passed about four or five years back was very close to, uh, us, uh, close to our heart.

Uh, in fact, uh uh, Right here is our dogs, uh, my daughter. A little decoration on top of it, too, but, uh, yeah, So it was our goss.

Just, you know, dogs and how they're a part of our lives. And a bit of a geeky reference to, uh, 00, which is greater than ourselves and our own understanding of ourselves and what it means in our journey. Awesome.

So, uh, are we If we were to start a start up idea now, uh, in the new platform shift of a I, uh What type of ideas would you work on?

And why would you pick those?

You know, that's that's That's probably one of the toughest questions people are taking today. So you put me in a bit of a hot seat here, but, uh uh, here's what I'd say.

I, um one way to think about it is go back to other similar big waves that happened.

Uh, both you think of mainframes, and, uh, back then you had, uh, Steve Wozniak and others who were part of hacker houses tinkering with different things with hardware connected with Steve Jobs and found a way to solve real consumer problems. And you think of the, uh, early two thousands, the emergence of the internet.

Uh, I was actually I see there sitting across this ID, ID and D machine, and, uh, we I got to see, uh, N CS a mosaic for the first time and we could. We realised it was world changing. But to go from that opportunity of Hey, there's the Internet. You could connect all this to solving real business problems.

Uh, Amazon really, uh, led the pack there, starting with one single category of Everyone wants books, and they're finding it hard today to find the books that they want.

And how could you solve that?

Well, with the Internet And from there they expanded category after category solving to become a far broader market play. So I think across these, sometimes a broader vision. But solving for a certain category where you can deliver outcomes really well, I think is key.

And, um, the contract of this is I'll give you Sometimes you can have a number of different ideas.

And, uh, in the two thousands, I was living in Menlo Park, and I used to order my groceries through, uh, Webvan. And we come to that saying that you know, there is Amazon, and then there was wean. Right.

So how do you pick?

How do you know?

Right?

And so, um, it's not The Web band was a bad idea, but it didn't last more than a year back then because the complexity of all the things that it took to get to product market fit were too high.

But, uh, a little over a decade later, you know, you could be in Bangalore, or you could be in a small town anywhere in the world in a in a places where you can use an app and get things delivered home. So sometimes these great ideas take some time to, uh, pan out.

So, uh, making careful choices about areas that you can get traction figuring out which is the right ones, where you're getting enough traction that you can double down on.

So if you're a large enterprise, understanding the broader patterns of where there's traction and areas that you could double down on where you have more leverage or if you're a start up figuring out where you can accelerate the key levels, that will give you the product market fit, acceleration and then broader leverage, I think a key at at a point like this. So you talked about leverage.

But then there is also this element of timing, right?

Like you said, uh, Webvan back then did not work, but now you go to any small town or city. You'll find that there is an Instacart or there is a Ziggy and, uh, it it's basically same web and, like, you know, 10 or 15 years later, so timing is also key.

So how do you actually make those choices around timing and key?

Uh uh uh, Aspect of leverage. Yeah.

No, it's a great question. And I think, Look, there's the business outcomes that we want to solve for ultimately. And I think those still need to drive our focus and not let technology drive, uh, choices, because sometimes with a I as an example, we spend a lot of time talking about technology.

But we don't talk about the business problems, uh, that we're trying to solve, but we're also the phase where there is a balance of the two.

And, um, I'm gonna take a couple of minutes to share an example here.

Before I talk about this, I'm gonna go back to, like, literally, like mid nineties, uh, while I was working with Parham and later at grad school in the States, uh, I worked on grand challenge problems like meteorological like weather prediction which involves using large data sets from all around the world and running complex meteorological models to predict the weather.

Now this prediction models were there even in the mid nineties, using high performance computing to do. And you can think of next token prediction as further evolution of that. In the various ways those models back in the mid nineties were based on, uh, finite difference and finite element analysis.

All, uh, calculus, Right?

So it was complex calculus based models that scientists had come up with which were being implemented in code, a lot of it in FORTRAN and then later in C and C++ as it is set up with PB M MP I huge complex clusters. So that was an old computer, and we talk about Gen. A.

I and oh, it's hitting adoption issues. It's not always accurate. And we have all these problems and two computers faced with these challenges of, uh, I don't know if I can rely on it. I don't know. The outputs are, uh, trustable hallucinations.

So the analogy I want to pull us back to here is, uh, example of, uh, Malian, uh, India's own moon Mars mission, um, and things that we did in the nineties. Calculus can only take you so far. With calculus, you can send a spacecraft to the moon or to Mars.

But if you want to find the right landing spot where you want your rocket to go, land at the right place and land without crashing, it's all probability. Catalyst cannot solve the problem.

Well, today, probability is used to determine where to land, how to land and all those details. So even today, in the real world, if you want to solve a problem, well, that is so critical it's probability that is able to solve it well for you.

So the stochastic nature of those systems, as long as they have access to the right data sets, is still able to get this to you at higher fidelity than traditional calculus based systems. With weather forecasting, you could get by with calculus because it, you know, at at occasions of rain and so on.

Uh uh, there's some level of fidelity, but when you're landing something on the moon or on Mars, you still need probability. So part of the generational shift that we're seeing here to me foundation is how software is getting built, Uh, we have traditionally always been in this mode of Hey, someone is gonna go talk to all the users, gather requirements, come back. Requirements may not be accurate.

They come down, you build some software engineers may not have understood what logic to plug in.

Finally, it gets tested. You don't know if that's right or wrong. It gets out of the market, and then you realise this huge disconnect.

Hm?

In the last decade, we've all spoken about how we've got to be more data driven. As product manager, we've got to be more data driven and thinking about how to do things.

What are the insights that you want to gain from the market?

Customer feedback. All of this to think about what is the right thing to build now, this data on it through a different base And part of the generational shift that I think we are seeing here is how data itself is being used, uh, to drive the logic of the system and learning how to use this new computer is part of the evolution that I think, uh, we're on that.

Does that make sense?

Yeah. So are you saying that you know. So for your for for somebody to get the answer right from a leverage and the timing perspective here, you're dealing with things which are, uh, like, you know, we use the word stochastic.

I mean, when we use words like stochastic and probability, we lose half the audience, what you're saying is something which is not very predictable.

And, um, like when you deal with, uh, environment where a lot of unpredictability is there, so you cannot use things like calculus, But you have to use things like probability.

And most of the software that was built was actually in an environment where many things are very, very well, uh, sequenced as a logic, right?

And that's what you are calling us as calculus.

But like now when we have to actually build systems the leverage and like, you know, the timing would be more around thinking about how you use, uh, like, uh, uh, math that will help you deal with things which are unpredictable, right?

That's right. That's right. And until the, uh, uh, probability based systems have access to enough data and the right data, um, they will not be as mature as systems where they're entirely programmed with logic. But as they have access to the right data sets and as they apply to the right problems with the right expectations of fidelity and accuracy, those will improve.

We've seen the ship now the last year or two where we have moved from a I being mainly used to assist humans with different things, towards more autonomous ability to help understand problems, to help solve problems and to take action on its own and part of this evolution.

Here, too, is where, with access to the right data sets applied to the right business problems and within the right boundaries of either organisational knowledge or personal knowledge and various other types of actions that you try to enable. These are starting to happen.

So today, whether you see a broad range of applications of a I, where it's using different data sets either within an enterprise for consumers, uh, that use them, use it to help them with writing. It helps them with making PowerPoints. It helps them with creators the number of TIR designs that you can make.

I think that there is a very strong proof point when it comes to creators generative a IS ability to give creators plurality of choice where earlier anything that you were trying to create and you could maybe come up with two or three. You can easily come up with 25 very, very quickly. Get to that 95% point as a creator and then choose how you want to go from there.

And it could be that the last five person is still something you do yourself. Or maybe you do the whole thing from scratch. But the plurality of choice widens your horizon and lets you go broad, almost like a design thinking principle of Hey, go broad before you go narrow. And it's letting you go far broader as a product leader, as a designer, as a strategist, and how you would approach that.

So a designer first becomes a curator rather than a creator, right?

Because he has, like, you know, hundreds of options.

And then he's picking which one of the ones that are generated as a design that he wants to sort of suggest, Uh, instead of saying OK, how do I like, you know, draw the outline?

And how do I actually colour the pixels and so on and so forth.

So are are we, um how why don't we make this a little more concrete?

Um, like like when?

Like whether it is probabilistic or whether it is like, you know, deterministic.

Um, like, you know, when our founders is building a start up or an idea.

Um, there is this element of going to the market. And then there is competition.

And, uh, also from a shift perspective, there is this discussion around saying, Does the does the David have the advantage of the Goliath?

Have the advantage And how do you actually think about it?

And like, make it concrete with an example for like, let's say, an Indian founder.

And that's why I asked you the question, saying What would you pick?

Where is an opportunity for an Indian founder to win and how to think about from a generative idea, Start up perspective. Yeah. So I think if you're sitting out of India often at times, one of the biggest challenges is, uh, distribution and proximity to the actual users of the software.

So, for an Indian founder, I would say optimised for considerations around these two because, uh, these otherwise get in the way.

So proximity to who the buyers and users of your software are, uh, so distribution with respect to how you can get to them easily And, uh, how you'll understand, uh, whether they you are meeting their outcomes and to be able to quickly hydrate with them is to be singularly, uh, an immediate focus.

So if you're building a gen A based software, for instance for creatives it could be in the movie production industry, It could be, uh, generating PowerPoints. It could be helping people with, uh, generating websites from, uh, things that they have.

Then here, uh, you want to optimise for How are we going to get to those users and work a little backwards from that aspect of the journey and then really optimise for the fact that there's so much of noise out there today that you really want to be able to deliver an outcome for them very, very, very tangibly and quickly. So let me interrupt you there.

Um, so when you say distribution and you need access to that and if you're an Indian founder and you use the examples of movies and making creatives, um What is your view on India versus US?

In terms of geography for customer market. Yeah.

So, uh, for certain types of markets, for instance, especially enterprise.

Uh, it it, you know, here, uh, aspects of go to market.

Uh uh, business reach if you're sitting out of India, uh, then establishing that is critical.

And, uh, as I think you know, many founders have discussed having people in the US to build that bridge and link is incredibly critical to ensuring that you succeed, because those strategic partnerships are incredibly important for having that leverage.

But once that aspect of distribution is something that you have a bit of a handle on, then, uh, I'd say from there, figuring out for who who uses your product and tries it for the first time or for the first couple of weeks, you want it to be something where they can immediately see, uh, benefits. Because Gene is also a point today that a lot of people have tried it.

And there's a sense of Hey, what?

I don't know if it's going to fully meet my needs. So you've got to pick an area where, uh, you can move the needle from Hey, it's 80% there, 90% there to 98 99% there or, ideally, offer something a little more turnkey for them, and we can talk a bit more about that.

But I think that here's an element where, um, I think for Indian founders the opportunity to depart from some traditional thinking on Hey, I'm building a SAS application.

Um, maybe the opportunity there, since there are enough SAS products, enough dashboards, enough of those today and today. Programmatic interfaces. The ability to interfaces with low no code platforms give enterprises the flexibility of choice but also have a way in which you can land where someone's able to quickly get started and see some business outcomes out of the product. So if it's like movie production, you wanted to you.

Let's say you pick documentaries as an example, and if someone who's able to quickly come in there, they have some existing assets with those existing assets, you're able to give them, uh, 20 different types of documentaries that they could right away, publish and start to use. That's very, uh, tangible. And if there is a 12 3% I'd say that as a I continues to mature.

Um uh, in my opinion, to solve for the business outcome, we shouldn't hold back on a service based model. If Indian founders had the leverage to partner with other services companies in the ecosystem who could take it through to the 100% in some form. And that's a workable business model.

Uh, II, I think that, uh, fundamentally breaks up no software, uh, mode towards leaning in a construct Where, uh, it's a service at the end of the day and software drives it, uh, it could be humans.

Uh uh, training a IA I training humans, and these constructs will improve and various examples of how the humans are interfacing with a I that will help drive continuous improvements to take you from that 9798 closer to, uh uh enabling a I to do it more autonomously.

Yeah, uh, that's a That's a very, uh, fascinating area. So one of the things that, uh uh, like, you know, I also, um, covered in a report that we, as ope I had published at the beginning of the year saying that the software and the services market is gonna be pretty big, and then India has a big advantage.

One of the the thing about the evolution of that market is is, um I think you know, it's it's companies like Accenture and T CS who are, like, you know, far ahead and leveraging the trend. I think the start start ups. I'm yet to see them, uh, like, you know, pick that and really build on that. I'm completely optimistic about it.

I know that it is bound to happen, and it's an area of advantage for founders. But just in India, uh, like, you know, start up. Founders adopting that are like, you know, few and far examples, uh, at the moment.

But that's a great opportunity, right?

That's a great opportunity in terms of leverage. That's a great opportunity in terms of, uh maybe, um, like, you know, solving some of the issues that you said with respect to go to market. I think it's a huge opportunity. I. I think people have tapped into India in the past for labelling services, but, um, there's a creator's economy that's opening up.

But there's also a curator's, uh, economy that, in effect, is opening up the curator's economy. I don't think you know anybody has written something like curator economy. And that's where I I'd like to see it.

There is literally a curator's economy that's sort of opening up here where, um um uh I think the job creation opportunities here where I think new jobs will start to get created where there are people who would be able to come in, look at 20 different types of things that are possible to offer inputs because today, with a I, you can get from 0 to 90 95% on a lot of this very quickly.

One part of it is literally the curators input on which one to go with. The other part of it is literally then the part of going from 95 to 100. So go from, um, good ideas through the completion. And to explore a wider range of them is increasingly now getting more, uh, easier. And then that this is getting automated for more, too is on personalization.

And today there's a multitude of choice, right?

Personalization, Uh uh of at a scale that is now starts to be possible.

And, um uh I think the interesting part of that that starts to also start to break down some barriers. In my opinion, is concerns around, uh, ethics, Uh, diversity data sets.

What of these are models trained only on data in the US?

Uh, and I actually think personalization can start to break that frontier.

And, uh, in my opinion, a I there starts to become a lot more personal, Uh, with respect to your personal A i and other A I That starts to then protect the knowledge of the creators and the curators because it's their own in that economy and how they can work together. So for founders here, I think the key part here is thinking about the service.

And how can you enable someone to just get the outcome that they want?

Today, people go out and hire various employees to do different things. And what if, uh, it's almost like a BPO type model that Indian companies have gone in there and said, Hey, we've got this whole building. I'll move the whole building over, run the processes and all of it.

And, uh, I think in the globalisation play here, uh, enabling, uh, an entire function and using a I first to do it. And those who figure out the balance of this human in the loop with this and how it matures, I think is key, but picking a few segments where you can deliver an outcome really well, literally the Amazon book equivalent.

If you're a start up picking that, that kind of a category.

And if you're an enterprise, uh, figuring figuring out the access to data sets that you have, uh, leverage and levels that you have combining that with areas of the market that are seeing traction to optimise for it are I think the two things to, uh, look for what would be a guess for you on what the book like category is today.

Uh, to me, it's the creator, uh, economy today, uh, for the next 12 months, just given that, um, with JA, I, uh uh, it's solving.

Well, for creators today is giving them a plethora of choices.

Uh, but to go from there to getting something done is still, uh, quantum of work. But this category is somewhat clearer.

Uh, where you're selling into usually the units of a company Where, uh, there is some level of uh uh, tolerance for inaccuracy. As long as the curator and the creators involved are able to look at that, uh, with, uh, leveraging how Gene broadens the opportunity while, uh, leveraging human capability and capacity in a nonthreatening way where you're really up levelling human ability there.

And, uh, I I see that as being very tangible.

It it it, uh, up levels, human capacity.

It, uh, improves business outcomes and customer experience, uh, respect to who's involved.

And, uh, I see that as one category.

And, uh, the second is things involving automation where, uh, people are looking at old PC and new PC How these come together and bringing that you would go for, uh, the developer, um, like, you know, redefinition.

Like, you know, from a from a full stack engineer to an a i engineer where, uh, they are using things like cursor and clots on it and, like, you know, writing a lot of code. You you you don't see that as, uh, in a very similar way, as is the curator.

Uh, opportunity. I think the curator and the generative area is like a natural fit.

Your thoughts on developer?

No, you're right.

No, There's a natural fit between the two, and there's significant overlap between the two. I think the the only difference to me, though, is, uh, violin underlying stack. Uh uh. There is overlap, uh, from a business outcome perspective, there's slightly higher tolerance at one end.

Where, um if the creator that goes to you telling you about Hey, uh, oa has got a new, uh, podcast or this thing that's coming up, it's slightly off. There's a bit more of tolerance to that versus a developer who, uh, ship software, which suddenly brings a bunch of servers around the world down.

Right or so?

Um so I think, uh, that balance is important. So on on for the next 12 months, I think the creator part is a little clearer.

Uh, there's no doubt that the stack and as it matures, will evolve how we build software. But we also have to keep in mind that building software today touches many parts of the stack right at a very fundamental level. So underlying infrastructure operations Uh uh, data centre, security and all of that.

And there's actually an opportunity there, which people aren't looking at as much uh, it's a little unsexy, but to look at the more boring areas of the stack on, uh, quality monitoring, reliability, Uh, and all of this and one example I tend to think of sometimes even like browser stack, like one of the start ups that em out of India. And it took a fairly bro boring problem of, uh, browser compatibility.

It solved it incredibly well. So I think there's an opportunity there. I also think a little eye labs, right. At a time where a number of people were building mobile apps, little labs came out with a building block that, uh, where everyone else was digging for gold.

They built a spade, right?

And, um, sometimes, uh, being able to solve for platform infra that enables an A I gold rush might also be a good place to be.

Yeah, then, like, you know, for for for an Indian founder that you need, uh, you need to be in places where the stacks are layering and coming together, right?

So if you you earlier spoke about like in an Indian founder's opportunity, as in making sure that they access distribution, uh, and distribution needs like, you know, global presence.

And, uh, if you if you're focusing on building on the ST part of it, then you need to be in the right places.

So are you saying that these are two different sets of opportunities?

One is on the application side and the other one on the infra side.

In fact, you know, what I was gonna say is that I love the boring framing I.

I say that anybody even ask when someone asked me, Uh, what are you focusing on?

I say vertical boarding applications A I applications. Right.

So, uh, where, like, you know, others are not coming. And incumbents are not spending time. And therefore you can find a space or a spa. You can go and build a solution and then, you know, become really big before the incumbent catches up. And that's the only way to sort of beat incumbents into building large businesses.

Otherwise, if you are in a sexy space, then a hyper scalar could come and say I will make this infrastructure component, uh, available for free, or they'll throw like, you know, billions of dollars on GTM and then can crush a start up.

So how does the How does the start up pick right?

You know the application or on the infra site?

Yeah, You're right.

Um, so this new PC is kind of coming together, right?

Uh, I tend to think of the, uh, large language models is almost like a chip, right?

And the chips themselves. There are many types of chips, some of the commodity, many different variants now in place.

Uh, we're starting the evolution of, uh, short term memory, long term memory coming together, short term memory with various options rag and so on And how this is getting managed.

And, uh, the pieces around the new PC are coming together. There's definitely opportunities there in that stack. There are challenges with respect to go to market partnerships, distribution for an Indian founder sitting here. You build something, and, yes, it could be an item there.

Um, it's a tricky choice, but, uh, III I think the since we have seen examples both in the valley as well as in India of some of those start ups getting picked up, um, there will be choices larger companies have of solving those problems while solving broader problems.

And, uh, I think there are opportunities is there. But the part isn't necessarily, uh, entirely clear.

Um, at a higher order layer, the challenge is while this new PC is still coming together, what problems can you solve?

Well, uh, and I think both matter and both will continue to evolve.

Uh, so this whole, uh, old versus new If I were to sort of shift the gear and look at it through the lens of a product manager, right?

How does things change for them?

Um, how do they have to think about, like, you know, building products.

You know, what's the role of, um, gathering requirements that, like you said, like having customer empathy?

And then what is the role of user experience?

And, like, you know, how does this paradigm this platform should change paradigm for product managers.

Yeah, No great question.

Um, so, like we saw, I think with early days of personal computing, early days of a lot of these shifts.

Uh, early on, these things are very expensive, right?

So the cost of, uh uh a I at this point is still very high. Some of this will go down over time.

Uh, but given the cost of a IJA I overall, um, making, uh, very careful choices about where you can deliver outcomes and, uh, where value lies.

Uh, what the proof point is and how you can show measurable progress, uh, is important.

So anchoring, uh, on what the business outcomes are, how you will measure it anchoring, uh, in close partnership with, uh uh, your co-founders or technical, uh, leaders in engineering to be very pragmatic about what can be solved well, today And that definition, uh, both from outcome perspective, requirement, perspective, and seeing it all the way through, uh, and sequencing of, uh, how you get to those markers that will guide it, uh, are the hard choices if you go back to the Amazon example So anyone can put up a website that lists a bunch of books?

Uh, one part of the complexity is ease of experience. The UX making it easy to come in, find books that they want, uh, the catalogue broad enough.

Uh, am I sending it to the right people?

When they come in, do they find what they want?

Are they able to search?

Are they able to pick going through the ordering flow?

But then all the other operational mechanics, too, Where this alone doesn't solve the entire problem. It comes down to, uh, sourcing logistics delivery replacements.

Right?

So, thinking of it end to end and, uh, figuring out for the category that you're going after, even if it's narrow with a broader vision. How you solve for that end to end, I think is key.

So with JA, I given the current limitations that it has, what can it solve?

Well, today And how can you land in the market with something differentiating where In the middle of all the noise, you still have something where the value prop is clear. Someone who tries it gets to the point of value, and then you're able to expand from there, irrespective of which category. I think this is key, uh, whether you're a start up or an enterprise to get very, very clear on this.

And, um, don't forget the fact that, uh, access to data and, uh, it's still a very key pivotal, uh, factor here, uh, from a strategic level standpoint, so are we.

Are you investing in general or investing in the space?

And if you're investing, how do you actually think about, uh, opportunities.

So for me, I'm actually mainly looking at the unsexy, Uh, and the reason being I, I think these had the most, uh, longer term leverage.

Uh uh, at this point.

And, um, I think when you look at a 34, year horizon, it's a lot of those unsexy areas that I think will start to matter more. I think on a 1224 month horizon, we'll see various ranges of applications and then we'll see a way where a lot of these will consolidate.

Uh, a number of these, though, will, uh, start to be breakthrough companies.

Uh, but that will be a very small percentage, but I think, uh uh, the key when a new technology emerges to see through towards the business outcomes, that matter is shaping of a lot of those little unsexy areas and how they come together, because without it, the promise will not be met.

And, um, I see those areas of when you have 100,000 different options that could be put out.

Uh, what represents good quality?

What represents good experience?

Uh uh. What are all the pieces around it that need to enable that last mile to come through. It's in that unsexy that I think a lot of opportunity lies while a lot more of attention today is placed on that.

How do you do that?

90% with an application. And I do think that that matters as well. And whoever lands that part will own the distribution, the marketplace, the central points of a platform that have leverage. But I also think that this part, uh, could have significant strategic leverage depending on, uh, how things go. So I think both are relevant.

How do you think about the more question?

Is it even relevant?

I, uh, to be honest with you, start up. I don't have any good answers here for this.

Um, I think the speed of change we've seen in the last few years has been so much that, uh, who really has any answers. We thought it as LL MS LLM is a commodity today.

Uh, NVIDIA is the only one who has every other aspect of the S A. There are so many pieces that, um uh, to me, this it it's it's, um it feels like access to data, data sets and LL MS so is a commodity.

And, um, if you ask me, I think it's it's who owns the relationship with the end consumer. That has most leverage because I foresee a future where, as, uh, job shifts start to happen, uh, people will protect their knowledge. People will protect their abilities. People will protect their capacities that exceed a I.

And when they start to protect their capacities that exceed a I, it will be, uh, some version of a personal A I That will be the interface and how that gets regulated.

And, uh, it's it's just human nature. And I think those boundaries will start to, uh, form.

Uh, So those boundaries and the boundaries at a geopolitical level, I think are two that, uh, may, uh, affect, uh, the moat question a little bit.

But, uh, what do I know?

How do you keep up with all the changes in a I?

Who do you read?

What do you read?

Where do you hang out?

I actually don't keep up that much. To be honest with you, I, I mostly like I'd say, like, once every couple of weeks.

I, I try not to uh, I'm not too active on most social media I every couple of weeks, I look at research papers. I look to look at, uh, start ups that are gaining traction. I look at enterprises and broader strategies that they have, uh, across the different sources.

I have a a local setup, which does, uh uh, I have an agentic framework in place with multiple agents for me that do a lot of this research a lot of the curation. And they serve up for me, Uh, information that, uh, I'm looking for so in sort of a little a I doing a bit of research for me and serves it up periodically.

Tell me a little bit more about that agentic workflow.

And like, uh, what does it mean?

And why is why?

Why did you pick that?

Um, it's part of just tinkering to stay close to the latest of what a I can do today.

Uh, for me is also just, you know, uh, adapting the latest products, staying at the bleeding edge and continue to use them to also know where we are and what is real.

And, uh so yeah, so I have, uh um uh an agentic framework. Sort of running personally where it mines data from, uh, Twitter, different sources. Looks at different research papers, curates it, puts it together, uh, tracks those, uh, colleagues and summarises that for me. And that's kind of mostly what I read. And then I use that to decide if I want to dig a little further.

So what are some of your favourite A I tools that you use in our day to day life.

Uh, so, uh, Cisco's A I products are amazing.

And, uh, those are what I use day to day, uh, cross Cisco's platforms, things that are woven into all the software that we build and use.

Which ones do you use personally?

So Cisco A I assistant today, uh, is integrated across the entire portfolio across plunk across WebEx and, uh, use it on a day to day basis for crafting messages that I send, uh, use it to look at summaries, action items, various things that I need to do every day.

And then at a personal level, I, uh I just play around with the late you know, the newer frameworks that are in places the newer models, uh, experiment with little things that I do even for my daughter. So for her birthday, I use JA I to, uh, make a little game for her, the little playing card game. And I asked her for the favourite thing she likes to do.

And I came up with a little Sunday fun thing where I use JA I to come up with little visual cards, like little playing cards with pictures of like, you know, like you're going to science museum or you're jumping in muddy puddles or you're you know, we're going, uh, hiking together or teaching a dog a new trick and, uh, generate them into, like, about, uh, one card for every Sunday of the year and put those together.

So I keep trying to use it just for, uh, curiosity and to know what's possible and applying it just to day to day things that I do, uh, outside of work, uh, as a maker creator and figuring out how ideas come to life. It's another way of staying in touch. Awesome.

Anything that you think that we should have talked about in this episode that we didn't cover yet comes to your mind?

Yes, I think, uh, the only thing that comes to my mind is, uh, a little bit of data and, uh, geopolitics, sovereignty, Uh, and some of that. And I think that, uh, continues setting be an open, uh, unclear question. Sometimes the start ups.

You think?

Hey, um uh, there's always this question of Was it a need for a WhatsApp for India, Right. And many starts. Think that.

And would you want to build products for India or not?

Um, and I said twitter for India or the Facebook of India, Right?

Right. Yeah. The Facebook and Twitter. Exactly. Right. So I think there's a lot of that open question.

Um, I think potentially, uh, the opportunity that shifts us a little bit is when you think a I is a service.

Uh, because access to large data sets access to people who can drive those outcomes.

Uh, and helping you to find your product, uh, can enable it, But price points are a factor. That's the one thing that gets in the way a little bit there.

Um, but that part, uh, that ability to run that loop end to end where you have access to large data sets access to being able to deliver those outcomes. End to end, uh, with human in the loop for labeler Curators, creators, you name it.

Uh, you actually have access here today, which can then, uh, let you not have to worry about some of the other, uh uh, complexities. But it means investors who are willing to look at this with the lens of, um there is an opportunity to incubate Grow this here, and then you could take it anywhere.

Um, And just the fact that, given the diversity of data, there can be opportunities to have and a I calling for India just because the data sets are completely different. So I think there's one part we didn't touch upon, but I, I think that's one which I suspect that enough people are already having conversations about awesome. Awesome.

Avi, this is, uh, such a great conversation. Thank you for taking the time.

And, uh, please thank you for being, uh, on this podcast. Thank you so much for having me always a pleasure. Awesome.

Latest Podcasts
Ep
1
What's Real in AGI: A Conversation with Peter Voss
November 1, 2024
36:15 Min
Ep
2
How AI is Actually Changing Accounting with Dr Mfon Akpan
October 30, 2024
39:14 Min
Ep
3
AI's Leverage - Timing, Data and Human Relationship: Insights frm Aravind Krishnaswamy
November 1, 2024
42:19 Min
Ep
4
The Surprising Upside of AI Regulation with Maya Sherman
October 30, 2024
36:07 Min
Ep
5
Robotics Inflection Point: Arshad Hisham on The AI-Robotics Stack & Traditional Labor Markets
November 4, 2024
28:52 Min
Ep
6
When Laws Meet AI, with Laura Carr
November 19, 2024
19:10

Get Connected with Upekkha

Want to understand if Upekkha is right for you? or Get AI SaaS Report