Getting Ready for AI-Powered Everything - Srinath Sridhar - Revenue Today - Episode # 105

1, 2, 3, 4.

Sales, marketing, and RevOps. It's sink or swim out there, and yesterday's strategies and tactics won't help you today. This is Revenue Today, and I'm your host, Jared Robin. Join me as we interview revenue leaders in our community to learn what steps we could take right now to help you scale yourself and your company. Revenue Today is sponsored by Rev Genius, and we're on a mission to bring inspiration and creativity to all revenue professionals in the world. Want to shout out our sponsor Demandbase. Demandbase is smarter GTM for B2B brands. They help marketing and sales teams spot the juiciest opportunities earlier and progress them faster by injecting account intelligence into every step of the buyer journey and orchestrating every action. For more information about Demandbase, visit demandbase.com. Hello friends, I'm Jared Robin, host of Revenue Today, and I'm joined with Srinath Sridhar, the CEO and co-founder of Reggie.ai. Hey Jared, hey everyone, it's so great to have you, and I wanted to push to, to bring you on the show especially because we are now officially as of a couple weeks ago in the age of AI. And, and somebody like you, you've been living in this for so long, but as soon as the commoditized GPT-3 became available to the public use, things changed a bit, right?

Things changed a lot. Yeah, I mean, um, since it's hard to believe, first of all, that ChatGPT has been around just for 10 weeks now. Uh, that's pretty much mind-blowing. And GPT-3 itself has just been around for, I think, about January of 2021 is when it came out. At least fine-tuning came out in '21, so it's been out for 2 years and change. So yeah, it's unbelievable how much progress we've made as a community in the last 2 years and in the last 2 months even.

Well, I'm excited to dive in a bit because if I haven't learned anything recently, it's people are enthusiastic about this, but they're also very curious about what AI means to folks. So, you know, set the record straight. I'd like you to debunk some myths or one big myth that you've heard being said about AI.

Uh, sure. So Especially as it relates to us, let me talk about something that's AI for sales specific, although this actually applies broadly. Most people think that AI is gonna replace humans. You know, that's one of the first things that people think of, and that people think of any automation that way. And what, if you look back, what tends to happen is that people end up doing new things. Mm-hmm. As opposed to people being replaced by AI. So, in a specific example, although I'll try not to talk about Reggie too much, but specifically, you know, if you look at sales and personalization, people used to write templated emails and you had to do that because you couldn't write personalization at scale. So, it's not like AI is going to come in and replace your jobs, but what it's going to help you do is to be able to personalize all your communication as opposed to writing in templates. So, the big thing is, if at all, any workforce usually expands, then contracts. So, that's one typical thing that people get worried about. And if you look back in history, that's always been the case. It's more expansion because of automation and AI than compression.

But I hear all the time— so, there's some Is there any instances of AI taking jobs or does it just create new jobs? Or perhaps, uh, because, because it fixes inefficiencies, right?

Yeah, it fixes inefficiencies for sure. But at the same time, even, you know, let's take automation as a more general example. Let's say Outreach ended up automating sales emails in the sense that you could send one email and then Outreach could follow up by sending a bunch of emails. Before that, that would've been done manually. So you would think that, hey, Outreach can actually now write a bunch of emails in sequence and therefore the SDR jobs would go away. But actually what ended up happening is probably Outreach has been the biggest influence in terms of increasing the number of SDRs because you have one platform through which you can monitor all of your SDR activity and perform all of your SDR activity. It's sort of codified everything. So I think the same will happen with AI, which is definitely, it'll cut down on inefficiencies. So it's not like now that you're using Outreach, you're gonna go send follow-up emails one at a time yourself because you'll just use Outreach to do it. But at the same time, you have a system of record to do your phone calls and one-off emails and personalization in one platform. So same thing will happen here, which is sure, it won't be the case that You will have to send all the manual emails yourself and so on, but, or look up prospects, go to LinkedIn, read up about them, and all of that sort of stuff can be automated. But you will still have one platform to view all the AI sort of activity, whether that's personalization or sequencing. And SDRs can all live there and work out of there. So go ahead.

Let's unpack this a bit. Like, like how will it change the way that we sell? You touched on personalization, but let's, Let's dive into different ways because I know Reggie in particular is a writing tool, but is it gonna, is it gonna just be that like with AI as a whole? Like, what do you see?

Yeah, great question. So it's definitely a writing tool, but for us, it's writing tool for sales, right? So we're not writing in the business of writing novels or in the business of writing YouTube descriptions or Amazon reviews or any of that. For us, writing means Writing as it relates to sales teams. And for us, for the most part, it means writing as it relates to writing emails right now. In the future, we could expand into writing sales collateral as well— case studies, testimonials, that sort of stuff.

But what about like AI in general in sales? Like what else are you seeing like outside of your capabilities and your roadmap?

Yeah, great question. So for example, I mean, here's a cool idea that anybody can go implement if they wanted to. I don't believe anybody's doing it today, but if you wanted to do personalization, we can actually do personalized voicemail drops as an example, right? I mean, people do calls and people drop voicemails.

Yeah.

But it's just generic voicemails right now, but doing a personalized voicemail drop is very doable. So now you get into this modality of doing voice communication on top of emails. Same is true with images.

So it's going to take our voice.

Yeah.

And change it based on the person, the context.

Exactly. The prospect, what is the prospect's pain points? What is your value proposition as it relates to the prospect? What's the prospect's company doing? And drop the voicemail that's most relevant to them.

So let me ask the million-dollar question. Is this going to replace cold calling?

No, I, again, I don't think it's going to replace cold calling because first of all, even from a technology standpoint, I do think that it's a little bit far out. Um, but even today, if you look at, and it's the same analogy applies if you think about it. So today nobody leaves voicemails. They record the voicemail once and they use something like an outreach to drop the voicemail. You're not leaving actually voicemails every single time. But what you actually ended up doing is instead of dropping a standard templatized voicemail, now you can actually drop a personalized voicemail. That's what I think will change because at the end of the day, Google, Microsoft, everybody will come clamping down, FCC for that matter, on people sending thousands of automated emails. So, I do not see that being the future at all. There's no future in which that, that's going to happen.

So, you know, some of the points I've heard brought up is it could be, but do you have to announce yourself first? Like, hey, this is Jared's AI bot. And then, and then, um, I heard Elon Musk talking about the obvious, and, and we're not talking 2023,

right?

We're talking 2033. Or, well, you know what, a funny quote that I heard— not a funny, an interesting quote— is People overestimate what you could do in 1 year and they underestimate what you could do in 10 years. So this could be 2025, 2026.

And usually that underestimation happens because of technologies coming together, because you otherwise think of things in silos. You don't see the advances coming in. I don't know, telephones as an example. And that comes together along with advances in AI and sales and It's very hard to anticipate how things come together, which happens in multiple years. But I do believe that all the modalities will be impacted with AI. So, in that sense, there are startups like Tavis that does personalized videos already. And Hippo Video also does personalized videos. So videos is coming, uh, or it's already there. Uh, voicemails will be coming if it's not already there. Calls is probably a little bit further out. And then of course emails is coming, but I do think it won't be the way people anticipate it to be. It's not going to be a drop-in replacement for an SDR where you're actually going to do calls and emails. It's going to look completely different from that. And it's going to be higher quality. more personalized, more one-to-one communication rather than bots spewing lots of random stuff. It's actually, I think, going to be more human than more automated at the end of the day.

Nice. Well said. Yeah. And, you know, I just got off the phone with another technology that does call recording. You know, we've, we've, we've all seen that a dozen times plus. Transcribing, which we've seen just as many times.

Yep.

But giving, you know, notes to CRM, follow-up email.

That's right.

Based, based on the context. But then even more, you could ask the notes a question like, what was the most important thing to Srinath? And, or what, what was his title again?

Yep.

And it'll say CEO, assuming that he, he mentioned that in the call. So like the search functionality and, uh, and it's fascinating. Now GPT-3, people are getting it in their hands now.

Yep.

How could, how could somebody just like, how could a revenue team best leverage that tool in its rawness in the $20 a month if they get the upgrade or the regular to do some to, to, to add some efficiencies to the tasks they do without having to buy a software? I think that's kind of like a question that's going around now too.

Absolutely. So a few things that you can do out of the box that's extremely useful, you know, again, it's very weird how you think AI is going to do something, but it actually ends up doing something else. So I'm going to say something that none of us anticipated will happen, but it's the most obvious use case right now, which is, you know, we all thought in some ways creativity will be the last thing that AIs will touch and it'll take the mundane stuff and automate that away. And in fact, the opposite has happened right now, which is, let's say that you're staring at a blank screen and you're trying to figure out, okay, I need to write a blog and I need to write something about, how GPT-3 is transforming sales.

Yep.

But I have trouble getting started. ChatGPT is an excellent tool, and you can write that, ask it to generate a 100-word starting point, ask it to generate an outline, ask it to generate what is, what are different ways in which you can organize this content, and it'll do a fantastic job. Of course, it's up to you to Fill in the details, make sure it's correct, do the citations, all of that stuff. But funnily enough, we actually thought that AI would come in and fix the grammar and the citations, but it has actually gone the exact opposite route, which is it has actually helped you with the creativity side. And then you are now here trying to fix the grammar and the citations and all this stuff. So those are some simple ways. So let's say that you're writing a blog post, let's say you're writing a social post, Let's say you're doing some thought leadership. Again, it's not like it's going to spew content for you, but you should not have any problems getting started anymore. And that is a pretty big win. So anything around getting started and creativity side for long-form content out of the box, ChatGPT is fantastic.

I've also seen, I've been a part of this integrating it with Google Sheets.

That's right. And there are a whole bunch of plugins that help you integrate to Google Sheets. And that's very early stages, very exciting. Um, and I'm sure, you know, Google will just start putting this into their own, um, platforms, as will Microsoft and so on. Very exciting. Same thing will happen to PowerPoint and Google Slides. Yeah. Uh, same thing will happen to Google Docs and, uh, Microsoft Word. And, you know, like, let's say you're trying to write a letter, whatever that letter is, going forward, you shouldn't need to just pick from templates. You should be able to just describe the letter and then just write the letter, right? And then you can obviously take the last 10%. Yeah, so I think all of that has changed completely.

You brought up an awesome point, like, you know, in the Reggie tool, You know, writing sales emails. And, and I think this is an important thing because, um, I think that people conflate, um, GPT-3 as like a standalone AI needed for your company. And, and I think it probably is for like some early-stage MVPs, but you need to train it, you need to tease it, you need to add other algorithms because It's not learning fast enough off your own specific use cases for your own specific CTAs or conversions or whatever.

That's—

Can you walk me through, like, build— because GPT is great, it's going to help you get to a starting point, but if you're selling a $200,000 product to a niche audience with certain conversion metrics, it's not going to work outside the box and optimize on that, right?

Great point. And that will be the next frontier for all enterprise applications. So for Reggie, as it relates to us, you know, we connect to your Outreach instance and Salesloft instance and so on. And we look at the data that exist there. So what are the sales sequences that are running? How are they performing? How do I write more like that? What are the personas you're trying to target? What are the value propositions and pain points you have set up for those personas? And then how do we make sure that if I'm sending an email to this persona, we actually talk about this value proposition and pain point and so on? And ChatGPT out of the box has none of this content. It just has internet content, right? So a lot of the next frontiers on the enterprise application side will be on tapping in private data sources that only the companies themselves have access to. So imagine for us, not just connecting to the sales engagement platforms like Outreach, Salesloft, HubSpot, but also connecting to the CRM, Salesforce and HubSpot as an example, or connecting to the CMS like Highspot, Showpad, Seismic. How do I learn from the data that you have, your brand, your voice, and then how do I write for you? That's going to be the next frontier for us. But if you leave us out, same thing is gonna happen. Let's say that you are in e-commerce and I'm trying to build a search engine. Mm-hmm. You're now trying to build a search engine for e-com commerce based on your data. What do your customers look for? What has been the search results that I've shown you before? What did the customers end up buying? And therefore, what should the new search experience be? So the next frontier is all going to be on taking GPT-3 and applying it to private datasets.

Yeah, I mean, I, I was in machine learning, um, over 10 years ago, or 10 years ago, and even though we had basic— not basic, you know, sophisticated algorithms for different e-commerce sites, uh, for e-commerce. Each site that we put it on, we had to have a data scientist, um, figure out the nuances of that site. So maybe 80% could be repeated, but the, the 20% secret sauce They needed to learn.

That's right. And that's personalization at different levels as well. So, which is what is also going to happen. So for example, for the most part, what we're talking about here is personalization from the company. So like, let's say AT&T is a customer. What is AT&T's brand and voice and so on? The second level of personalization is what the recipient is and what do they care about? as it relates to AT&T. And then the third, which will absolutely happen, is what is your voice? What's my voice? What's— I don't wanna write like how Jared writes. I wanna write like what Srinath writes. Yeah. And so, I want the AI to help me write as Srinath, not help me write as Jared. And all of that personalization will absolutely happen. And That's true for sales emails and that's true for e-commerce. Like, you know, if I'm Macy's selling a bunch of products, what does it mean for Macy's products? And what does it mean for the individual buyer who's looking to buy Macy's products? So on both sides, it's private datasets as well as personalization.

Yeah. So just breaking it down, you have AI, In general, that's now commoditized. Now you have AI through Reggie's lens, which is more targeted for a sales audience and all of that. And then you have AI through Reggie's lens for their AT&T client versus their Salesforce client, which is targeted based on industry. Then you have AI through that.

That's right.

Based on the recipient.

Yeah, that's

right. And, and the sender specifically.

That's right. Exactly. And all of that is going to happen. So it's very exciting.

So personalization at scale has just got— oh my gosh— more granular than it ever has. And that's right, only AI could do this because otherwise, um, it's a human making some assumptions.

Yeah, and that's impossible to do, which is why we never practically ended up doing that. And so you end up with templatized emails, whether that is on the B2C side, doing some cohorts and sending some emails as cohorts on the B2B side, doing templates and sending emails as templated emails and so on. So absolutely. So that's what's going to change from all of these less granular approaches to much more granular.

Do you, do you actually have to be the sales engagement platform as well yourself to get even more data on the testing of each?

No, thankfully, you know, Outreach and Salesloft right now have pretty open APIs. So we have no ambitions of becoming a sales engagement platform right now. And I think that battle, first of all, I think there are fantastic companies doing this. But, you know, all of us just build on top of each other. You know, Outreach and Salesloft could have become CRMs, but then they built on top of Salesforce. We see it as building on top of Outreach to build content. So content is our main thing, just like how Outreach and Salesloft did sales engagement as their main thing, as opposed to CRM. And then you have your Salesforce building out the CRM. So. Yeah, no ambitions to becoming a sales engagement platform.

I'm just thinking of in the API, that makes complete sense. Because I'm thinking, how can you get all the data for each test being run and suggest other tests, not just other content? Well, I guess it's content, but it could also be timing and stuff. I guess let Outreach and Salesloft put that AI level in for that part themselves, right?

Yeah, so AI is also gonna be in different places, right? So for example, lead scoring, we're not going anywhere there. There are enough companies doing that and there's gonna be lots of action there. The places around Apollo and ZoomInfo and so on, on the prospecting side, there's gonna be a lot of action there. Yeah. You know, looking at pipeline and what deals are gonna convert or not, There's gonna be lots of AI there. So yeah, I think the field is pretty rich for us. We're just focused on the content side, but I think all of these will be impacted for sure.

Now, I know most companies have, you know, certain KPIs. Like what are some KPIs that y'all are focused on right now? Like the core ones?

Yeah. What we wanna make sure is that people can actually communicate one-to-one. So this is our big push right now. So the more personalized emails are, the better it is. So ultimately, let's say that you are an SDR team, the biggest KPI that we look for is how personalized is your communication? So Reggie can actually help you write the sequences, of course, and then it can set up manual steps in that sequences. And then when it comes time to execute those manual steps, Reggie will help you with the personalization. So what we want to make sure is that the humans are actually there in the loop to do the personalization. So it's just the number of personalized emails that you're sending. And as it relates to that, the amount of time saved and so on. But for us, a big chunk is just personalized communication. Communicate as humans one-to-one, and how can Reggie help you enable that?

Are you able to enable your team selling outside of email with an AI tool that's Obvious. Do they write their call scripts out with AI too, and are they able to feed that in?

That's exactly correct.

Because that's— I, you know, I'm just reading a post that you all had your best quarter in a down market, and, and I think in, in a difficult time. Yep. And I can't help but think, what a good time to be in AI.

Yeah, and the reason also is, look, at the end of the day, our revenues are small, right? So it's not like we are hitting against TAM necessarily. So for companies where you're hitting up against TAM, I think the economy matters a lot. For us, you know, we're such a small company that— Mm-hmm. Clearly we need to work more deals to have better conversion, but for us it's just, yeah, so AI maturity, people are talking about ChatGPT, everybody wants to know what they can do. And so that greatly helped. And to answer your previous question, yes. So Reggie can help you write sales sequences. Reggie can help you write call scripts even as part of sales sequences. In the future, Reggie will be writing personalized call scripts as well. Reggie can help you write personalized LinkedIn messages as well. So LinkedIn call scripts and sequences, as well as personalized emails, all of those touchpoints, Reggie is going to help you.

Now, my big question is outside of email, because email is very easy to attribute. On the phone and on email and on LinkedIn, is, is the feedback loop automated or does it need to be manual with like the reps putting it in?

Uh, today I wouldn't call it feedback, but today the reps will have to go to LinkedIn to actually be able to write the LinkedIn message. And we do not see it being automated away. So the reps would actually have to go put in the LinkedIn message themselves. So that part is still manual, but Reggie will show up inside LinkedIn. And Reggie also pulls information out of LinkedIn. So Reggie will do both of that for you, but you have to copy paste your own content.

And then how is Reggie learning every time?

Absolutely.

So, so the rep is after the LinkedIn interaction saying no response or positive response or negative response? Like, how does, how does that come back to influence the next? And then same thing with the calls. I'm very intrigued, uh, by this, especially when the— especially when a human, uh, has to, has to be part of the cog.

Yeah, that's, that's a good point. So Right now we rely on the SCPs. So what happens is, let's say somebody picked up the phone and you stopped the sequence, you would write up a call disposition and you would stop the sequence. So then we would actually know that, okay, this ended up being either a meeting that's getting scheduled or the person is not interested and the sequence is getting stopped. So this would follow from the SCPs. Um, so we just get that information back from the SCPs. So if the SDRs don't fill up that on the SCPs, we actually don't know either.

Got it. And, and the AI from Reg, like, can— do you integrate with the, the APIs for like Gong and Chorus and, and all of that? Because I'm thinking, you know, the bigger picture with AI going into sales to really influence content, which is presented in a multitude of ways, as you know, and right down to LinkedIn posts and calibrating who's liking, who's commenting. When I say calibrating, taking in the information of who's liking, who's commenting. Yeah. And correlating that to your CRM, your pipeline.

Yep.

And, and geez, dare I say bringing community in, it's attribution for sales that marketers want. And, and, and, and then, and then hold on, because then we could also tie in Reggie for marketing copy and emails from like, uh, a marketing automation platform, right? Like, uh, technically all to the same persona or ICP group of companies. Um, isn't that how you're gonna get— and then, and then of course segment it in there to understand, you know, channel marketing and, and clients and all of that. Is that what you're seeing as the future? Uh, For AI and sales, AI and go-to-market, right?

Absolutely. And today all of it is done in a piecemeal manner, right? Like you can go to Reggie today and write a blog post. You can go to Reggie today, write a social post. But like you pointed out, we actually can get you the social post that someone has written. but we actually don't take that feedback in, in terms of who liked the post and compare it to the CRM and so on. But I think the next iteration is when all of this comes in together. So today we do a great job of closing the loop on the sales engagement side, but to your point, like how do you close the loop across sales engagement and CRM and CMS and website? And LinkedIn, that's very exciting. So I'll give you some other. So first of all, your point on LinkedIn is absolutely valid, but I'll also throw in a couple of other integrations that we have talked about. So the way CMS comes into the play is like let's say that you're doing a sales sequence and attaching a PDF. The CMS knows when that PDF gets opened. So, you know, you have signals in the CMS that's actually missing from SAPs. And, you know, so that is a lot that we can actually learn from integrating to the CMS. Same way on the website. Like, let's say that you wrote a blog post and you actually have people coming in to see that blog post. Then the question is, what are new blog posts that you should write? Should you write an email sequence that references that blog post? And then going as far as dropping a pixel on your website so that you can actually figure out who is the person who is reading the blog post and then sending an inbound message to them as part of a sequence. So yes, the more of these that come together, the more exciting it is. And absolutely, that's the future.

It's, it's, it's a really exciting time as well because We're at a time— so we have AI, but we're also aggre— we're also starting to see rollouts of tools aggregating all of the tools in your stack from a data standpoint, right?

Right.

And it— and, and how serendipitous is it that it's happening at the same time? Because now you could just say, okay, I don't need to plug in to your 50 tools in your stack that actually take data, let alone the 50 that Actually, probably most of your tools take some sort of data.

Yep.

Um, we could just integrate with an operating system, like, like, like, like a sinkery or something, right? Like, and, and look through—

That's exactly—

Interesting things and, and fig— yeah, and figure out, um, what's interesting. And interesting is defined as what, what lever can, um, And, and, and in the past, this would be us doing it, right? Like with, with like a BI layer looking to, um, have an analyst look in. But, but theoretically, you could ask the question to AI or, or, or give it a goal of what category is most productive for us for pipeline. What, uh— That's

exactly right.

What deals close the fastest? And then can you, can you then— so you optimize on the outbound side, but then can you recalibrate it to inform your ICP and inform—

Yeah, even today we do, we close the loop pretty nicely on the sales engagement side.

Nice.

So for example, we can actually figure out who are the personas that got targeted, What were the reply rates when you targeted this persona? What was the reply rates when you targeted this persona, this pain point and this value prop combination? So we can do a good job of closing the loop on the SAP side. But to your point, the future looks like asking questions like, can you create me an outbound sales sequence and target the people who I've not targeted in the last 6 months? who are in the healthcare industry in the Bay Area, in companies of at least 500 employees. And you can now go query Apollo or ZoomInfo, query your CRM to get whether you have contacted them or not, prepare the lead list, prepare the sequence, write the sequence based on the best practices like we do today, based on what's performing with the SAPs in your own content and language as you got from your SAPs and CMS and put that all together, that's very exciting. And asking questions like, what are the 10 new blogs that I should write based on the blogs that have been receiving traffic so far? What are the 10 new blogs that I should write based on the keywords that I'm not getting any traffic? For right now. And once those things come together, that's, that's really beautiful.

What are you writing full blogs for, for your company with Reggie? Or how does that look?

We do. It's not our main feature.

How does it rank?

We were actually just looking at it earlier.

You're like, we got the content out.

Sure.

If it's driving.

No, thankfully, It's getting better and better and we're getting more and more. Most of our actually traffic today is organic traffic. So most of, by far the biggest lead source are direct traffic. So people just come to the website. So whatever it is that we're doing, whether that is SEO or marketing or brand marketing or—

Direct traffic is great.

Or funding announcements or the community spreading the word. But, you know, honestly though, the SEO has definitely been useful for sure. And absolutely. But again, that's not the main thing. There are so many tools today that are equally good at writing blogs as us, in all fairness. ChatGPT might already write a great blog for you out of the box.

But it's closing the loop, right? Like, because—

Correct.

And it's, yeah, it's understanding that that blog connects with a full content go-to-market strategy in different things.

That's exactly correct. So you need to actually specify what are the brand messages that you want to convey and the blogs should obey that. But you also want to know, you know, I, not my previous startup, but the startup before that is this company called BloomReach. And one of the things that Bloomreach did was on e-commerce websites to just look at webpages that did not get traffic for yourself, but got traffic for the competitors and therefore suggest new webpages that you should build. So as an example, you know, at the time, 10 years back, people had queries like red pumps, you know, where pumps refer to red heels. Like women's shoes, and you would have the heels, but you would not have the word red pumps, and you would definitely not have a page with all the red heels in your website. And so what Bloomreach did was simply say like, hey, let's look through your website, we'll find all the keywords you're not getting traffic for and your competitor is getting traffic for, and therefore we'll suggest new web pages and what the content should be that goes on those web pages. So exact same thing on blogs that nobody's doing today, right? What you should be asking is, what are the keywords that's driving blogs traffic for me, which is very easy to find from integration to Google Analytics. But in the best world, actually see how they would rank for your competitors. And the places you want to really focus on are keywords where your competitors are ranking and you're not ranking and you do not have any blog content around that topic, then suggest the topic and write that blog. And that's really closing the loop.

Yeah, I, I can only imagine what's keeping you up at night. Do you have time to sleep? You're, you're on the, the verge of creating, um, the future. It's pretty, uh, it's pretty

Outstanding. A very exciting time. I feel blessed that we were in this space in 2020 when it was just getting started because starting a company, it's very hard to time things correctly. You're either too early in the market or you're too late in the market. Getting it just right like this is a once-in-a-lifetime opportunity. So, definitely trying to do the best we can with that opportunity that we've been given.

You're speaking to somebody that was in blockchain about 3 years ago. Okay. Or before. You know what's funny about that space? The fact that so many people are throwing around the word utility with such importance.

Yeah.

SaaS, it's like everything has utility.

Yeah.

If it doesn't, it's not getting funded.

That's exactly right, especially in this market.

What a luxury. That's right, to be able to have a company without utility. Yeah. What's keeping you up at night with what you're building and how your team's humming?

Yeah. I mean, the biggest thing is everything is changing and that's both good and bad because everybody's trying to grapple with the change all the time. And everybody's reacting to it today. If you compare that to 2020 when we started, Nobody was reacting to anything. There were early movers like us, and there were a few other companies, and nobody was reacting. They were all thinking ahead of what the future might look like and planning around it and so on. So it's lots of planning, very less reaction. Today, it's sort of the opposite because you see Microsoft reacting, Google reacting, and soon I'm sure AWS reacting, and Salesforce will be reacting. Everything today is reactionary and everything today is about what somebody else is doing, as opposed to what your plan is and what your goal is and what you think is the future. There's lots of FOMO, there's lots of, okay, I need to do this thing too, as opposed to you coming into and thinking about it from first principles to say, okay, what do I think the future is going to be and how do I put myself in that position? So what's keeping me at night is to distill that in the sense, not get dragged into what others are doing and not worry too much about reacting to what others are doing, paying attention to it, but not letting that drive your roadmap and to think clearly what your roadmap should be for yourself. And so that's what keeps me up at night, is just making sure you don't get into this FOMO trap like everybody else.

The FOMO trap is, uh, is a good thing if you're not getting into it, from the sense that it'll take care of your competitors on its own.

Yeah, exactly. And it's a long game. Like, look, everybody is looking at ChatGPT last 10 weeks. that it's been out and everybody's like, what's my ChatGPT strategy? Realistically, you know, this is going to play out over the next 10 years. So sure, you can react to it in the next 10 weeks, you can react to it in the next 10 months, but let's see where everybody is in the next 5 years because the good stuff will take time to come out. It's not going to be instant.

What do you think is the future of ChatGPT, AI? I mean, I know GPT-4 is literally the future, but like, what do you think is gonna come out with all of that?

Yeah, one of the big advances that will happen is logical and math reasoning will get dramatically better. So, you know, the beauty of the AI systems that we have today is It has been very, very hard to teach AI English grammar, if you think about it. And Grammarly, the state-of-the-art system, really actually doesn't know too much grammar. All that it, for the most part, does is look at 4 words at a time, count the frequency, and try to see if the 4 words that you have in sequence is less frequent than the most frequent. And if so, they have a human annotator to look at all combinations of 4 words and tell it why this is probably grammatically incorrect. But we could never teach computers grammar. But the beauty of it is we did not need to teach computers grammar for it to write long-form content. It just learned it by itself, right? Like, we didn't teach it anything if you look at OpenAI and GPT-3. We just circumvented the problem altogether. Same thing will happen to math. We have been forever trying to do these techniques of theorem proving and so on, where we have been trying to teach math to computers and we made very incremental progress. Exact same thing that happened to grammar is going to happen to math and logical reasoning, where the computers will just start picking it up, as opposed to you trying to teach it addition and multiplication and all of this stuff. It's just going to pick it up. So I think math is going to see a lot of advance. Logical reasoning is going to see a lot of advance. So those are immediate, and this is where again we will be caught off guard because you don't know how they will intersect with the language tools that we're building today, right? And those are the kind of cool things where you will have these intersections of things that's hard to see today, but that'll happen in the very near future. And then in the medium term, you're going to have multiple modalities for sure. So images, videos, audios, text, all coming together simultaneously, those are all going to happen.

The images is an interesting one because my partner is a creative director for a major magazine, and it rubs her the wrong way for a couple of reasons. One, because it really could put people out of business in, in that space, but also because to create a new image You need to infringe on copyright.

That's right.

Okay, they're taking images that are copyrighted to create either inspiration or pieces of, um, something else and cutting out the other folks. And, um, and it's a very slippery slope.

That's right.

With, with art in particular, like math

That's right.

Writing, sure. And with math, like, if you were to ask it the hardest question ever and there's no citation or anything and people don't know it, how do you know the answer is right?

That's right. No, not just that. I mean, to a certain extent, math has similar problems as well. Like, I can't— I guess I didn't give my introduction at the beginning. You know, I— my background is—

I was about to go into Tell me more about you.

Going through a little bit of my background, you know, I come from the academic world. I did a PhD in computer science from Carnegie Mellon. So I come from the AI computer science side of things as opposed to the sales side of things. In academia, people would get pissed off when you do not cite them. Like pissed is like an understatement, you know, And so that is a little bit of problem even there. Like, if it's where you draw the line, nobody's going to be pissed off if you used a simple, well-known fact in math, but everyone will be pissed off if I actually came up with a theorem that you used to prove your own theorem, but you did not cite me. I will be pissed off as hell. So, the same copyright stuff applies there as well. But it all depends on what— just because you use yellow or red doesn't violate copyright. But if you took my picture and you ripped it off, I'm going to be pissed. So, there is a line. I think all the agencies are going to have to figure out where that line is clearly. And that's also where—

It's going to be hard to trace because if there's 200 pictures to make one, 2 people might recognize and, and 198 might never realize that they were infringed.

Yeah, same is true even with, uh, citations and so on in academic papers. Like, you know, because it's such niche, uh, communities that you may not cite somebody, you— they may have already proven it before. That's the other thing. If somebody has already proven this before and you actually pretend as though you're proving it for the first time, they will be pissed. Um, you know, and these fights get fought all the time. Um,

but, but now it's the AI.

Correct.

Serving it up.

That's exactly right.

More than just one person who's a renegade. Exactly correct. Serving it up to dozens of people.

That's exactly correct. So I think A bunch of regulatory bodies will have to figure some of these things out on the general purpose side, but that's also where I think specialized systems are in the clear. So like, let's say that I work off of your CMS data, you're in the clear. Let's say that I work off of your brand images and I'm trying to help create a new brand image based off of your copyrighted content that you own. you're in a much better place than saying like, hey, here's a bunch of copyrighted content that I'm just gonna steal and remove the copyrighted symbol in there. So yeah, I think this is where the private data sources, where you're actually building it from your repository of content, is gonna become much more relevant. I don't think that images, I don't know if, They're going to be put out of jobs though, to be honest, but it's another discussion. Um,

so, um, stock images, right? Like, you could just theoretically replace all the stock images, which, believe it or not, traces back to individuals that are getting paid for those. So there, there's, there's some— there's, there's a lot of images in the image world So, and interesting conversation nonetheless. But, you know, now learning a bit more about you, like, like diving in, you, you were in Facebook early. How did, how did things change from being, what was it, one of the first 100 engineers in Facebook?

Yep.

To being the CEO of Reggie, like, forget title, like,

how

has technology evolved?

Yeah, you know, at the time Facebook was really not doing any AI or machine learning or anything like that. You know, Facebook was just focused on growing its community. Um, and I was fortunate enough at the time to catch the tailwinds of the entire big data movement. And big data was coined around 2009. I got my PhD in 2007. So a couple of years inside Facebook, you know, that was the first time social was a thing. You know, until Facebook, social was not a thing in the sense that after Facebook, everybody said, similar to GPT-3, everybody said, what's my social strategy? I need to be selling in social. You know, what's my Facebook strategy? What's my Twitter strategy? Then the same thing happened with mobile. And then everybody said, what's my mobile strategy? So it was nice that I caught the wave of the entire social movement when I joined Facebook. It was very early days, less than 100 million users in total. We had just hit 50 million users. So very early days, great to see that movement. Then I caught the big data movement with Bloomreach. 2009 was the first time people were using AWS at scale. Cloud computing was getting used at scale for the first time with Rackspace and AWS and so on. So, we caught the big data movement with Bloomreach. Yeah, I'm so very lucky to be in these places. It's very interesting always to be working on the cutting edge of tech.

Now, outside of thinking of AI and work, like, what do you do for fun? Solve quadratic equations?

Thankfully quadratic is not as much fun, but I do do some recreational math, but not solving quadratic equations for sure. But my— so I'm married, we don't have any kids, but my wife and I do a lot of road trips around California. So we mostly end up driving north. So I live in San Francisco. So, but we do north, south, east, you know, on the east side we have the mountains, Yosemite and so on. And then we have Tahoe a little bit higher than that. Mendocino area, we were there just there a couple of weekends back. So, we do a lot of road trips on the weekends. Sometimes it's camping, RVs, lots of times it's Airbnbs, but glamping, my wife doesn't like the cold. But yeah, so lots of camping, lots of outdoors. I find it very peaceful to disconnect, be with nature as well. So yeah.

I like that. AI cannot take the place of nature.

No chance. No, of course not.

Now AR/VR can. Just kidding.

Good one. No, I mean, I have—

AI infused with AR/VR. Watch out.

Well, I don't think so. I have an Oculus, but you know, it's collecting dust. Um, so, uh, no, look, you have to—

That was like Kickstarter's best program ever that like Facebook bought, right?

Exactly. Uh, look, I'm not— I'm a technologist for sure, but it's not like Uh, I, I also believe in humans a lot, and I believe in nature a lot. So animals, humans, plants, trees, all of the above. So, um, you know, so definitely both sides.

Do you, do you think, uh, evolution as a whole, right, like, uh, nature, humans, all this, do you think AI is part of evolution? Of like the human race.

Yeah, I think so. But a lot of— some parts of evolution had to do with things outside of— I mean, this is a complicated topic to begin with, but some parts of evolution has to do with things outside of genetic material as well, in multiple ways. You know, typically when you talk about evolution, you're talking about genetic evolution. So changes in DNA and how that changes people, uh, or changes our—

But environment might have something to do with that too, right?

Environment has a lot to do with it, but also things that you pass down. So for example, you know, if you're building tools, that has an effect in terms of even evolution and so on. So like if crows are building tools to create nests, Or to find insects, then that puts them on a different evolutionary path than otherwise. So you can compare crows that invented, you know, things to do things and crows that did not. And that has an actual impact that does impact evolution, right? So there's lots of other— and language evolution is a whole nother thing.

Same as with bow and arrow to guns.

Yeah, exactly. And that puts you on a different path. So yeah, I mean, one way or another, it's definitely going to impact the way we evolve. That's for sure. No question about it.

This has been a fascinating conversation. How could folks get in touch with you?

Yeah, I mean, get in touch with me to talk about anything. It could be AI, it could be math, it could be Reggie, it could be the outdoors. Yeah, so LinkedIn is great. Of course, you can find me on LinkedIn. Email me, srinath@reggie.ai. No problem at all. Either, either is great. Doesn't have to be Reggie. Always available for anything.

I love this conversation. It was, it was very insightful and piqued a lot of curiosity. Thank you for your time.

Of course. Thanks for having me.

Of course, and for the audience, thank you for listening. This has been another episode of Revenue Today. If you enjoyed this or you learned something, tell somebody else about it. We'll be back next week and look forward to having you there too. Thanks so much, Srinath. Whoa! Another great episode of Revenue Today. For show notes, links, and mentions, visit revenue today dot live. For all my friends. Friends in the Rev Genius community, thank you. It's been awesome to spend this time with you. Please DM me any feedback and ideas in our Slack channel or on LinkedIn. If you're not in Rev Genius, join us at revgenius.com. It's free and it only takes like two seconds, and you'll be joining a group of 27,000 revenue professionals strong. We've got it all. Looking forward to seeing you there. Catch you on the. Flip side.