Leveraging Adaptive AI For Sales With Jags Kandasamy
AI has been helping companies boost sales because it can augment the sales process of a business. Chad Burmeister’s guest for this episode talks then discusses what adaptive AI means for sales.
Jags Kandasamy is the co-founder and CEO at Latent AI. In this insightful conversation, he shares insights about artificial intelligence, technology, automated emails, lead generation tools, and the best practices his company executes in AI for sales. They also delve into the challenges and strategies in becoming an operator for optimization.
Listen to the podcast here:
Leveraging Adaptive AI For Sales With Jags Kandasamy
I've got a gentleman with me from the Bay Area, which is probably more than half of the folks that are on my show. Jags Kandasamy is with Latent AI. We'll get into a little bit more about what it means and how it works. Think of it as you wouldn't want a big heavy AI system running on your iPhone, on a refrigerator, or on something like that but you want the heavy AI running in the cloud. Latent does its thing on the back and we'll dive into that. Jags, welcome to the show.
Thanks, Chad. Thanks for having me. I'm excited to be here.
I love talking about this topic because even though it's built for how sales people leverage artificial intelligence, it starts to open up people's minds on what's possible and where the industry is headed when we talk about AI for transportation, AI for Amazon Alexa, all of the different areas. I'm excited to drill in. Before we go into artificial intelligence, I like to let our audience get to know you the founder and the CEO. I like to go back and ask the question, where were you raised? What part of the world? What were you passionate about? When you were younger, did you know you'd be running Latent AI or perhaps not?
I grew up in a Southern Indian City called Coimbatore. It's known as the Manchester of South India because it's famous for its cotton mills and fabrications. T-shirts and stuff were made there. The spinning mills were there as well. Growing up, I was lucky to go to a private school. I was the sixth youngest family member on a single income family. My dad was a cop. I was lucky to find myself into a private school and had a lot of who's who in the city kids studying with me. From then on, the entrepreneurship bug bit me. I was passionate about business. I was a hustler throughout school and college. I used to carry hard disks back and forth from the dorms to format them, load them with software, and take it back to my friends and customers then right in college. I was a hustler. I always knew that I would be running a business. I didn't know what it was. I did many things. I ran a restaurant in Atlanta for a bit. I did some export-import stuff. I did some medical equipment trading. Now I'm into the technology space.
I saw my niece on a trip to Arizona. She's getting into the hotel hospitality business. I've had interesting interactions with sales operations folks that the level of operating experience that they learned from the restaurant management is solid. You have to be very good. What is it about restaurant that causes you to become a good operator?
You have to think wearing multiple hats. There are days that you're spending time in the front office. Then there are times you're in the kitchen. Sometimes when the dishwasher doesn't show up, you're there. When you are there standing in the dishwasher shoes, trying to get the line moving, you understand the difficulties they face, the challenges that they have and what are the things that you can optimize for them so they can do their job easier and be efficient as well, things like that.
Attention to detail and helping people unstick their production line is essential to succeed.
It's note to sales leadership. Anybody who's reading, if you haven't looked at former restaurant hospitality management folks, I highly encourage you to do it, especially SDR managers. I've seen a lot of success. It's attention to detail and helping people unstick their production line. "I don't have enough leads. I don't have the phone thing to make the dials properly," all of those concerns, the restaurant manager would help them through that, so that's pretty cool.
Including objection handling. Sometimes there are objections that get thrown at you that look like a complete left field but if you're sitting there or if you think like them, you would know how to answer it on a rapid fashion.
I heard hustler was the term. I certainly was a hustler. I would sell suckers on the bus to kids when I was a young kid. I learned the trick. Let's say my cost was $0.10 a sucker, I would normally sell them for $0.25. The margin was pretty good but by the time you get to the end and you're down to scarcity of the last couple, I would get as much as $0.50 to $1 for those two.
It's supply and demand.
Let's pivot into AI. Latent AI has AI in the name. I'm starting to see more companies that are focused on AI. The .ai URL is a newer thing in the last several years. Our audience is probably more business folks. SDK might be a new term for them. Tell us more about how AI is leveraging your product.
I'll draw a parallel to the gold rush in San Francisco. I'm here, so I'll draw. If you remember, Levi was one of the most successful companies that came out of the gold rush. They didn't do anything with gold itself but they were the provision store for all the gold rush people, all the people that went North or East to go dig. They were the provision supplier. They were providing the picks, shovels, and other things to support. What Latent does is in a similar way. We do use some AI in our technologies. We are providing tools and platform to help other AI developers to build efficient models so that they can deploy them into energy efficient processes. We've come up with this terminology is called the edge continuum.
Adaptive AI: Salespeople leverage AI as it starts to open up people's minds on what's possible and where the industry is headed.
Right from the center level, let's say a speaker, a microphone on your phone, or a heat sensor in a factory across different gateways, different devices all the way into the cloud. By the time you type something, it goes up, there are multiple devices along the way. How do we put intelligence along that pipeline efficiently? That's what we do and provide. We are providing tools. Think of us like spanners, screwdrivers, and other stuff that makes your software work. Making the job of the developer easier so they're not spinning wheels spending nine months building something where they can just call in with two lines of code, they can call our tools and be done with it.
Is it more for the energy efficiency side of the equation or does it also have to do with all other kinds of complex algorithms and software decisions?
There are multiple things that go into factor. Size is a major thing. Probably folks have heard about the memory size, gigabytes and megabytes. Before, it used to be megabyte, your computers are now 8 gigs, 16 gigs. There are processes that have one gig. There is the cache memory where the processor works on things. You cannot put too much stuff in there. Your process is going to overheat or it's going to slow down. How do you efficiently manage the size of the memory in processing? That's one. Second, in relation to that comes the power usage. If you have too much computing to do, you're going to waste power. If you are running on a battery, how can you push that battery for a longer life? The third one is speed. A simple example, you talk to a Siri or an Alexa. If it takes about 1.5 seconds for it to reply, you feel like the conversation is not natural. There are so many other use cases like this. Time is critical and it needs to happen locally.
If I was driving and my exit's a half mile away and I say, "Siri, do I exit here?" Four minutes later, "Yes. It's your turn off.” I could have used that because my router was plugged into a car. The audio was gone, so I didn't have that verbal cue. I couldn't hear it here. It's interesting. Adaptive AI, I've heard that term before. What does that mean?
I'll give a simple example for this. You all heard of the doorbell cameras, ring, blink. All of that is based on motion sensing. If a bee comes in front of your camera, you will get a notification. If a postman comes, you'll get a notification. If a bird flies by, you'll get a notification. We can solve that by putting a human detection model. We've compressed it into a very small size that can fit on that camera and run on an AA battery for two years. It's always on. It only wakes up when there is a human present. You can do that. That's based on AI.
You want to know who's on that scene where you need to employ facial recognition. You cannot run a facial recognition algorithm on that camera because it's has very small processors. Laws of physics cannot prevent you from doing that. You have to run that at a different layer of computing. It could be a gateway, a set-top box in your home, a CDN in the cloud, or a Telco edge, whatever that network side of things. You can run that. How do you connect this and that? That is Adaptive AI. How do you make sure that you are passing the right information and getting the right inference?
Incorporating AI in business enables tools for salespeople to build models.
When I was in Arizona, my ring doorbell notified me. For the first time, I noticed it said, "A person is in your backyard." I was like, "They've upgraded." It used to just say something other than a person. Now it can tell a person. Similarly, I said, "When's it going to get to the point where it says your wife's in the backyard, your son or your daughter with the facial recognition PC?" I was happy to see it move from a bee to a person.
Probably they are still doing that in the cloud. As soon as a motion is sensed, they capture the frame, send it to the cloud, and then they say, "There's a person there,” and then they send you the notification. With our technology, you can do that person detection, animal detection, or car detection right on the camera. I don't even have to go to the cloud. That is Adaptive AI.
It reminds me of RingCentral and Zoom, where a lot of people used to have a lot of equipment in their closet to manage the phone. Now it's all managed via the web. You're taking that piece of it out and connecting it directly to the phone. That's outstanding. We talked a little bit about use cases. One is a very top secret use case and another one is automotive OEM. We probably can't talk about the top secret one but suffice it to say Latent AI is being used at some pretty top secret kinds of places and we'll just leave it at that. When it comes to automotive or any of the other non-top secret use cases, share 1 or 2 of those ideas.
We have a customer that's an automotive OEM. They want to put object detection capability on their rearview camera. When you put your car or your truck in the reverse mode, you get the screen. You're watching what's going on but there are times that you run into objects that you didn't see or didn't appear in there. How do we detect those objects and enable Automated Driver Assist System or what we call as ADAS? The problem is with every automotive company, now the semiconductor shortage is high, but even before that, they try to optimize for cost. What is the lowest processor that I can put there and get the best benefit from it, best bang for the buck? They picked a processor and they wanted to run at a certain speed. You don't want to be notified about your exit five minutes later. It's the same thing if you're backing up into things, you don't want to be notified after you ran over it.
I ran into the trash can in the driveway. It's a green trash can. Usually, I would think it would beep but it wasn't beeping.
Those are the things that the camera would be able to detect and provide information to automatically break. Now we have the breaking from a frontal collision perspective in certain high-end cars so we're trying to get this to the lower-end cars with the low power process. We are providing tooling to the automotive customer, so they can build models. They can deploy on this process.
Adaptive AI: Objection handling is important. Sometimes some objections get thrown at you, and you have to respond in the right manner.
We're talking a lot about physical appliances kind of insights or connectivity. Is there anything where you would help a software company do what they do or does it usually involve a device or a physical component?
There is software enablement as well. I'm drawing a blank on an immediate use case but think of us being able to run certain inferences on your browser. We can enable inferences to be run on JavaScript, on the browser and also to be run on the CDN. Salespeople will understand this very clearly. There are several companies that do natural language analysis. As you are interacting with your prospect, you want to get the right queue at the right time of like, "He's getting disinterested. He's moving away from this topic. I need to get his attention back." All of this was being done remotely in a data center in the cloud and it takes about 30, 40 milliseconds to come back to you. You already moved onto a next topic but what did appear right there on your screen as you were talking? Simple thing like closed captioning system. Close captioning is delayed. By at least 20, 30 milliseconds, you get the delay. How do you reduce that and you get it instantly?
You should talk to a company called Barco. They do real-time conversation analysis. If I'm talking to you and I happen to say something or you say something about a competitor you're looking at, let's say, it will show you in near real-time what I should say back. The biggest challenge they face is that it could take 1, 2, 3 seconds. I’m having to shoot the breeze on something then it tells me what to say and how to say it. I bet you by cutting out that latency and they may have solved it already in a different way, I'm not sure, but there's probably something that this software could help them do in a better and faster way. From a sales perspective, a lot of times as a founder and as a hustler, I have to assume you're probably leveraging some kind of AI in your outreach process. Are you finding AI being used in sales?
Personally, we haven't gotten to that stage yet. We are still early stage in the company. I have a big network, so I've used that network to get my initial design customers and building on the product from there. I still have a lot more awaiting customers to test the product. We haven't gone fully into the deal or lead generation process. One thing that I'm very familiar with as we're getting ready and we will be starting to implement that is some of this voice recognition, voice analysis companies during a sales call. You've got Chorus, Gong and stuff like that where AI is helpful. I have been selling for years. I have an understanding of how to bring a customer along a journey and to buy.
There will be new folk that will be coming into the team in the inside sales or in SDR motion. How do I train them with some of my knowledge or bring the best practices to them and help them understand? They need to learn this on their own as well. You can't be handling them for too long and they will be waiting for the cue to come all the time. You don't want that to happen either. These guys need to graduate from inside sales to field sales. They're sitting in front of a customer. There is no going to be any prompt showing up here and there. They can't be looking at a phone. It's not going to be helpful.
Also, a little chip in their ear or something, maybe ten years from now.
Several companies do natural language analysis for effective interaction that leads to the right queue at the right time.
We could have coms like the agents do. Audio check type of thing, that somebody is helping them out there.
Some presidential candidates might get pumped into their ear while they're speaking.
Those are the things where AI is going to be helpful, I believe. For me, AI is not artificial intelligence. It's more of Augmented Intelligence for the human.
A lot of companies like ours start in automation, speeding things up, and doing the low value tasks at a higher velocity. There's a piece of code we purchased at the end of 2020 that analyzed, let's say you give me 100 prospects. I got to get a meeting with these 100. How do you know who in your network knows the 100? What the AI does is goes out and looks at entire network of first degree connections and second degree connections. It does one year worth of human work. I ran one before this call. It took 25 minutes. It said, "You have 1,143 pathways into these 242 companies." There's no way that I could have gone and clicked through all that in 25 minutes. It would take a year but now we can automate the email to go out and say, "Jags, I see you know this guy. Would you be okay to make an intro?" You said most of your prospects are coming through referrals. This helps you automate referrals much.
Years ago, I started a project, working on some lead generation stuff. It was a project like the joint. The fundamental concept behind that was if you are my customer working at company A, when you switch jobs from company A to company B and if company B is not my customer, you need to be on the top of my prospect list. I was working for a large company and we didn't have that system in place. Salesforce didn't have that in place. There were other CRMs out there that didn't have it. We thought we would solve it. You're going to use LinkedIn, job change notifications to populate that list. You have to use machine learning. We were getting into the early stages of machine learning then when we were shut down by LinkedIn because they said they were going to launch their own product on it, so they didn't give us access for it.
Hitesh Shah was the CTO of ConnectAndSell. He created something with the team that did the job hopper tracking as well. That was our second highest performing lead generation tool. It also creates another opportunity. Number one, it creates the op at where they landed, but number two, it creates an op at the vacancy that it just filled. They maybe a buyer of the product but there's a new contact. You can solve for churn and you can solve for upsell by going to the vacant spot. I forgot about that one. That might need to be another focus area for reinventing that product. Thinking about the future, last question is where do you think it's all headed from AI, especially on the sales side? What changes in a couple of years from now for salespeople?
One thing that I strongly believe in is that relationship sales will continue to surpass other types of sales but getting that relationship built, that's where AI will help and should help in the sense of if I cold emailed you, there is 99.99% of chance that you're going to spam me out and move on. The way you were talking about it, how do you find those connections? How do you find commonalities? "Jags isn't even interested in flying. He's interested in soccer and cricket. How do I make sure that I'm presenting things to him that are of interest to him or going through a reference with those that interest him in mind?" It's establishing that relationship and then taking the conversation from that.
Adaptive AI: AI goes out and looks at the entire network of first-degree connections and second-degree connections, and then do a year's worth of human work.
We have a partner that looks at your LinkedIn profile and your email address and then builds 12 to 15 personalization sentences based on that. There are always a few false positives. "Congrats on that local community college that you attended. It looks like a prestigious school." A lot of the reps were like, "If it was Harvard or something, yes but Arapahoe Community College here in Denver, you might not want to say that's prestigious." What's interesting is as the rep see it, humans and managers say, "You can't call that a prestigious school. I see you went to the local school, Arapahoe Community College," it gets better. What we're doing instead of serving all fifteen up or the best one, we're giving them three options. They go to outreach. They go send their email. Three personalization options are there. They delete the two they don't want or all three and create their own. By opening with that, your emails go through. Your open rates are higher. Your reply rates are higher. You build credibility and trust but the AI did the research.
Salespeople shouldn't get lazy. They cannot just trust the automation. How many spam generated emails come through on LinkedIn or through my email? I have not responded to a single one. They all end up and I blocked them for good measure. The ones that come with referrals and with the proper referral as well, "I saw you on this article. You talked about this. This resonates with me because of this. Can we have a chat?"
You've earned the right. You're different. You're a zero where everyone else is a one or vice versa. That's good. This has been a fabulous conversation. If someone wants to get ahold of you, Jags, how would they reach you?
LatentAI.com is our website and then they can email me at Jags@LatentAI.com.
Thank you for joining the show. It's a fabulous conversation. Congratulations on your successes. Continued success to you, Jags.
Thanks, Chad. I appreciate it.
Jags Kandasamy, Cofounder CEO of Latent AI. We'll catch you on the next episode. Thank you.
Important Links:
About Jags Kandasamy
"I do not accept cold connection requests, even if you have a great catchy line. I connect with people that I've met or spoken or referred by someone I know and trust." - Jags Kandasamy
Jags Kandasamy is a co-founder and CEO at Latent AI, an early stage venture spinout of SRI International, dedicated to building solutions that enable the adaptive edge to transform AI processing. Latent AI is well-funded by industry-leading investors with support from Fortune 500 clients. The Latent AI Efficient Inference Platform (LEIP™) brings AI to the edge by optimizing for compute, energy, and memory without requiring changes to existing AI/ML infrastructure and frameworks.
