← The Digital Executive
The Digital Executive
Krishna Gupta on the Future of Vertical AI and Voice Automation | Ep 1076
00:11:35
Loading summary
Transcript13 lines
- [00:00]
A
Foreign welcome to Coruscant Technologies, home of the Digital Executive podcast. Welcome to the Digital Executive. Today's guest is Krishna Gupta. Krishna Gupta is an AI founder and investor. He is the co founder of Presto, which is the leader of using voice AI to automate restaurant drive throughs. He also founded his venture capital firm, Remus Capital while an undergrad at MIT and has a deep focus around vertical AI with a particular emphasis around applying voice AI and the industries that power America. He's been investing in voice AI for the last 15 years. Krishna received his bachelor's degree in engineering and management from mit. Well, good afternoon, Krishna. Welcome to the show.
- [00:50]
B
Thank you. Great to be here.
- [00:51]
A
Absolutely. My friend. I appreciate it. You're hailing out of that Cambridge, Boston area in Massachusetts. I'm in Kansas City, so just an hour apart today, but I appreciate you making the time. And Christian, I'm going to jump right into your first question. Your journey began with Remus Capital. While still an undergrad. What drew you to invest in AI so early and how did that lead to founding Presto?
- [01:14]
B
Yeah, look, I grew up in a world. My mother was a computer science grad and she had studied AI back in the seven, you know, back in the 80s. And so I grew up sort of learning or hearing about Marvin Minsky and that approach to AI. So it was definitely not unfamiliar to me. And while I was at mit, there was a renaissance of machine learning happening. And so, you know, I was reading machine learning papers and pretty quickly realized the power of broadly big data and the ability to algorithmically play on big data, to create insights that could be beneficial to man and machine. And so that's how we started investing in applied AI. And as I was doing that, one of my first, in fact my very first investment was Presto, which I effectively put in business and co founded. And so that's sort of how I got involved with Presto and with AI investing.
- [02:07]
A
It's amazing. I appreciate that, of course you had a big influence from your mother, you know, being in computer science. I think that's pretty cool how our parents kind of shape and form, you know, our career trajectory sometimes. But I like how you got into machine learning and, and big data. I think that was important and that also had a big influence in what you did. But I did not know you invested into Presto and that's how you became part of that organization. So appreciate the backstory on that. And Krishna, Presto pioneered voice AI for restaurant drive throughs. What unique challenges and opportunities come with applying AI in Such a high volume, fast paced environment.
- [02:44]
B
Yeah, well, look, I think there had been attempts to apply, you know, just purely natural language processing to the drive thru prior to the LLM and the Transformer revolution. I think once that came about in 2020, we realized that there was a really big opportunity to leverage this new wave of technology and AI to be able to create a product that might actually work. And you're spot on. It's a very challenging product to make work because you have a lot of noise. So it's a very kind of classically noisy environment and the noise is different from place to place. So you know, whether you're in Kansas City or Boston or in a suburb or in a city like all of these different settings means that you've got different backgrounds to deal with. And so that's one part of what makes it hard. The other part is that this is an industry, the restaurant industry, that is very, it's full of unstructured data and a messy matrix of integrations and you have to sort of be able to do all those integrations into the menus, into the loyalty system, into the POS system, et cetera, while dealing with the complications of outside noise and then being able to deliver a highly accurate order in a very time constrained manner because you can't be sitting there for three minutes taking an order. You've got to do it in 40 seconds and then do it again in 40 seconds and again in 40 seconds. And you've got to do all of this while ensuring accuracy and contextual relevance. So first of all, this ended up being the first verticalized application of voice AI in the LLM era. And it also, because it makes so much sense and the ROI is so clear, but then it also ends up being one of the first places where I recognize you've got to use a neuro symbolic approach, you've got to marry LLMs with NLP. And in fact, in the early days we had two different teams, one working on LLMs, one working on NLP approach to this. We married the two and I think the marriage is what makes this work. There are companies out there that only have tried the LLM approach or only have tried the NLP approach, and neither one really works. So marrying the two on the AI side and then ensuring that you've got hum in the loop in the background as you get started is what creates a product that delivers what the customer wants, which is ultimately what matters.
- [04:58]
A
Thank you, I appreciate that. I just know that this industry, we've talked a lot about this on the podcast restaurant industry is particularly the fast pacedness of the industry. What I really took away from this is you marrying that LLM and nlp, right? That those large language models and that natural language processing and you were successful at that and I think that's really cool. Obviously there's always got to be a human in the loop and hopefully we'll always have a human in the loop as fast as the technology is advancing today. I appreciate the information and insights. Krishna, you've spoken about the advantages of vertical AI over general purpose AI. Can you break down what makes vertical AI a more immediate and ROI driven opportunity for industries like QSRs, logistics or manufacturing?
- [05:44]
B
So, you know, I've always been focused on vertical technology and applying, you know, modern technology and AI to verticalized applications. I would say that it's not always the case that a vertical application is better than a horizontal one. In certain, you know, applications and certain industries, perhaps a horizontal one will be better. But when it comes to certain industries and restaurants is one of them, a verticalized approach is necessary. And that's for three reasons. One is what I mentioned earlier, this mess of unstructured data that's out there and the matrix of integrations, etc. Etc. The second is that you need to be able to have deep understanding of the industry to be able to build the product so that the ROI matches the industry. So in each industry you may have a different ratio of ROI between labor savings, revenue optimization and customer experience optimization. And for the, for the restaurant industry, even at a franchisee level and a state level, that mix of ROI is different. So without that deep understanding of the industry, it's really easy to speak in extract and very challenging to deliver a product that works on the ground consistently and uniformly across the country. And then the third part is go to market. You know, you can build a great product, but if you can't, basically convince customers on the ground to use it, bring an operations team to implement it, a customer success team to make sure it continues working. You won't be able to be successful as a company. And those in certain industries, those are highly verticalized relationships. You've got to spend your time going to franchisee meetings. I just flew to la, you know, on Thursday, literally for five hours I went back and forth just to have lunch with a large franchisee customer. If you're not focused like that, you're not going to win. So Those are the three reasons for which certain verticals like QSRs, you need a verticalized AI approach.
- [07:31]
A
I appreciate that and obviously you need to maximize every angle, every efficiency to gain that success by really specifically knowing that business inside and out. You did highlight a few things. You, you know, I know you're particularly focused on leveraging AI in these, as you say, verticalized approach, but it is challenging, unstructured data. You know, every industry obviously has unique roi, but that's why you focused specifically in this area. Again, last thing I took away was obviously your go to market, right? You may have a great product, but you got to have that customer buy in at the ground level. So I appreciate that. And Krishna, last question of the day. Looking ahead, how do you see the vertical AI ecosystem evolving over the next five years and what role do you hope Presto and Remus Capital will play shaping that future?
- [08:16]
B
Great question. I think, you know, a lot of these verticals are starting to what I call instrument their companies with the first order of AI. So let's take QSRs for example. I often talk to our customers. Hey, our voice AI is what enables you to start instrumenting your stores with an AI native approach where that leads. And the product roadmap of these platforms like ours is going to be very different. Right. So we start thinking about an end to end AI connected platform. Starts with voice AI, but incorporates things like computer vision and robotics, ensures very real time management of these companies. I've always believed that one of the challenges in any industry is that the lag time between getting data back on how you're performing at, on the ground level, let's say at a store level, at a customer level, and then being able to actually make changes and tweaks in your business based on that data and then seeing that whole feedback loop again. I mean the lag is way too long for you to actually have any real chance of managing the success. But what AI will enable us to do is turn those lags into a much more real time dynamic thing. And you'll be getting data from the voice AI, you'll be getting data from computer vision and you'll be able to feed it into robotics. And so I anticipate in every one of these industries to have, you know, perfect visibility or near perfect visibility into the business, which enables kind of an automated way of having the businesses run. So the other thing is a lot of these actions that a business might take based on certain data. So for example, if I, if I have voice AI running in a drive through, I'm going to start learning that certain specials, I can test certain specials and in real time get feedback whether customers like them or not in this location or that location. Then what do I do based on that? Well, right now you've got a very lean management team that has to take that data and then cut it, slice it and dice it, and then get into some meetings and then make decisions based on those meetings. Then tell a bunch of people to go implement those decisions, then measure those decisions. Right. I can effectively automate, at least the vision here is that we can automate that end to end. And so now the management team can actually focus on what's important without getting in the way of this dynamic management style. So that's how I look at every vertical. That's restaurant. You know, we have Voice AI companies in home services, in healthcare and financial services. I look at Voice as being a powerful first adoption tool for large industries to start instrumenting their companies with AI.
- [10:47]
A
I think that's amazing and you really hit the nail on the head. I really like this AI native approach that is truly foundational to the business success, what you're doing to help these businesses. Going back the current way we talk about providing feedback, right. We're way out of touch at this point because as you said, humans are involved. There's an analytical piece of it, measurement, feedback, decision making. And that is way too long. That feedback loop needs to be real time at this point, as you mentioned. And I know we can do that, leveraging the power of AI. So I appreciate that. Krishna, it was such a pleasure having you on today and I look forward to speaking with you real soon.
- [11:23]
B
Thanks so much, Brian. A great way to start my day.
- [11:26]
A
Bye for now.