Episode 03: Connecting the Dots: How Sean Grant is Transforming Data and AI at SAE International
Sean Grant: [00:00:00] We have a saying in the Marine Corps, improvise, adapt, and overcome. And that's what you have to do, right? You got new tech or a new approach or something comes in, you got to learn it real quick, get on board and start rolling it out. And it can be very overwhelming. It's just a matter of focusing on the longer term, knowing that technology is going to advance, things are going to move along.
You just got to adapt to it. Roll punches, just make sure you have your strategy and your vision in mind and how that's going to relate to it.
Brooks Canavesi: Welcome to the sales stage podcast. I'm Brooks Canavesi, your host. And if you're tired of playing by the old rules, you're in the right place. We're here to give you an unfair advantage where AI becomes your secret weapon to outsmart, outmaneuver and outperform the competition.
No gimmicks, just real strategies that top performers are using to stay ahead. Ready to tip the scales in your favor? Let's get into it. I'm your host, Brooks Canavesi. Really excited and thrilled to have Sean Grant, the Director of Data Products at SAE International. Sean is a Marine Corps, Iraq veteran, turned tech leader with an [00:01:00] expertise in knowledge graphs, AI, and data analytics.
He's also an alum of Harvard Business Analytics Program and passionate advocate for Bitcoin. We just talked about that a little bit before the show. Sean, it's great to have you on the show.
Sean Grant: Awesome. Thanks for having me.
Brooks Canavesi: Yeah. So, Sean, your career began in the Marine Corps and transitioned into IT. Can you share a little bit about that experience into the military and how that shaped your approach to leadership and problem solving?
Sean Grant: You know, I've always been interested in technology and computers growing up. My family was really good about, you know, emphasizing that and encouraging that with me. And then You know, when I got out of high school, I didn't want to go to college. I was actually supposed to go to Weber State University out of Utah.
But then military had been a big thing in my family. My dad and his two brothers were in, his dad was in, and I was like, well, I don't want to do college again. I want to do something different. I need something to help give me better direction. So I just started looking at the military branches. I took the ASVAB, which is the test you take to get in to see what your [00:02:00] aptitude is, or, you know, are you good at mechanical, electrical or these different things that you can place in.
And I took that. And I went to the recruiters, went to the Air Force, and then went to the uh, Marine Corps, and asked what kind of jobs do you guys have available, you know, I may just do this for a couple years, pay for college when I get out, and, you know, just see where life takes me at that point, just need to get some, some discipline, some order inside of there.
Air Force, I asked them questions about different jobs, they were like, oh yeah, you can do this, this, and that, didn't seem really exciting or anything like that. Then I went to the Marine recruiter office and he's like, well, we have the longest bootcamp. What's the toughest. We don't have a lot of money. We don't have a lot of people, but you know what?
We have the highest esprit de corps. We have drive around what you see a lot more Marine Corps flags and a lot more people who are proud to be Marines. It'll definitely give you discipline and give you organization in your life and help set you in the right direction So we want to do you want to sign?
I was like sure let's do it, right? So looking at the jobs they had available I qualified for Intelligence, so I was like, all right. Well, [00:03:00] this will be good. I can maybe get a security clearance and this sounds cool It's like spy stuff. I was like, oh, yeah, I mean don't get me wrong, you know shooting guns and blowing stuff up It's great, too But when you get out not really much of a practical application for that You Um, nothing against the grunts that are out there who did that, you know, that that was for you guys.
So, signed up, uh, went to boot camp and got out, and the Marine Corps, in there you only choose like the occupational field. In this case it was just general intelligence, and there's many different jobs within that. And, The Marine Corps decided that I was going to be what's called the geospatial intelligence specialist.
So this is using GIS, geographic information systems, building maps and doing analysis with spatial information, spatial data. And so this was allowed me to really see, okay, how can I use. Data and analytics to help solve problems in this case help, you know, plan for missions help, you know, basically save lives That was a really big turning point for me into there.
That's Really where I got started into real technical stuff [00:04:00] getting hands on with hardware Started going into software started doing a little bit of programming as well And that's what you know at the time I enlisted in 2000 We had really nothing not much going on, but then come around I was at school my schoolhouse in fort belvoir, virginia You And 9 11 hit, right?
So, at the time, we're like, alright, something's gonna go down. What can we do to help, right? And that really, like, cemented us for us. Like, yeah, we're really in this. We're really in the military. As soon as we got out of school, we got deployed to our duty stations. They're like, hey, stuff's ramping up. Okay, let's get you training.
Let's start practicing and getting stuff ready, because you're going to have to start learning and jump in the deep end here, right? And so, you know, the great thing about the Marine Corps, it focuses a lot on small leadership, really about the unit level down to the NCOs, the corporals, the lance corporals who are in charge, and really gives you the tools and has really good mentors in there to help you become a leader.
And that really helped me a lot. We had a lot of great leaders within our platoon. I think we were probably [00:05:00] more fortunate than some of the other platoons out there. I don't know what it was about ours, but maybe just lucked out. When I got my duty assignment, but we had some really good leaders that helped us one helped lead us into getting trained and getting ready to go out to Iraq, go to war, but also just being better Marines and just better humans and especially leading in the tech space because it wasn't just that, you know, we were out there running machine guns and doing those kind of drills.
We also had to learn GIS analysis, the geodetic surveying, going out. Um, you know, doing surveys on helicopter landing zones or finding the most optimal zone for a helicopter to land on in short notice or, you know, what other kind of products we had to build really required us to really think of on our feet.
And, you know, when I was out there in Iraq, I did get the opportunity to lead one of the teams on the night shift there in Ramadi. And that was my first exposure sort of in to tech leadership was, okay, well, I've got to manage these three Marines. I've got requests coming in from all these different [00:06:00] units.
I've got to stack and prioritize. I got to know what data I have, what I can do in order to help satisfy the mission and prioritize accordingly. And so that was a really great experience for me to, to get into that. And I really, really enjoyed that part.
Brooks Canavesi: Just as you moved into your early career, what sparked your interest in AI and data driven solutions outside of that initial experience in the military?
Sean Grant: I did my two tours in the Middle East and decided to get out. And I got a job working for Esri, which is the company that made the GIS software there in Redlands, California. And I, once again, I started at the bottom, became the DBA, the database administrator, and just started working my way up. I was really, really loved doing the technical stuff.
You know, learning how to set up Oracle Database, SQL Server Database, Postgres, learning SQL, started teaching myself Python and started doing analysis with that and some of our products and building the software. It didn't occur till maybe five or six years ago, I'd say. Um, I'd worked my way up to be the product manager of the aviation solution [00:07:00] at Esri.
And AI was starting to take off a bit more, right? The people started integrating it more and more. It wasn't just sort of an abstract thing that some people used here and there. Now businesses started seeing the advantages of using AI. And even Esri has started setting up their own geo AI group on how we can leverage the spatial analytics piece in combination with AI.
And that's when I started looking at it and what are the use cases that they're able to do. And it just kind of blew my mind with some of the stuff where initially it was just sort of computer vision use cases. So on the aviation side, we want to provide our customers the ability to help, uh, understand their assets better.
Like they're on the aviation realm, like you have runways, you have taxiways, tarmac where all the planes are, they're in constant need of maintenance, right? But being able to, you know, identify. Deformations, for example, like normally somebody, they have to close a runway down, because you have to go out there and visually inspect all these.
What if there's a way that you can do this with AI? It was one of the first things I, when I talked with one of my mentors, Omar [00:08:00] Mehar, who started the GeoAI group, was that he told me that if, if you can see something and recognize it within, you know, a couple seconds, you can probably train an AI to do it.
It's like, okay, well, let's see, I can look at this and see like there's deformations here. There's maybe some cracks on the pavement. Why don't I try to do it myself and see if I can do it? So I taught myself how to get a whole bunch of high resolution imagery, download that, clip it out, identify, just literally drag squares around it, tag every single one that I thought was either a crack or a repair.
Yeah. All these different things. And just manually go through that process myself and figure out how do you do this from the backend side. Once I was able to get that model, I trained it and then I ran it on my test image on a different airport and it sort of coming up with. Here's what I think it was.
And it was like a, whoa, it was able to do that. That was amazing. That was really cool. And so from there, that's how I actually got into the business analytics program at Harvard was really like, okay, I can do this from the [00:09:00] technical side. How can I understand the business side of this as well? How can I leverage this for business?
Right? And so being able to see, have that sort of aha moment, seeing the AI do it to then the next sort of pivotal moment was, okay, what's a business use case for, for AI? How can I integrate that with my customers? What we were making at the time FAA and them. And I wanted to see, okay, could we use like machine learning models to calculate the time it takes to make a chart.
Right. We know how many changes we get. We know how long it takes one of their analysts to process those changes. Can we just build a simple model that says, okay, this month's changes came in, not even using machine learning, deep protected, just using just regular analytics to figure out how we can help improve them, increase operational efficiency for them so that they can better plan ahead, like, oh, this month's got a lot of changes.
So this one's light, we can take this guy and put him somewhere else. Those are some really key moments right there of how I can, you know, seeing it firsthand [00:10:00] of what it can do, and then really thinking about how I can actually apply this to a business case.
Brooks Canavesi: When you think about your day, I mean, director of data products at SAE kind of walk us through what's the day in the life like responsibilities.
And then what are some of the visions for growth? You know, that you're focused on there.
Sean Grant: Well, the first thing on sort of day to day is I always have one on ones with all my team members. That's the biggest thing. Like my team is important. I want to make sure that they know what their expectations are and you know, how they're doing, how, you know, if they're running into any personal issues or whatnot.
And so I have one on ones every two weeks with everybody on my team. My team also, we practice agile for our development of our products. So, you know, we have our daily standups every day and I pop into those really just to see how things are progressing, but also to make sure we're not being blocked by something on, as they're waiting on the ops team, or is it one of our vendors is slowing us up or something like that and really get a sense from the tactical level, what, you know, where we're at with certain things and how things are progressing along.
And if there's any sort of tweaks or anything else, and, you know, it gives them the opportunity to [00:11:00] ask me any questions. Sort of clarifying questions, and then usually after that have other meetings because we service the entire company. That's three different affiliates that we have multiple different groups inside there that are really trying to leverage data or trying to get access to data, you know, they'll come to me and be like, Hey, I need access to this marketing data, this event data, or do you have the standards content, you know, where I can get access to that.
And I usually try to work with them on getting them access to our knowledge graph, but also going out to people and asking them, Hey, where do you have this data? Can I get that data? And I pull that data in. We want to do some interesting products with that. And then also the backside of that meeting with folks who have ideas, right?
Thankfully, more recently, we've seen a lot more interest in AI and a lot of people thinking about, Oh, how can I use, you know, AI machine learning in my daily jobs, or even just, you know, descriptive analytics or predict, or, you know, any color analytics that we have to help, help me. So let me go talk to Sean and see, I have this idea.
Is this. Worth it. Is this a good idea? Can we refine it? And then we [00:12:00] start building that out and seeing, okay, what, from a product perspective, do we need to start building that out and start prioritizing that? So it's a lot of different conversations with a bunch of folks around the company and then, you know, my boss, the CIO, senior leadership and then executive leadership and the other, you know, Teams around there and we all just start coordinating because you know, we've got you know Two main initiatives happening right now where we've got our unity where it's getting data off prem onto the cloud more modern systems Then we have the data program So a lot of stuff moving around and a lot of a lot of people we got to talk to you about what's going on
Brooks Canavesi: You mentioned knowledge graph.
Can you talk a little bit about what has been the evolution SAE from a data perspective? When did you land? At the graph. Why did you land at the graph? And what does it mean for those in the audience that aren't familiar with what a knowledge graph is? I think this would be a really good opportunity for the listeners to understand, how do you make that decision?
When did this come up for you at SAE? Where did you start from and why was it the best solution? [00:13:00]
Sean Grant: SAE started off, like I said, they have this unity project where they want to get all this data that's been siloed and move all this old stuff onto the cloud. But then they have, they said, you know, we've been around for over a century, you know, we produced our standards in the early 1900s, so we've got literally a century, over a century's worth of data that we've generated.
And they start seeing the value, they're like, okay, well, we see value, there's value in data, right? Data's the new oil, as the saying always goes. But we have, you know, three different affiliates and they have multiple different business units underneath them. How can we get all this data connected together so that we can really gather some insights on how things are connected.
So, you know, SAE is focusing on the standards, but PRI audits, the standards, ITC helps create the standards, but everything's siloed in there. How do we start connecting all these people in all these groups together? So that's where. The leadership, uh, CIO came up with the idea, well how about we use this thing called a knowledge graph?
It's basically a structure that allows you to connect data to gather insights from it. People probably don't realize they're interacting with knowledge [00:14:00] graphs and networks every day, like you go onto social media, you go onto Facebook, you go onto LinkedIn, whenever it says, you know, these are people within your network, that is a knowledge graph right there.
It's basically just a node with a, a line connecting you to something else or someone else. It allows you to see these sort of insights. Uh, and how things are connected together or what patterns may emerge inside there. So they quickly realize that, yeah, let's get a knowledge graph going. You know, we can get all of our data, we can get them into our data lake or whatever, but we still have to, like, join it together like you would in a typical, um, infrastructure.
But with the advantage of the knowledge graph, it's already connected for you, right? When you put it into the knowledge graph, the data is already connected. So you don't have to worry about trying to find the different keys or which one goes to where, how does this customer match up to this order or this training course that they purchased.
The knowledge graph allows us to have that information already connected and easily allows us to input and update that data and knowledge graphs can scale large. I mean, we have over a billion objects within our graph right now, and it's growing by millions and millions [00:15:00] each day because we're getting in like web event data and things like that.
So, you know, with knowledge graphs, the evolution of them, they've been around for many, many years now. They've just been off in sort of their own little niche world, right? They're like, Oh, you can do some knowledge graphs. Some companies are doing one. LinkedIn has theirs. Facebook has theirs, but they don't really talk about it.
But you're starting to see, I think about three, four years ago, a lot more interested in knowledge graphs, really to help connect data together as an alternative to data silos. Now, more recently, like you're mentioning, how do, do knowledge graphs play a part in these RAG systems, the Retrieval Augmented Generation systems, part of Gen AI, right?
So typically you have, like you said, vector stores, right? You have some data, you use an embedding model that converts that to vectors, a bunch of series of numbers that the machine understands, stores that into a database. You ask the LLM a question, it takes that question. Converts out the numbers and then compares.
Is this similar to this one or this one or this one? Okay, it's similar to these three. Grab these three. [00:16:00] Go give it back to the user and say, here's what we think is the Now, you know, the issue you have a lot of times with these LLMs that they do hallucinate and they kind of make things up a lot because they're full of information and they're kind of like a child.
They always want to please you. They're like, I will give you an answer even if it's wrong, but I will be really confident in that answer. The knowledge graph paired with your graph rack system allows you to what's called grounded. So our knowledge graph is actual data. Okay. Connected into this system into this network that the knowledge that the LLM can now talk to right?
So you ask it a question? It'll convert that question to a query instead of into numbers to try to search for content similar It takes that query goes against the actual data in the graph and pulls the actual data back out and then responds to you So it's you're starting to see a lot of these graph rag applications pop up Microsoft just came out with a really great paper not too long ago on their graph rag Neo4j who we use for our knowledge graph They have a new [00:17:00] GraphRag Python package, which is quite impressive.
It's actually really good. They've got the ability to do this all out of the box with a lot of the big name providers, OpenAI, Anthropic, Vertex. And you can just simply plug it in, put your key in, ask it a question. It'll convert that. You point it to your, your knowledge graph. And it'll pull the data and get that back out to you really easily.
So that's how we've been leveraging it. But the main thing was that knowledge graph, the decision to try to connect all this data together. And like I said, we just, we chose Neo4j. We prototyped a few others, AWS's Neptune, and a few other different, other open source ones. But we ended up deciding on Neo4j, mainly because, you know, they've been around for a long time.
They're the world leader. They're a graph native database and not sort of a, Take an existing database and we're just going to make it look graphy as some vendors have done, but the graph data science, and now the ability to add vectors into the graph kind of makes it now a hybrid approach, which is really optimal for these graph rack solutions.
Brooks Canavesi: So kind of thinking about from a sales and marketing perspective, is it safe to say like an analogy would be [00:18:00] traditional data structures or like the old filing cabinet where you would have categories, different files that you're pulling out, you know, different drawers that you're pulling out for. These are all the customer files and this is all the inventory information.
But when you go into a graph, it's like having a web where it's interconnected. So you can now the customer information, but also what deals they're related to, what notes. How they've migrated over time, the number of addresses that they have and products that they provide, all of that would be interconnected through a web that you could then search and make those interconnections amongst that.
Is that more? Exactly, a hundred percent.
Sean Grant: Yeah, that's exactly what it is. Because if you take that analogy of the filing cabinets, if you want to see what customers have what orders, you got to open them up, take them out, take the sheets, go through. Try to match them up as best as possible. And then it's like, well, Hey, well, this customer has orders.
Are these orders have invoices? Let me grab that real quick and start like pasting all these together. Right. So in your traditional relational database management system, it's just that it's a series of tables [00:19:00] that are connected, but you have to make those connections inside there in order to gain those insights.
And that's kind of where the graph came about was that's really expensive, especially that's actually how I got into graphs back at Esri. We had a data model that was based off an international standard that was highly, highly relational. And then we tried to implement that into a database. And we were doing queries across nine or ten tables to get attributes that we needed.
And this was very computationally expensive. And so, talking with my developers, I was like, There's got to be a faster way to get this data. It's all related. Shouldn't we have this already, like, pre cached or something? And they came out. He told me about an idea. He was like, Well, there's this concept of a graph database I had learned about.
Take a look at it. So I was like, Okay. Put our database into the graph and a little proof of concept. And I was able to pull this information once I linked it all together in, like, less than a second. Whereas other before it was taking minutes to get this information back. And I was like, see, this is exactly, this is another, I guess, another one of those key moments of understanding for graphs, what [00:20:00] the power of it is.
And yeah, definitely from sales and marketing perspective, we're using that now where we're doing a full customer 360, where we can implement, you know, we have a 15 line query, even less than that. And we can give it a person's an email address and it just crawls the graph, grabs the information, So, tell me everything about Brooks.
What has he done? Well, he's been to this committee. He's authored these papers. He bought these training courses. He's part of this company over here. Um, he's been to these events that we had, and he was a present, presenter at this one, and all this stuff can come back, and it's, it's great. Everything about that person so you can get a full understanding of that customer and who they are or even a company as well because like you said it's a web of connected information and you're able to gather a lot more better insights from that
Brooks Canavesi: so using that example.
data isn't stored in the same system sometimes as who purchased this training, this product. And that's not in the same system as this other piece of data about the customer. And the graph is [00:21:00] able to bring all of that together. And I think that's really important for the listeners to understand is many sales marketing organizations, regardless of your industry.
Have a lot of systems of record, right? And data is dispersed across everything from their intranet to the systems of record. But with the graph strategy, you're able to actually link those together to get that comprehensive view. Is that how it works? Yeah,
Sean Grant: exactly. So we're bringing in data from, you know, Salesforce into, into the graph.
That's the customers that we have. We have data from C vent, which is our event management application. That's coming into the graph. We're connecting those up together. We have data from their table, which is our training one once again. So now you can start seeing how we can start connecting all this together.
We have legacy data and Oracle that we're pulling in financial data is coming from NetSuite. And, you know, for us, we have NetSuite in two affiliates. So we have two different NetSuite instances. We have to pull into to get the financial information with adaptive and the graph lets us connect that all together.
So I can say. What are, you know, [00:22:00] Acme, Forge, company that's in here, what are they doing with their employees? Or I can go and see an individual person, what they purchased, and the graph just lets me navigate through and get all these different touch points. It makes it a lot easier to get this connected information.
And the advantage is that it sits on top of all of our data. our data. So we have the knowledge graph. You can still access all those other data in those native systems. We also pull it together within our data platform, Databricks. And so we have all the tables inside of there as well. So we can still generate reports, sort of a typical table, RDBMS kind of situation, a data warehouse data lake.
But yeah, the graph just sits on top and we feed it, you know, all this connected information and it's able to access all these different systems of a record. And it's actually helped us uncover a lot of like duplicates and stuff we've had across systems. And. It's actually led to another initiative where we were trying to dedupe all the company records inside of there.
Like initially we had, if you went to look at Boeing 168 Boeing customers, well, there's, there's one Boeing company. Why do I have 168 Boeing's [00:23:00] in here? Right? So then you can start using third party data, like say Dun and Brad street, their businesses, data, and these companies, they have all this information about these companies who's a subsidiary affiliate, global parent.
That's got fed into the, into the source and the Salesforce that floats downstream. Now you start seeing, you know, clean data starting to flow through and get reconnected back up into the graph. So it's, it's pretty, it's really cool. It's really cool to see all this stuff connected back together.
Brooks Canavesi: When you think about AI at SAE, you've already done some prototyping.
You've already started working. So you have a good data governance strategy that you're working from in data science and the practice and the knowledge graphs that you've built for other applications as it relates to AI now. How does that accelerate what you can do and what have you tried thus far?
Sean Grant: The first thing is that when we set out to build this, the data strategy. I'm focusing on our three main points. It's people, infrastructure, and value. So with people, we have to one, build up a data team. Uh, that was the big thing for us. Like, all right, we [00:24:00] got this data initiative. I came on board in January of 23 and we had the VP of engineering.
We had one software engineer turned into a data engineer and we had one part time contractor and three. Other contractors helping us out, right? And they're like, Hey, let's build a knowledge graph. Okay. Well, we have some, you know, some AWS instances. We're blowing away the graph every day. We're not updating it and we're loading some data in.
We got this proof of concept. It's like, okay, well, to build this program out, we need to build a proper data team, right? In order for organizations who want to use data driven analytics, you have to have the right people in the right place. And for us, that was building out getting data engineers. getting data scientists, data analysts, and machine learning engineers and building that out.
So within the first eight months I was there, thankfully we had, you know, part of this digital transformation from the CEO was we had to go ahead, build the team you need, let's get it, let's run it fast. Thankfully, we were treated more like a startup within this century old company. Skunkworks. Yeah, like a little skunkworks thing.
And we were given [00:25:00] the autonomy, which was fantastic. And so within eight months we hired 14 people. It was very, very fast. So, like I said, I just did two interviews, people, they did technical. And then I talked with them about what we're doing and what are they going to match with the team? And we were able to get that team up and running fast.
And so, and then the other flip side of that and the people within the company, we had to educate and empower them, right? Hey, we're starting this AI initiative. We're changing the way we do things here. We're using analytics. We're using data to drive decisions to better inform decisions. Need to start learning.
How to query the knowledge graph. Start learning things like Power BI. We're standardized. Hey, if you're building any sort of analytics or visualization business intelligence. We had people using Tableau, Looker, everyone uses Power BI. Stop using Excel, go to Power BI. And so now we got the people, the infrastructure was building out, what vendors we're going to choose, um, Neo4j, Databricks, and then value.
So, we can build a lot of cool applications with AI, a lot of stuff that does really cool, fun stuff, but if it's not translating into value [00:26:00] for the company, we're just burning, burning money, right, at this point, or any value to the people or to our customers. So it really comes down to what are we going to build and why are we building it, whether it's operational efficiencies, whether it's optimizing existing revenue sources or even generating new revenue sources, right?
So a couple of the initiatives that were initiatives that we did was, let's see, you know, we've got a membership renewal, right? Most companies have like a subscription that they have with their customers. And we wanted to see people will renew. Some people won't renew. Could we predict that? Is there factors within the data that we have?
Could we predict if Brooks is not going to renew his membership next month? So let's take in a bunch of these different factors. So we've got data from all over the affiliate, from all over. Are they involved in training and committees? Any bodies have gone into any events and what's their web activity?
Things like that. Just trying to see what factors, you know, how long have they been at SAE? What are their job title? Have they been involved in any other activities there and seeing if any of these [00:27:00] factors would allow us to predict if somebody is going to renew. So we took that as a proof of concept. We were doing it locally.
It's like, okay, Databricks up and running. Let's go ahead and do this prediction. And then let's get this into production. So we were able to. Take this, I'll put a, you know, a pretty high percentage and then got that into production. So it runs every day. Once new data comes in, we get the new data comes in.
It does the prediction. Is this model better than the old model? If it is, okay, replace it. That's the one that gets published out. And we tell the marketing team, Hey, here's your Power BI report. Here's the customer IDs. Go at it. Go have fun with that and go see if you can reduce the non renewals. Right.
They're tying it to value. We can have 99 percent accuracy, but it really comes down to what are the business metrics, right? So if the membership teams target for their metric is going to be, we want 5 percent decrease in non renewals or whatnot for this quarter. That's our, that's our goal right here.
That'll translate to, you know, say 200, 000 for argument's sake in revenue that we can recap, recapture. [00:28:00] Right. So we need to find a model that's going to meet up with their business metric. And that's really what it's been coming to for some of these. Other things we've done, just simple financial forecasting.
Take that event or take that financial data that we've been collecting in NetSuite from the different affiliates and the sales that we've had. Can we just do some time series prediction, take old historical data, see if we can predict out with 95 percent confidence intervals, where we're going to be next month, right?
To get some little bit more accurate predictions and then break it out by business unit, by group, that kind of thing. So a lot of these, and they were able to put into Databricks, publish out to a power BI report, send it to the business unit. They just open up refresh. Oh, what's this business unit. All right.
They look good this month. Let's move on to the next one. Now we have more accurate ones and then a bunch of the LLM work that we've also been doing as well. That's been interesting stuff too.
Brooks Canavesi: Would you have believed when you were doing the work on the tarmac, drawing boxes, that you would be able to do what you can do today with the technology?
Like just in that short amount of time?
Sean Grant: [00:29:00] Not at all. I mean, I just. Maybe I was just a naive about it, but it wasn't until I got to the business analytics program at Harvard because I was just so focused on our little use cases, computer vision, the spatial piece of inside of it, trying to understand that.
And then once I got to the business analytics program, which it was a really great program that covered a good range from for, you know, Professionals who want to get into AI, like we had a bunch of people from marketing, HR folks, they really wanted to use AI. It dug down to the technical level. We were writing Python code, doing our statistical analysis.
They taught us what the statistic models are, what the types of statistics there are, and then backed it up on the top side with the Harvard Business School portion, digital strategy and innovation. How to compete in the age of AI, like how to take this Build a business model and operating model around it.
Value creation, value capture, all these different concepts. I had no idea about. And then in there that a couple of short courses on leading analytics teams and talking about these different capabilities of technology. That's when I was really [00:30:00] introduced first into the capabilities of gen AI. And I was like, this is crazy.
How, how fast this has come in just a couple of years from me tagging some images for a computer vision model to, I ask it a question and it responds like, it's a computer, like it's a rational human being. And the capabilities have just even skyrocketed from there. So no, definitely did not see, see that coming.
Brooks Canavesi: Looking ahead. What do you see as some of the most transformative AI advancements and mobility engineering coming up, like around vehicles, you work a lot, obviously in standards around the space, just speculatively looking out. What do you see as some of the most transformative AI advancements? I
Sean Grant: think it's, you'll see a lot with a simulation, but like on our side, Being able to get to information faster and easier.
Is going to truly speed up innovation. Like we're doing a lot of work, obviously, you know, our bread and butter SE is society of automobile engineers. We started off with [00:31:00] building standards for internal combustion engines. When those first came out in the early nineteen hundreds, I mean, you know, if you have your oil, you look at us as SAE 10 W 30, right?
That's that's us. Right. And we've, we've had to adapt to the times you see. Internal combustion engines are going, but they're still here. But now you see hybrids coming out right now. I see fully electronic vehicles, alternative fuel, um, vehicles, alternative fuel airplanes as well. And so, you know, one of the things is we have a lot of innovation happening in these spaces and the standards are tying that all together.
And that helps to increase adoption, right? So we helped come up with the standard, the J1772 standard for electronic charging. So Tesla had their own standard, but then all the other automobiles, manufacturers were like, Hey, we want to get on this electric car thing, but we don't want to make our own, so that only you could go to Nissan Chargers if you had a Nissan, and I can only go to Toyota once if I had a Toyota.
We worked with the industry to create this. That standard for the charging. And then our recently we worked with Tesla. Now we're [00:32:00] coming out with a new North American charging center that everybody will use. Right. And so you'll see standardization and electronic batteries, right? The more, a lot of innovations happening in the battery space because they want to get rid of range anxiety.
So, so for us really transformative, the things like building gen AI chatbots that can work directly with engineers to have conversations with them, to help them answer questions, get the information they in this obscure standard of that standard. Or even, you know, from events and things like that. So they can help form and get better ideas on how to innovate and work faster.
And then obviously work with the industry and the consortia on that one as well.
Brooks Canavesi: A lot of our listeners are starting their AI journey, their corporation, their organization, whether they're the. CEO and executive, or they're just working in a specific line of business within an organization like sales or marketing.
They're starting their AI journey. What could you give as advice to someone just getting started? Maybe they don't have a knowledge graph. They don't have data scientists on staff. What are some of the key [00:33:00] things to making their?
Sean Grant: That's the big thing is the use cases. How do you want to use or what problems, you know, going back, identifying what problems you're having in your company or ways that you think AI can improve it, right?
There's a lot of, a lot of great information out there. How AI is solving problems from things like membership subscription. There's plenty of use cases out there. You just have to see how that's going to apply to, to your business, to your corporation, right? You know, like I said, with our CEO, he saw, you know, we have, we have Data dispersed across all these different affiliates and these silos.
He's like, I want to get these connected together. And he saw a thing. Knowledge Graph can, can help me with that. I mean, we do some research into that, or, you know, a lot of people are using these gen AI. You know, for chat bots to help people understand their content and stuff better. Is that a problem that we have here too?
Is this an internal or external problem for us? For example, you know, we have people constantly coming to me. Hey, do you know where I can get this data? Or how do I get, you know, what's the answer to this question? How many sales did we have last year? I need to do this quick answer. [00:34:00] And somebody's got the contact this person and gets, you know, net suite data and all that.
Just ask a chatbot, even if it's just an internal tool, right? So the big thing is really figuring out what problems that you have, understanding what the capabilities of AI is, what it can do. Next is building a strong data foundation and data platform so that you have know where your data is at, what quality your data is, and how to get that so the AI can consume it and to utilize that and really build out that product management.
piece as well. It's like, okay, well, I have this problem. There's really no alternatives for it, but there's a solution we can use AI for. Okay, let's explore this use case in this product, whether it's an internal one that we have to help with operational efficiencies or to help automate and help improve people's workdays, or it's a new revenue stream, new, new thing we want to do.
Hey, we got our own little chat GBT. If you want to access all of our content, yeah, sign up. You can get access to all this stuff a lot faster. And that, that's kind of where I've been guiding people and talking with folks about is, you [00:35:00] know, really get an understanding that, I mean, there's so many great resources out there now, like I use LinkedIn learning a lot, Coursera, a lot of free stuff out there as well that you can get access to, to, you know, podcasts like this and listening to podcasts helps just get an understanding of how people are using it and how you can adapt that to your use cases as well.
Brooks Canavesi: As you're creating AI solutions or things are rolling out at your company, what role do you see in getting people comfortable from a change management education perspective? Because AI can be somewhat dualistic as far as people's vantage point of it. Some people have a very pessimistic view of AI and where it's headed and maybe what it means for their job.
And others are over optimistic seeing it as augmentation and the ability to assist them in their role you. As you've rolled a lot of these tools and functionality out as a product owner at various companies. Where do you see, like, the training, upskilling, education, change management aspect of having a successful rollout?
Sean Grant: It comes down to [00:36:00] leadership, right? So leadership really needs to come down from the top, like, Hey, this is what we're doing. We're moving everybody forward. Like, with our case, with the digital transformation, we've got Unity, we've got the data program. They're coming in, we're building AI tools. It got a lot of folks scared, like, Oh, the AI is going to take over my job.
It's It's not that case. I think, I don't know who said it. Somebody said it, but AI is not going to take your job. Somebody using AI will take your job. It's sort of a hard thing for some people to understand. Like, okay, well, I've been doing my job this way for a while. It seems to work. Why do I need this tool?
This tool will make it better, but also it'll allow you to automate and make it more efficient. So you can do other things now, right? You can start learning about other things or grow, go to a different one. So it's definitely important for the leadership to one guide everybody and tell them like, this is the direction layout, a good strategy and a good vision for what they're doing.
And this is the mission that the company is on and this is where we're going, but also provide them the tools, right? So, like I said, educate and empower. For my team, for example, we were onboarding a lot of folks really fast. [00:37:00] And when we made the decision to go with Databricks for our data and analytics, but data analytics vendor, one of the first things I told my boss is like, okay, well, Databricks has their own little academy, a little thing where you can go through these learning paths, get your badge and put on LinkedIn or get your certification.
I want to make that available to everybody here. So let's pay for their first certification and they can go through and get their associate level data engineer, ML scientist, data scientist. Also within the tech space, if you're familiar with the O'Reilly books, the books with the animals on it, great books.
And I've got a few of those at home and I still have a bunch of, I got some, I have some read yet. Unix
Brooks Canavesi: and shell program. Anyone's on my shelf.
Sean Grant: And so I reached out to O'Reilly media and I was like, Hey, we love your books. I don't want to buy one off two off. Do you guys have a subscription service or something that we can get some?
And I didn't realize what they had. The O'Reilly media was fantastic. We did a little trial run for, for 10 users on the team. And we had full access to all of that as well. So I let the team members like, Hey, you know, here's the book on a data catalog. Right. So go ahead and read that. There's [00:38:00] podcasts, there's videos.
Like I was looking. I was watching podcasts that they have on there and videos, and they have events that you can sign up for as well. You know, we've got free LinkedIn learning for every employee in the company. So the education and the tools are there and the everything's right there. And plus, like I said, with leadership's ability, come in and tell me, this is what you're doing.
Here's the tools. Let's go forward, right? This is what we're building. It's not to take your job. It's to improve your job and help improve, you know, set you up for there. So it really comes down to making sure you've educated and empowered your, your team to go forward with it.
Brooks Canavesi: How does that relate to your military background?
When something's coming down as new technology, a new process or procedure in the military, how is that approached from a military aspect? And then how has that influenced your leadership style and how you see things working?
Sean Grant: I mean, it was very applicable because with being in the intel, we got new software.
I mean, things move slow in the military, but within our tech space, we were getting new things probably more often than others, especially with the way tech moves, right? We [00:39:00] have a saying in the Marine Corps, improvise, adapt, and overcome, and that's what you have to do. Right. You got new tech or a new approach or something comes in.
You got to learn it real quick, get on board and start rolling it out. And then, you know, we had people going to advanced training courses back at the schoolhouse. And as soon as they go, they'd come back and distribute that knowledge back to everybody else. So we made sure that if somebody learned something new, like when I was running the surveys, survey teams.
I wanted to make sure that everybody knew like the output of what the work they did because previously you go out you could do your land surveys, you get that data, it's geospatial processing, you just go throw it over to the other side of the house, they make the maps. But I wanted to make sure that they saw from start to finish and so that they would take the data, we'd go over to the other side and we go build the products ourselves.
So being able to adapt on the fly and especially within the LLM space, it's extremely fast moving and it can be very frustrating. I mean, when we were building our LLM last year to talk with the graph, every week a new model was coming out that was going to be better. I remember one [00:40:00] time we built it, I think it was LLAMA 2, and we were generating Cypher, it was doing really well.
Then they announced, hey, there's this code LLAMA, it's better generating code and queries. And we're like, huh? We just did all this work with Llama 2. Okay, let's backtrack. Let's get the new model now. And then, you know, at some point, you're just going to be chasing your tail. You're just going to be going around circles because you're trying to chase the latest and greatest thing.
And it can be very overwhelming. It's just a matter of, you know, sticking to what you have, focusing on the longer term, knowing that, you know, yes, technology is going to advance, right? And that's, that's what I always tell everybody in the Marines that technology is going to advance. Things are going to move along.
You just got to adapt to it. Roll with the punches. And keep going, just make sure you have your strategy and your vision in mind and how that's going to relate to it. And just keep going along that route.
Brooks Canavesi: We get to this point in the show where we talk about the genie question. Our genie only has one wish, though.
So if you could get one wish, and this would be the impact of AI on the future, what would that wish be?
Sean Grant: Safe and responsible. I think that's the biggest thing. People, I mean, we've seen, we've all seen [00:41:00] the sci fi series come out, AI takes over, destroys humanity. I don't know if that's going to happen anytime in our lifetimes.
But. You know, we've seen some of the dangers of AI and some of the unrestrictedness, uh, that happens. And, you know, without the proper guardrails and governance in place, it can get pretty scary and be abused rather quickly. So I would say like the future of AI to ensure that it's, you know, safe and responsible, but helps us innovate and improve our lives.
Brooks Canavesi: Just really want to say thank you so much. Thank you for your service. And thank you so much for coming in today and sharing your knowledge with us. This has been fascinating and, uh, really appreciate the time.
Sean Grant: Yeah. Thank you. This is fun. I love talking about this stuff. So it's good times.
Brooks Canavesi: That's it for today's episode of the sales stage podcast.
Remember the game has changed. Now it's your turn to tip the scales in your favor with AI as your secret weapon. Keep pushing the boundaries and we'll see you next time. If you like what you hear, please leave a positive review on our podcast and share it with your peers. Please also visit [00:42:00] our website at salesage.ai to learn more about equipping your sales teams with AI superpowers.
