Austen (00:01.154)
Awesome. I believe we are live. thanks Vince so much for joining us and for spending a little time with us today.

Vinit Mehta (00:09.076)
Absolutely. Thanks, Austin. Pleasure's all mine. I'm looking forward to it.

Austen (00:12.184)
All right. And and before we kick off, was that is that your smoke detector that you're in the background going off?

Vinit Mehta (00:18.624)
It i it is, it is, and I apologize for that in advance. Yeah, it's it's a little too high for me to reach and it started beeping just earlier today. So so apologies. but I I I I think we should be fine. It it it's you you're gonna hear a beep every few seconds, but yeah.

Austen (00:27.376)
Yeah, we'll be just fine. It just makes me laugh. Yeah. We I bought a I bought a house about a year ago. and the it has tall ceilings and we have a light bulb that's been burned out and I have yet to replace the it actually just was a year yesterday and we still have yet to replace the light bulb. So I feel your pain. I'm way too short to be reaching high ceilings e even with a ladder. So good times. awesome. Well,

Thank you everybody for joining us. We're excited to get kicked off here. we're gonna talk about a bunch of different things today. But before we dive into it, why don't you tell us a little bit about who you are and what you work on? And I I know you've got a lot of experience, but we'll we'll jump off from there.

Vinit Mehta (01:20.664)
Well, absolutely. first of all, welcome everyone. if if if you are looking at the news today, you know, you might be confused between the two very contrary views. one view will tell you that AI is pretty much running the world, and then that you will see the naysayers talking about how this massive AI infrastructure investment is always. well, from my point of view, you know, when you dig deep.

into enterprises and look at the reality. On one hand, you will see like a massive pile of very ambitious AI pilots. And you will see a real relatively smaller size of true production live AI implementation. My name is Vineet Meta. I I head customer engineering for Google Cloud. I've been with Google Cloud for

a little close to ten years. Actually I'll finish ten years this November. And me and my team, we sit right in between some of the greatest and the best frontier AI models and the reality of messy enterprise business processes. so look I I I think I'll I'll tell you outright that the problem isn't the AI model.

It's the problem lies in in the ability of enterprises to really cope up with the innovation that's going in the AI space. So Austin, I'm I'm super energized, super pumped. I'm happy to share you know my learnings from being in the field with Google, talking to my customers across different verticals in the enterprise space. I have also, you know, I've I started my career as a developer.

dwelt into professional services and I had customer engineering, AKA sales engineering for Google Cloud. And I'm happy to talk about you know, my own journey and share nuggets from my recent experiences around FDEs and whatnot. so yeah, everything is on the table, Austin. Let's hit it off.

Austen (03:36.092)
Perfect. Perfect. so yeah, you work with a bunch of different enterprises, you know, high-tech, ISVs, manufacturing, all over the place, right? software vendors. My my first question is from what you've seen, I'm sure you can see companies that are doing a really good job of adopting AI and some that are struggling a lot or doing a poor job. What is it that you see in the companies that are what is the difference between those?

What does it look like? what are they doing differently when they're adopting AI really well and really utilizing it versus when they're not?

Vinit Mehta (04:12.833)
Yeah. And and and since you since I have a lot of experience working with, you know, 200 people company through, you know, thousands and thousands of, you know, people employ big enterprise companies, I'll tell you that the divide isn't the budget. I have seen companies with massive AI budgets. and I've seen companies with you know that are nimble and have smaller AI budgets and and and you know

the success is not the the AI budget is not the success factor in in in in companies in in one being you know thriving in AI versus the not. I think I think I I think the divide is is something else. And to understand the divide well, I would like to talk a little bit about about the evolution of AI and and where where we were three years ago, where we were a year ago and where we are now.

and and that's the crux of it. so so you know when when three years ago when the whole you know AI boom started with the advent of you know open AI releasing chat GPD and then you know Google, you know, then you know coming up with Gemini, we see Claude from Anthropic and so on and so forth. Three years ago, we heard a lot about generative AI.

This is when we you know, companies started releasing chatbots and the voice agents and so on and so forth. In just literally months and maximum a year's time, we have evolved. We don't hear generative AI that often now, right? Now, what we hear is agents and how AI agents are driving complex autonomous workflows. So the so the reason why I say this, the

The pace of innovation has outpaced big companies' ability to really, you know, adjust and adopt through that change. So that is the crux of the problem. I think the companies that are able to move along with this pace of innovation happening in the AI space are are are able to deliver way better results faster and are are able to get greater economies of scale.

Vinit Mehta (06:34.099)
I I predominantly see this. you know, the I I I I I look at this in three areas where companies, the customers that I work with do really well in AI versus the ones that are still struggling. And and and pretty much it it boils down to the three areas. And I I like to call it as business, data, and people. from a business perspective, I think it's the the the most common trap I see.

that the that a lot of companies fall into is they treat ai as a technical problem than a business problem. These are these are the executives that are that are looking at at a at at this amazing technology AI that is in search of a business problem. The ones that do really well they really flip it around and have a complete different perspective on things.

These are the people, these are the executives, these are the visionaries, I call them. They look at, they start with the business problem. They look at where AI is gonna drive most meaningful impact for my company, and then they kind make their way work from there from there on. So I think that that's number one. I think it's the perspective of of treating I I I hear this a lot where you know smart executives.

you know they they think of IA they put IA before AI IA meaning information architecture before artificial intelligence and they look at it that way. So that's number one. The number two, the second reason I see a lot of companies lagging behind in AI is they they don't look at data. They they they don't look at AI with as from a data perspective.

I think your AI is just as good as your data. and companies that do phenomenally well in this have identified a unified data platform, a data platform that is able to scale to their AI needs. They they look at data platform that is extensible to not just one model but to multiple models at the as the models leapfrog each other, as we are seeing every day, right? One model outpaces the other one.

Vinit Mehta (09:01.409)
having a data foundation is the key. in in in my first statement, as I said, like you know, the piles of pilots into realizing actual enterprise value. If you have the data foundation figured out, you will you will be able to move projects really fast because that's when you would have already thought about about compliance, about security, about extending, about scale, and all of that. And the companies that really struggle.

they they they don't put data in the forefront. They are they are there to r you know really create some cool search bar and and this is when they don't really truly realize the business value out of AI.

Austen (09:45.093)
So what does that look like? If you if you go into a company and they say, Okay, we probably don't have that data layer put in place, obviously it varies a lot company to company, but i if you were if you were going to give them advice or wave a magic wand, you would say, What what would you tell the generic company they need to do? What does that actually mean?

Vinit Mehta (10:06.795)
Th yeah, that that that literally translates into pretty much making sure that the data foundation is in place. there are a bunch of vendors out there. I don't want to do a Google Cloud pitch here, but you know, this is where having a platform you know, that is end to end, you know, that that that makes sure that you have a consolidated data platform and you are able to build AI on top of that. So so so we start with you know, making we we we

Like literally, we would have deeper architecture sessions, understanding the data pipeline, understanding the data residency requirements, understanding the data lifecycle, if I may, and then you know, have them think through that. Because a lot of these big enterprises are sitting on legacy systems. Their data is siloed, it's all over the place. And if you are if you're gonna just put a shiny AI layer on the top.

Austen (10:54.992)
Mm-hmm.

Vinit Mehta (11:02.613)
you're never gonna reach you know achieve results that you intend to because you know as I as I said your AI is just as good as your data and you know the results are not gonna be as good. So so this is what we pretty much do. We we make sure that you know companies really understand and have invested in in in in their data foundations well before they can think of AI. so so that's the second one. The third one, the third pillar which I think is is very important.

Is the human and the culture pillar of this. look, I'm a firm believer that technology doesn't drive value, human does and people do. So so the ones that really succeed at this really understand this very well. And they they invest a ton in AI fluency, in making sure that each and every organization within their enterprise business is fluent in AI in their own nuanced way.

I was exactly a year ago, I was at a CIO summit in in Washington, DC. And this was a year ago. This was I I think agents were just picking up. And I I spent I know, right? Yeah, exactly. it was it was beautiful this time of the year in DC. but you know, I spent two days with these 40 CIOs and a lot of them really, you know, they really spoke about the culture and the human aspect of it.

Austen (12:12.184)
There's ancient history for AI. A year ago is yeah.

Vinit Mehta (12:32.435)
Then there were some great ideas, and one of the great ideas that was you know that was spoken about a lot during the CIO summit was the concept of AI COE or AI Center of Excellence. so so they spoke about how you know a big enterprise company you know put a team, put a tiger team together that was representing every BU.

within their within their organization had representation from HR marketing, finance, engineering, product, so on and so forth. And and how this Tiger team together came with a charter of of driving AI fluency across their enterprise. they they came up with you know we they really invested a lot. They hired consultants from you know they would hire consultants from Google and elsewhere to to even help them with

job descriptions of you know how they were gonna repurpose the existing roles into the roles that were more AI aligned. And then this committee would then you know talk about how do they drive AI push top down. They would have their CEO or their top executives send a company wide email that these are the expectations. How they would go bottom-up driving enablement you know for their customers. I have this this very big manufacturing customer of mine

and you know we are we are talking about amazing use cases and their CIO, you know, he and I met like a month ago and one of his asks of me was, Hey, I have these massive manufacturing plants in Guadalajara, in Chennai, etc. you know, I want to drive AI awareness on the factory floors. you know, can you can you help us train those you know those workers you know who might have an iPad with them and how can they

build low code, no code agents and and you know automate their work. So so a lot of so so the the the third third one is is you know just being super crisp here is the culture and the human aspect. I think that's a clear distinction. I've seen companies, you know, that purchase AI and never really care about enabling their works workforce or have a sophisticated way through center of excellence of of really you know delivering that outcome, making sure that each one of their

Vinit Mehta (14:54.003)
Knowledge workers. Are you familiar with this term called knowledge workers? So knowledge workers is anyone who has domain specific knowledge, you know, be it a customer support staff or be it a FinOps person or be it a marketing person. And and the nuanced AI knowledge for their domain is very different. So so these are the three, just to sum it up, I think number one is how the leaders look at AI as a business problem, not a technology problem.

number two is is the data foundation, which is super important. And just to summarize, number three is the human and the culture aspect. I I I posted on my LinkedIn for those of you who want to check it. One of my customers who's pretty much the backbone of the biggest telco companies and the broadband, you know, they they posted a LinkedIn article recently. They were talking about how the

How the domain experts, their customers, being the broadband providers, these are the folks who have spent 30, 40 years building robust systems, how they are outdating way faster than those companies can hire new talent. And the way this company is solving for that is they are creating agents that are being the force multipliers into these to be hired or newly hired broadband employees at these broadband companies.

So that they can fill that knowledge gap you know, of of those people that have been around for years. So so so that these are all amazing perspectives that I'm hearing. And I would I I don't know if I answered your question, but these are the three ways and three very distinct ways I see you know, companies looking at these things and how they they you know how

Austen (16:41.475)
No, that's a that's a great answer. Yeah. I mean, I I think it's rare to see someone who sees such a breadth of companies. And it is interesting to me, and you you addressed and acknowledged this pretty early on, but very little of what you talked about actually reference the AI, the technologies, the model. You know, your solution to how does a company adopt AI isn't

You need to use this model or that model, or you know, it's about data and people and purpose more so than the the technology itself. So that's that's probably a good thing. if yeah. So so jumping back a little bit, let's say you're you know, so I invest in a lot of startups on the side and we talk with a lot of folks at at Gauntlet. three years ago.

you would have two main expenses at most tech companies. You would have people, and then you would have the AWS or the Google Cloud bill. And those were the two things. Now there's a third line item that's starting to creep in there, and that is, you know, inference or AI spend or whatever else that falls within that category. how are you seeing CIOs and CTOs think about that today? and

Yeah, just take me through what you're seeing. How are people justifying or not justifying the spend? I feel like for a while we had a you know, everybody had unlimited money to start throwing at AI and then they saw what that meant and are starting to pull back to some extent. What are you seeing on your side?

Vinit Mehta (18:21.607)
I'm I'm seeing all of that and more. look, it's it's a real problem. And I think I think a lot has to do with the whole innovation, the pace of innovation that's b that's been driven by not just the frontier model companies, but you know, the industry as a whole. I think I think the CIOs, the CXOs are struggling with with the ROI question from their board. I I cannot name names, but I've had some of the biggest companies in the world.

you know recently had a couple months ago we had a conversation with their CIO and he was he was being very vulnerable and he was like well AI is is a cost in my line item and we have still not been able to associate ROI to that and it's not just about throwing dollars at your at your AI ROI problem it's it it needs it needs a solution. Look I'll be I'll be lying if I told you you know

I know the solution, or for that matter, anyone knows the solution. I I think it's it's a hard problem. the ROI problem is the hard problem. I think I think industry as a whole is is going through the phase of figuring it out. and I I'm I'm super bullish that you know we will every enterprise, most enterprises will get significant ROI and more.

To continue to invest in this AI AI cycle. But coming back, you know, the the the boards, I also we also partner with a lot of private equity companies. You know, these are the conglomerate of many different companies and and and and they partner very closely with us and and and they are asking the CXOs of their seven, eight companies that they that they that they own the same question. I think I think there are three three basic questions from an ROI perspective.

Number one is who is gaining from the ROI? Like who is gaining from the AI ROI? How is AI driving ROI? It's the who, the how, and the where. And where is this AI ROI actually showing up? and these are the three very basic fundamental questions. And and and the and and I I've seen companies and CXOs that do.

Vinit Mehta (20:41.933)
Relatively well while still continuing to get better at this ROI thing. They do three distinct things better than the ones that don't. these leaders they make decisions fast. They do not just sit on something, they they they fail fast, but they they are they they they are opinionated.

And and they may be right or wrong, but but I've seen them make decisions fast. These folks are maniacal about AI fluency.

they they take it to their heart to make sure, you know, the the technology that's gonna revo re revolu revolu revolutionalize the human you know the mankind they are they are manacle about driving AI fluency within their own organization. And the third and the most important thing that they think about is is how to embed AI in their core business processes.

So let me expand this a little bit. Imagine, Austin, that you know, you are a CIO of a big manufacturing company or a big retail company, right? how what what does it mean to embed AI in your core business processes? Like what what what are the business functions you have in your company? You have HR, you have sales, you have marketing, you have legal, you have finance, you have sales, you have engineering. the the the the lowest hanging fruit was engineering.

right and we'll talk about the AI spend this is this is where it took off because of some of the amazing coding models out there I won't name names but you know they the you know enterprises gave you know gave these

Vinit Mehta (22:26.023)
Very, very powerful AI models to their employees to drive developer productivity. And then we saw 60, 70, even 80% of the code being written by AI, which is great, right? We've seen great successes there. And I'll talk about the you know the economics part of this as well. But I'm I'm right now addressing embedding AI into the

Core business processes. So developer productivity, employee productivity, you know, creating chatbots, creating agents for customer support, you know, giving, you know, Google Workspace, Gemini, and the like tools to your employees so that, you know, they can transcribe, they can, you know, write notes, they can write emails in way lesser time. So all of this is employee productivity. the customer experience, right? How do you embed AI in your own product to drive amazing customer experience?

Is like if you have a UI, if you have an app, how do you capture a customer sentiment? And how do you drive? At the end of the day, there are three things leaders are trying to do. They are either trying to save cost, they are either trying to create new value through ROI and upselling their products, and they're trying to achieve more innovation out of that, right? So I can talk about HR use cases where companies have developed embedded AI into their HR processes and output out of

their every HR employee is 10Xing. so these are the three things that I see you know, leaders who have a way better control of ROI question do. Summarize it, three things. Number one, they make decisions fast. Number two, they drive AR fluency. And number three, they embed AI into the core business processes of their organization. Now let's touch upon this very important aspect that you brought up is

You know, the cost of AI, right? So in the evolution, what happened was, you know, there was this FOMO of I'm gonna miss out on AI, and every organization had AI budget side, you know, reserved for that. And what I'm seeing right now, I'm not even making this up. I'm seeing, I don't know if you're familiar with these terms of AI sprawl or agent sprawl or tokenomics or token maxing. We can spend hours on each one of these. Basically, what what this is is what is happening right now.

Austen (24:38.958)
Mm-hmm.

Vinit Mehta (24:44.453)
Is in a lot of big organizations the AI spend has gone completely out of control. now let's take those developer examples, right? early, like you know, a a year ago, I would I would talk to my friends who are engineering leaders at some of the awesomest Bay Area companies. They would talk about how they have leaderboards on on

you know, folks who are maximizing tokens and are using most tokens and all of that. I was just having you know, we had a social gathering and I met a few of them. I was like, how how's that leaderboard going? They're like, shit, dude, like that's that's in fact now frowned upon is, you know it's it's like now frowned upon. Like, you know, why are they using so many tokens? Like, you know, those so people are token maxing and and I think the right way to look at it is

Austen (25:27.406)
That's shut down now. Yeah.

Vinit Mehta (25:39.998)
Are is is your token consumption driving the right business value? We we at Google, we you know, you know, we we have some of the amazing frontier models and one of the ways we are educating our customers is exactly on this very topic of tokenomics. we've built some amazing analysis that we walk through with our customers.

Austen (26:03.057)
you guys can.

Vinit Mehta (26:08.757)
All good. so so so we we we we we talk about token tokenomics, token maxing quite a lot. There is a whole business value side of this. And look, I think some of the smarter leaders, there's a lot of noise out there in the industry, right? There is you know, there are frontier models, there are open weight models, there are other Chinese models, and you you know, you'll see news from each one of those, you know, coming up with amazing power.

you know, the but the key here for somebody who's a visionary is to really understand, like you know, if you want to give our you know, our family a car or or or any tool for that use, like you know, this is more like giving Ferraris to your kids. Like you really need this very expensive model. So our talk to our customer is a lot about how can you number one do

Like 80% of what every enterprise tries to achieve today can be done through cheaper models. For that for those 20% of the use cases, you want like the best in class models. And this is what you know we do with our customers: you know, educate them, come up with an architecture that's kind of a hybrid of different frontier models.

Austen (27:15.959)
Certainly.

Vinit Mehta (27:33.889)
depending on their needs. and and you know, if you have a model which is costing you half and getting you eighty percent of the way that that your other model that is gonna cost you double, then you know it's it's a choice, it's a clear choice that enterprises and the leaders needs to make. So so so so that's that's my little POV on on ROI if that makes sense.

Austen (27:57.196)
Yeah, we've seen you know, we're on the training side of things. So there's some organizations we work with where their MO is spend as much as you want, go crazy. and then it takes about a week before they look at one, you know, wait, this person spent ten thousand dollars in inference last week? How is that even possible? Like, okay, maybe maybe unlimited isn't the right way to approach this. And then you look at what that person's doing.

There are instances and you know, we have customers where they have individual employees spending a million dollars a year in inference cost and they're thrilled with it because of what the output is. And there are times when, you know, you're using the 10x most expensive model to check your email, and that's you know, that's a little bit overkill. so yeah, it's it's been fascinating to watch. And there's a little period where

Vinit Mehta (28:42.914)
Yeah.

Austen (28:49.095)
Everybody was a little bit blind to what the cost would be. And now we're seeing a lot more. Okay, you need the right model for the right job and you need to scope it and the ROI needs to be there. we're still seeing companies that have a hard time attributing. I mean, it it all comes in kind of bundled, right? Like you get a bill from a model provider and it's pretty difficult to like distribute that out to which area of you know, is that being used to lower cost or to

mess around or and is that within which business unit is that happening in sometimes? so yeah it's been

Vinit Mehta (29:24.023)
So so so let me let me I I I I think you make me think about one other very important aspect. And again, I'm I'm gonna try not to pitch Google here, but let's let's talk about a platform as as such, right? And and you can just use Google you know as a reference. I think this is where the leaders I've seen, you know, who are the true visionaries, they bet

long term on any on any AI technology. One is, you know, model is a commodity. today it's it's some model on a number one spot, tomorrow it's gonna be someone else. that's that's not a question. I think I see a lot of

conversations now that the dust around model has settled down is around end-to-end platform, is is around governance, is is is around security, is around putting those guardrails. you know so so so so that's that's that's been a lot of you know and a lot of leaders pick a platform. I'll give you an example, right? Google is the only cloud provider that has everything from the chip

All the way up to the frontier model and everything in between. So, what companies like Google are gonna be able to do is they're gonna be able to provide you the economies of scale for tokenics that nobody else will be able to do that. So, so leaders are are now now that you know the whole phase of

what model and let me give the best to my kids, aka my employees. I shouldn't look at anything else but the best out there. Now that approach has started changing and people are looking at a platform as a whole, then than than just, you know, looking at one model. Does that make sense?

Austen (31:12.928)
Yeah, totally makes sense. fascinating. Yeah, I mean it it's not surprising, but it is it's kind of fun to watch everybody run into the exact same everything's changing so quickly. You get to watch hundreds of companies experience the exact same thing at the same time and everybody reach for solutions and learn. It's been it's been fascinating to watch. A hundred percent. Yeah.

Vinit Mehta (31:34.663)
And you're able to see the patterns, right? Like the patterns are very clear. The patterns are very clear. And you know, it's it's I'm blessed to be you know, at a place where where you are able to see, you know, you're you're able to adjust your Apache and you know, you see the lines very clearly. So yeah.

Austen (31:51.413)
Yeah, the number of data points that you have is is pretty wild and across so many different industries and yeah, that's really cool. okay, so I've got a question for you that is gonna be a little bit generic, but let's say you're talking to a company and they are deciding between hiring kind of one of two profiles. So one profile is the deeply technical expert, you know, and they're

New to the domain that the company is in, but they're deeply, deeply technical. The other is they're deeply domain experts and maybe not as strong technically. how do you what way would you be pushing companies to go? Obviously, it's a little bit specific to the details, but what what have you seen across big enterprise orgs?

Vinit Mehta (32:47.675)
my my short answer, and and just to clarify, this is the enterprise company. So I I I I think my answer would change depending on whether it's the Google or OpenAI hiring versus versus you know the other companies who are not the who are the AI buyers and not the AI creators or developers, right? So so from that perspective, my short answer is hire domain experts who are AI curious.

Austen (32:56.812)
Yeah, if it's a two person startup, maybe it's different. Yeah.

Vinit Mehta (33:16.449)
and and you know are are trainable on AI. That's my short answer. The reason why I say that is number one, I think the entry to barrier from a technology perspective has decreased significantly with AI.

what was what was the entry to barrier during our times, right? Like, you know, we had very specific swim lanes. Like I went to an engineering school, I did my bachelor's in computer science, I did my master's in computer science. I ended up my first job was a developer at IBM because I knew Java, I had learned programming languages, blah, blah, blah. now that has all changed.

Austen (33:58.125)
And then a company that used Python would say, no, you're a Java developer. That's too that's too far afield. Yeah.

Vinit Mehta (34:02.221)
Exactly, right? Yeah, so the entry to barrier, well, you still need to have an engineering mindset. You still need to, you know, have that aptitude to really understand concepts and tools and all of architecture and all of that. But the entry to barrier has from a programming language or a syntax perspective, has pretty much diminished. There are other entry-to-barriers, but that's that's why I say domain. Second one is.

there is a massive scarcity of deep domain experts. I gave you the example of of one of my customers, you know, who's the backbone of the broadband companies. And and and you know you can you can see the quote from their leaders. They like the the engineers, the domain experts at those companies are retiring way faster than they can they are able to hire them. So so so there is that, right? And I'll give you another example. I was

I was I was having a off side in New York with this retailer you know that's that's in women cosmetics space. They cater to Gen Zru and their their the a big chunk of their sales strategy is the TikTok ad campaigns for teenagers on on on TikTok and other social media platforms. and we had a workshop with them.

And we thought, like, you know, you know, it's easy to solve a marketing use case. We can easily come up with marketing content, so on and so forth. the the result out of AI was something that was absolutely not acceptable to them. That was because it did not speak about their brand. The content that came out of the AI was was not relevant to Gen Z. So there was a lot of domain.

knowledge, the company knowledge, the knowledge about their customers and what they like and what they'll not. So that's just another example of how AI is not replacing their their marketing and their content creators. What it is doing is it's enhancing the output. Now the real ROI for for this CDO, you know, at this retailer is that they're gonna save tons and tons in terms of services

Vinit Mehta (36:27.755)
that they would outsource elsewhere and AI will help their own marketing team drive tremendous outputs. So so that's why I think the domain knowledge is super important. but I rest my case there. I don't know if you agree with that or not.

Austen (36:43.633)
I think it it does vary company to company for sure, but I generally agree. the and I say that as someone whose business model is training the AI side, right? the we find more and more that our focus is more on deep developing deep expertise in the business side and the the non AI side as much as it is building the next generation AI workflow and

And yeah, we see that certainly cohort to cohort. What was a very difficult technical challenge six months ago is now free with the latest frontier model today. So that's, you know, not something we have to train people on anymore. so our our curriculum changes every cohort cohort, every quarter. It's very, very different.

Vinit Mehta (37:32.928)
I I I I truly believe there are there are gonna be like in AI there are gonna be two kind of roles. I think one is, you know, either you are like a a scientist with PhD and you are working at the frontier labs like DeepMind and others and you are like really working on these open mind you're doing the research. Yeah. Exactly, exactly. And then the other one I think is is the true application of AI is gonna be driven

Austen (37:50.229)
You're in the research and you're building LMs and yeah.

Vinit Mehta (38:02.537)
mostly by the domain experts who will learn AI to make their own field way more productive and give way more better outputs. That's that's my belief, but yeah.

Austen (38:15.487)
Yeah. We we we strongly encourage you know, we we kind of have this philosophy internally that look, the the AI models are going to continually be better and better at giving you the default consensus best practices of everything, right? Whether that's engineering or marketing or you know, so your job as the human in the loop is to guide and direct the AI, but also to

to have your own point of view that may differ from what you know, your own knowledge and expertise that's not going to be in the weights, so to speak.

Vinit Mehta (38:48.769)
Hundred percent.

Vinit Mehta (38:53.165)
You human in the loop. Human in the loop all the time. so so so I'm I'm I'm glad you mentioned that, right? Because most of the critical use cases. I I I was telling you know, just just before you joined, I was speaking about I was I was in Boston yesterday and I took a red eye. So I I spent time with one of my customers

Austen (38:56.479)
Yeah. Yeah.

Vinit Mehta (39:15.083)
Martech marketing technology team. We were building agents. We had a workshop in Boston. And one of the agents that they wanted to develop was how do they manage their ads budget?

And they wanted to write an agent because you know depending on the demographic and the and the and the shifts and the weather shifts and the changes, they might need more ads budget in Chicago versus in LA and that would flip the other way around. and and and we designed an agent and and the results were great, but what was at stake was way too much. So they definitely wanted like they you know, they could end up spending a a shit ton of money somewhere.

Where AI thinks it's the it's the right thing. and and of course it would improve over a period of time. And this is where human in the loop is so important between those deterministic and the non-deterministic flows, right? Especially in the non-deterministic flows, where where you know the results, it's not black or white and it could be anything, and that's where we need the domain experts.

Austen (40:21.419)
So along those lines, something that we have seen. I mean, the I'll I'll use a term that has been around for a long time, but I think the frequency of me hearing this term has probably 10x over the past six months. that's the forward deployed engineer. every every company that we talk to, every fund is obsessed with forward deployed engineers right now, and they're trying to figure out

what that means and engineers are trying to figure out how to get into that role for the next 10 years of their career. what's your viewpoint on this? And if you are an engineer who wants to move closer to being a forward deployed engineer, what what would you advise somebody to do or think about?

Vinit Mehta (41:09.729)
For sure. man, I've I've I've s so much to share on this. So FDEs, where well this this word was coined I think by Palantir, where they where they really needed you know

To that there was a bad need to embed engineers in the trenches with the customers to work hip to hip with them and and solve their problems. So FDs are nothing but you know using an analogy of Navy SEALs. These are your your engineers, these are your hackers or solution creators who you can embed in a customer with a customer.

to solve their particular specific problem. So so I I I I think especially with AI and the innovation around that being being so fast, I I see a real need of FDEs. Even at Google we have we have hired and we continue to hire a ton of FDEs. So so these are the folks the distinction between them is between FDEs and a true engineer is that

A true engineer ships code, FDE's ship solutions. So these are the folks who are. I'll I'll give you my example, Austin. I still remember the very first job out of my grad school in upstate New York. I ended up with with this company called Open Pages. We were in governance risk compliance. I had

no idea about what that mean meant grc. I was I was developing I was a developer I had zero idea on how how my work was getting leveraged by some of the biggest banks and and financial institutions to help them comply with Sarbins and Oxley right I had zero idea about it. Now now now now that's I was shipping code in this day and age

Austen (43:12.938)
So you'd just get a ticket, the ticket would tell you what you needed to build, you'd build it, you'd ship it off, it's gone.

Vinit Mehta (43:15.859)
Yeah, yeah. I I would own a domain and I would I would write a code on a particular module you know that would be connected to something else. I had zero clue about how my work was driving a customer impact. zero clue about this. So so that's an engineer. And FDE is somebody who ships solutions. So think of that think of I think of FDE is as look, we always had different flavors.

and not particularly every. I'll talk about my own journey, right? So so so back from being an engineer, I very s very quickly figured out that you know, this was not something I was enjoying. And I wanted I'm I'm more of an extrovert. I love to be in front of customers and I I was it was it was hurting me that I didn't know how my customer was using it. So I then you know I I I switched to a professional services role.

which you can call as an FD, right? A professional service is somebody who's paid to to advise and to talk and to deliver PSO, PSO, professional services organization. I was a consultant you know, with with this company called Apogee, that that was into API management platform. Great. Like I was I would try I got to travel the world, you know, worked with some of the biggest media companies, telco companies. I was deployed at BBC.

Austen (44:17.598)
What what did they call it back then? They probably didn't call it

Sure. Sure.

Vinit Mehta (44:39.627)
When BBC was coming up with their own BBC store. So I implemented a solution that was post sales. So I I think there are several aspects of because I was working hip to hip with a customer in a post-sale say in a post-sale situation. I was not engineering though, because I was building a wrapper on top of the engineering shift code to make it customized for my customers.

So so it comes close to FDEs, but that that was a role I I I was in. and and this is important because I'm going somewhere here. From there, I I was not learning much, I'd done many, many implementations, and over a walk, one of my friends convinced me that you should switch from post-sales to pre-sales. And and I thought, which was great. And I was like, I know the product in and out, and and I should be able to do this. I still remember.

My very first meeting as a sales engineer. this was like 13 years ago. and I was I was debriefing with my sales rep in one of the Starbucks in San Francisco, and I thought I'd crushed it. and she felt the opposite. I because see, remember, I was in PS, so I was paid to talk. Here, I'm I'm trying to solve the problem, but I'm also a consultant.

Austen (45:57.458)
Yeah. You're trying to solve the problem for and

Vinit Mehta (46:02.389)
I'm paid an hourly dollar wage to talk and advice. And I took the same thing here, which was absolute I did not listen. I did not, I assumed that I knew what they were asking about. I was according according to my rep, I was too dangerous because I knew too much about the product. Because you know, in a post-sale situation, you are very vocal about what you can and cannot do.

and so I had to go through a lot of unlearning of existing skills and learning of new skills. And where I'm going with this is FDE is a role I think that sits right in between the two. You can get away because I also built a lot of skills around relationship building, around executive presence and whatnot. So if you are an engineer.

Who is you know, who is wanting to you know go beyond your, you know, who is going through the same ocean as I was. Like I don't really understand how my customer uses this pro how is my work influenced at the end user or at the customer, then you can consider this where you do not have it's not a big shift like the one that I went through, from engineering to pre professional services to sales engineering.

But but these are the four or five things I think one needs to, as leaders who are hiring FDEs, they need to account for. I think first thing first is technical depth. I know I said the barrier to entry is bare minimum, but the technical depth around understanding enterprise systems, around understanding architecture, around understanding modern data.

science and the data platform and the AI tools. There's a lot of noise out there, but you know the fundamentals and the basics of the rag and everything should be there. That's number one. Second one is product intuition is super important for NFT. You need to think solutions and not product. Like not just what you have, you need to translate your technical skills to solve for a business problem. Number three is

Vinit Mehta (48:22.599)
One needs to have a high empathy. It's not just how I see things. The mindset should always be about how will my customer look at this feature or look at this particular solution and and how am I making their lives better? That's that's super important. I think communication is important because you're also sitting right in between product engineering and your salespeople and your customers. So you're you're gonna be under a high pressure to deliver and hence.

Communication and you know, building your own internal network and influencing strategy is important. But I would put that in that order. And and I think the last one is resourcefulness and extreme ownership. Because look, if you are deployed as an FDE, especially in this day and age, the expectations are super high just because of the hype that has been created out of for this very newly created role, for the rest of the

Austen (49:19.047)
You can fix all my problems and bring a magic wand and yeah. Yeah.

Vinit Mehta (49:22.045)
Exactly. Exactly. So so I think I think I think that that's the that that's my POV on FDs. I think I think this role too sh will and should evolve. And I I am like if going back fifteen years ago, I I would be, you know, jumping jumping out to to get this role. Like I was like, this is a great way for an engineer to drive customer impact. Back in the days there was not much of an option.

Austen (49:51.178)
I'll do a small plug here that if you want to become a forward-deployed engineer, come to Gauntlet. we've that's what we do. all right. So we're gonna open it up to QA for just a minute. those of you that are in here, if you just ask a question, it will automatically add questions to our QA channel. and if we don't have any questions coming in, then I'll keep asking just a couple more. but

Vinit Mehta (49:53.057)
Sure.

That's awesome.

Austen (50:19.049)
so go ahead and ask those and we'll wait for those to come in. but in the meantime, how has your role changed over the past I'm not sure what time frame is the best to think about this, because three years ago, I'm sure everything was completely different. but let's say over the past year, what has changed?

Vinit Mehta (50:40.599)
Yeah. I I in a in a big way. in in some ways a big way, in some ways it's still

you know it's still the same because as as as a GTM leader, as a GTM technical leader, I think some of some of the core aspects of the role, you know, I'm still accountable to to drive growth. I'm still accountable to drive technical excellence in my organization. And as a leader, I'm still accountable for the culture. One way in which my role has changed significantly and I think

I I take a pride in in in kind of understanding this very soon, sooner than than than others, is the hiring profile.

I I I think we have flip-flopped between generalists and experts and SMEs for all the right reasons. I think at this moment in time, companies like Google, Anthropic, and many others are looking for individuals with AI expertise. That's number one. back like two years ago, you know, I'm part of a cloud company, so we hired like data engineers and data specialists and infrastructure specialists.

specialists and all of that. I I agree that all of that still exists. I would add AI on top of everything. So so so so one thing I particularly look for in the talent that we hire is is how AI enabled or trainable they are.

Vinit Mehta (52:11.697)
so so that's number one. number two is look, I think we we we we should practice what we preach to our customers. And if I'm going to my customers talking about how they can build agents to you know do magic at their organizations, we should be doing the same. So in three weeks from now,

Our entire org, you know, this is this is from the highest level. We are all gonna go through deep, deep training and retraining. and you know, you you know, back to the the Michael Jordan example, right? At at his peak, he would still go back and practice the the very basics of basketball. And I think as technology leaders, you know, especially in this day and age of AI, we all need to. So so so so that that is

change we we we were kind of a little laxed you know i've been way too long with at at google but you know yes the technology evolution was happening but the pace at which you know the the google is shipping products and technology is is overwhelming to be honest so so that's that's the aspect you know i i i think you know from driving technical excellence and and holding oneself accountable and looking for hiring and and then you know

Time and again, s you know, stating out loud on what good looks like, rewarding you know, folks and calling out folks, you know, who are doing stellar work in terms of driving thought leadership in AI is is also super important. So so those are the three things I

Austen (53:51.571)
Awesome. All right. So we got we'll we'll make this our final question and then we'll let everybody get back to their day. but maybe it's you know puts a little bow on this conversation. will you say that the future of AI is going to be a technology problem or a leadership problem? Scott Roberts asks.

Vinit Mehta (53:56.407)
For sure.

Vinit Mehta (54:11.841)
I I I I think I I think the future of AI is gonna start with a leadership problem because I think the technology is will evolve.

I can tell you for sure. And and and and here's okay, here's here's here's a better way to answer you know this question and and one takeaway, if I may also kind of squeeze into this, right? If if if you have to take anything from this conversation, take this from me, you know, just summarizing everything that we spoke about. You know, I I think we should stop asking about what model we should use.

And we should start asking about what business friction or a business problem we are trying to solve. I can tell you for guarantee that the models and the AI stack and the technology in itself is gonna get better, is gonna get cheaper, is gonna get more efficient. So that's a guarantee. What is not guaranteed is are the organizations able, will they be able to adapt.

Will the will they be able to move beyond this somewhat?

True and somewhat perceived notion of AI is here to replace our jobs. If we are able to move beyond that, then that makes it a leadership problem, is where I'm going with this, than a technology problem. I I I think it will become a technology problem at some point when when we really see the unknowns that folks like Dario and others are fearing. Like the power of AI could really do more damage. Until I see that.

Vinit Mehta (55:52.797)
I I I I think from the perspective of getting most out of most goodness out of AI from an enterprise lens, I think it's a leadership problem.

Austen (56:04.05)
Well, I think that that pretty well sums it up. thank you so much, Vinique, for spending time with us today. if people wanna follow you on socials, what's the is LinkedIn the best place to go?

Vinit Mehta (56:06.369)
Great question, though. Yep.

Vinit Mehta (56:15.029)
Link LinkedIn is best. I know Austin, you've been posting, tagging me. Feel free to connect. feel free to drop me a message. you know, happy, happy to share my perspective. Austin, it was truly a pleasure. you know, talking. I I I think I I had great refreshing conversation myself. and I hope you know, my beeping in the background wasn't wasn't a show pooper. All right. Thanks a lot. See you folks. Bye.

Austen (56:39.291)
it wasn't bad. Yeah. Awesome. Well, thanks again. thanks everybody. And we'll we'll chat soon.