The innovation engine behind Samsara driving real-world impact: compounding feedback loops, data flywheels and embedding engineers in customer problems w/ Kiren Sekar #249

started with a vision that these are really massive industries that are vitally important to the world and they're kind of running in the stone ages. pen and paper and IBM mainframes and green screens and phone calls. And it seemed obvious that if you fast forward uh some number of decades that they're gonna be

fully digital and modern and and and automated. And we said there is an opportunity to play a big part in making that happen. You know, our angle is going to be simplicity. There are all these building blocks that exist, but if you're running a a production company, you can't figure out how am I gonna combine

sensors and connectivity and cloud infrastructure and analytics and build a great user interface, you actually need it all in one system. So we said that is going to be our our angle. Like let's make something that actually combines The full stack from capturing the data to aggregating it to all the way to you know solving the the end customer's problem through a software tool.

Hello and welcome to the Engineering Leadership Podcast brought to you by ELC, the engineering leadership community. I'm Jerry Lee, founder of ELC. And I'm Patrick Gallagher, and we're your hosts. Our show shares the most critical perspectives, habits, and examples of great software engineering leaders to help evolve leadership in the tech industry. This episode is all about how you build your innovation engine around real-world ROI.

Kieran Sakar, Chief Product Officer at Samsara, joins us to deconstruct Samsara's innovation engine, their decade-long compound product strategy, and their systems around customer-centered product development and product feedback loops. We start with an incredible example of real-world impact deconstructing the role that Samsara played, supporting their customers during the recent once-in-a-generation snowstorm that just happened across the East Coast. It's a crazy story.

We talk about accelerating product feedback loops, especially in an era where your bottleneck is shifting from code generation to customer feedback and absorption of product change. We discuss how their data flywheel expands the aperture of what is possible to build. And we dive into the system of customer-driven innovation and the power of embedding engineers in frontline environments to cultivate taste. customer empathy and trigger nonlinear ideas.

Let me introduce you to Kieran. Kieran leads Samsara's global product and engineering organizations where they harness trillions of data points and new AI technology to make customers' operations Safer, more efficient, and more sustainable. Kieran's led the product organization since founding and also served as Samsara's chief strategy officer. Prior to Samsara, he was VP of marketing at Meraki, now part of Cisco Systems.

where he helped grow the business from five million to over five hundred million dollars in annual revenue. And before that, Kieran held software engineering and management positions at various startups and at Apple. Enjoy our conversation with Kieran Sakar.

Kieran, first off, I just want to say welcome. Thanks so much for joining us. Uh I think today's Wednesday. Uh sometimes I you know it gets a little blurry, but how are how are you doing today? How are things? You said you were a coffee connoisseur. What have you been drinking today? I've been drinking uh cup of jose, uh, which is from the uh coffee shop half a block from my house. And it's a small local uh uh a coffee shop that was started by one of the early blue bottle folks.

And uh it's delicious and very convenient. So I'm fully caffeinated and and ready to go. I I love it. Well, to set some context for our conversation, there's there's a few things I kinda wanted to to to tee up. before we get into some stories. So one was this meme that you and I were sort of discussing a while back, which was twenty twenty six being called the year where AI sort of has to prove its real world ROI. And a lot of engineering leaders in our community are sort of

of uh feeling the pressure of that sort of meme or mandate or however you want to label it. But I think what's interesting is like samsara, like and how your product impacts customers is a fantastic example of that intersection in generating real world ROI that like literally impacts the physical world.

And people's safety and and like performance and efficiency. So to set up our conversation, I think what's interesting is we're recording this a few weeks after this once in a century snowstorm hitting the East Coast in the Midwest. And I know that this put an incredible amount of pressure on you and Samsara and many of the customers that rely on Samsara to support them in these types of things.

So I wanted to start there in that story. And so can you bring us into the recent snowstorms and specifically, you know, how did Samsara support customers on the ground during that crisis? And how does this story sort of set us up for a conversation around building an innovation engine, customer centered product development, and really connecting to real world ROI with these types of AI-driven products? So bring us into this, this big moment.

Yeah, absolutely. So crazy snowstorms over the last couple of weeks impacted all of our customers.

Before we get to how Samsara plays into this, a little bit of context. I know a lot of the engineering leaders I I've talked to have heard of Samsara, maybe l heard about us on a podcast, but don't really know what we do. So Start with who our customers are. We build technology for the world of of physical operations. So think the logistics networks, energy companies, the supply chains, infrastructure construction companies.

agriculture, food and beverage productions, all of these these businesses with Physical assets, frontline operations doing uh physical work that makes our economy go. Big, big world, forty percent of global GDP, billions of workers, and it's uh fascinating uh set of customers to to build for. Our our products, we start by gathering data from their physical operations. We actually make hardware devices, sensors, cameras, uh gateways. They're all

connected to uh the internet. They go on our customers' trucks and construction equipment and high value tools and assets. We have a mobile app that the frontline teams use. And then these bring data from their operations into our cloud where we use AI to find insights. We have software tools to help them act on them. So that's what we do at a very high level. And really, customers are buying our product to improve safety, improve efficiency, digitally transform their business.

So that's at like a 10,000 foot level. Let's talk about the the storms last week. We've got ice storms, we've got closed roads. This impacts our customers really in in two ways. First is is actually safety, right? They're driving hundreds or thousands of miles per truck per driver in these conditions. They're often operating, you know, 24-7, one in five fatal accidents.

Are weather related. So they're thinking, we're sending our teams out into these conditions. How do we make sure they come back home safely when we know that there's very real risk? Two. How do they actually deliver for their customers? How are they gonna figure out what roads are impassable? Uh, what's gonna be delayed? I normally do

five jobs in a day? Do I need to cut that down to two or three to allow for more time? What we do is we actually give them visibility into the actual environment they're operating in, what's going on, what are the hazards. And then we Use AI to figure out what does that actually mean for their frontline teams on the ground and how do we keep them safe?

So for example, we have weather radar data, we have traffic data, we actually have camera data to know in real time what are the roads like, what's the visibility like. We can figure out how high risk is this. And then we can actually push.

safety tips and safety thresholds down to drivers saying, hey, this is what an appropriate speed is. These are the precautions to kick to take care of. You've got hazardous weather and it's coming up 40 miles down the road. So plan accordingly. So people are, you know, have the tools to keep themselves safe. And then if you're a back office team and you've got a fleet of 10,000 vehicles, for example, and you've got heavy construction uh equipment out at job sites.

You've actually got got you know visibility into what's going on, uh what's getting delayed, what's clear, you know, what's potentially, you know, at risk and you need to go send out another team and you can orchestrate all of that. So we started, you know, getting text messages, LinkedIn messages from our customers saying like

This is this is game changing. You know, we've been uh using the platform for for years. All of this new AI is game changing and actually helping us navigate uh these conditions and You know, we got through the storm without accidents and our team's safe without letting our customers down. And, you know, that is a really awesome way to feel like, okay, there is real world impact and you're actually seeing it in people, you know, coming home at the at the end of the day. My my favorite.

applications or when I think about like the most aspirational and positive implications of AI on on our life, it's these ones that really bridge the gap between software products that just exist in your computer to things that actually impact the choices that you make in the the material like the physical world.

in terms of how you interact with a physical environment or how you interact with other human beings. And so like describing this environment and how it is changing how people operate. It's changing the decisions they make in terms of the routes that they take. make or changing the decisions that they make live, driving different roads and how that's all coming together to shape safer and more impactful choices for like the their goals as a as a business.

to be that like that represents like the most aspirational use case of AI, like in the best way. And that's it's changing physical decisions for people. Just me kind of reflecting on you hearing that. Like for me, I have family on the East Coast and in the Midwest.

You know, Noah, our producer, lives in Chicago. He had been sort of homebound for a couple of days because it was just too dang cold. Um, and so like for me, it's like I I just think about all of my family members in those environments. and how challenging that was. And like to have like that type of support that's changing your decisions in a safer or more effective way. Like that's what it's about, I feel like. Like that's the type of impact that we we hope these things create.

100%. You know, it's it's interesting. It I think it's actually an outcome of building for for this customer base. And you know, as An engineer as an engineering leader, you want to work on hard technical problems, but you want them to be useful, right? You want to actually uh have impact.

And whenever there's a new technology wave, there's so many cases where you work hard, you build something, and it ends up being a fat or a flash in the pan and it doesn't actually get traction. And The beautiful thing about working with customers and operations is they are just incredibly practical. They have really big budgets, but they run relatively low margin businesses. And so they will spend money if something works.

They will not spend money, they will not spend time if something isn't actually translating to results very quickly. As a builder it it it keeps you honest, right? They prevent you from uh going and and spending all of your calories into something that maybe seems cool or looks cool but doesn't actually uh create impact.

Because they're not gonna spend their money on it and uh uh you're not gonna have the uh the the boom that all of a sudden evaporates. And so, you know, we started working in operations before the the current AI wave, obviously, but Having these new technologies, having all this data creates so much potential. It's awesome to have customers to keep us grounded in these are the problems where if you can solve it, we will lean in because it's going to create real world impact.

So when you reflect on this experience during the winter storms, like to you, like how does this connect to the innovation engine that's been built at Samsara or the focus on customer-centered product development throughout?

the builder organization that you that you run, like how does it connect back to to that that area? Yeah, you know, it's it's interesting. It It's an actually a a great example of over, you know, ten years of having a specific way that we build products. And not compounding over time. Our philosophy has always been we have gotta be alongside the customer and working very deeply with them and figuring out what problems to solve.

and making sure we're solving them in a way that's going to be effective. And in a way, this is by necessity. We got really excited about this customer base. and these industries and uh we saw how big they were, how complex they were, but also how just underserved by technology they were. So there's so many problems to solve as a builder, you're like, I want to go, uh, I wanna go get into that, like clear, clear opportunity. But

We are technologists first and foremost. Most of us don't have a background in construction or agriculture or logistics or energy or what have you. The only way for us to figure out what to build and figure out if our products are working is actually to get make contact with that customer.

very, very early on and then run a run a feedback loop with them. And so in the early days, it was actually seeing, hey, we can take a inspiration from Uber and DoorDash with the idea that you can actually have real time GPS tracking. You know, something where 15 years ago that wasn't really possible, but because of cell phones and uh the network build outs that came up came with them.

It gave became possible. We said, all right, let's see if we can bottle that up into a technology that we can put in the hands of of operations. And then let's figure out how do we actually make this useful. We found that quickly people were using it to to manage logistics, particularly when there were challenges, right? Whether that was weather or a customer surge or, you know, whatever was going on. And then it turned into like Hey, well a lot of the times when we're using this data it's

Because there is a weather event, and we said, hey, what if we bring in radar data? What if we bring in forecasts and they have visibility into that? And I said, okay, this is, this is great. And then we started really kind of going down the the the camera rabbit hole, right? As we think about where cameras were getting inexpensive and the actually the idea that she could use your phone to record video and stream it

over the cellular network again, kind of going back uh seven, eight years, we said, Well, could we build a product that would enable that in our customers' operations? And they said, Oh, it's like Now the lights are turned on. Now, when there is a risk of some kind or a complexity I have to deal with.

I can now actually see what's going on. And that, you know, uh obviously translates to how they deal with with weather. And then, you know, you fast forward to today where we're able to say, okay, well, now we have millions of cameras, we can actually source data across all for all of them and create this kind of three hundred and sixty degree view of what's going on in the roads.

We have video language models that can take all of the the image data and understand what's going on. So it's not a human looking at all these feeds. We can automate the uh alerts. And the the training that go out to j to drivers. It's basically these kind of compounding stacks of of new technological capabilities, but it's rooted in this customer feedback. You know, the ability to be out looking in the world saying, what are the new tools in the in the toolkit based on technology advancements?

But then having our engineers actually really understanding the operating environments that our customers are are in and figuring out what are the creative ways we can bring them to bear on customer problems, I think is really what creates experiences like these.

There's a couple ways I want to deconstruct this because there's a few keywords here that I that you you've highlighted that I'm like, ah, we got to dive into this. So you're talking about feedback loops.

You were talking about assessing the emerging tools available in the toolkit and then identifying the best ways to apply that and getting engineers out there in the operating environments to understand the best way to apply and tackle those problems. There's two sides of this that I want to dive into. So one, I want to be in the conversation of the room in identifying these sort of compounding stack.

I guess was it phased? Was it one conversation where you're like, here's our vision, here's sort of the compounding stack, like this is what we elegantly think this will be like, or was it a series of conversations? So what did it look like to identify some of these like compounding stacks and the sequence of it? And then I think the second side of it is starting to understand this feedback loop. So I guess bring us into this this approach.

Yeah. You know, we started with a vision that these are really massive industries that are vitally important to the world and they're kind of running in the stone ages. pen and paper and IBM mainframes and green screens and phone calls. And it seemed obvious that if you fast forward uh some number of decades that they're gonna be

fully digital and modern and and and automated. And we said there is an opportunity to play a big part in making that happen. And where we sort of had conviction was these industries are going to go digital. And it Sounds kind of obvious now. At the time, like there were no Silicon Valley companies that had big multibillion dollar exits building for these industries. So it was a bit of a a leap of faith.

And then we said, you know, our angle is gonna be simplicity. Like there are all these building blocks that exist, but if you're running a a cement production company, you can't figure out how am I gonna combine sensors and connectivity and cloud infrastructure and analytics and build a great user interface. You actually need it all in one system.

So we said that is going to be our our angle. Like let's make something that actually combines the full stack from capturing the data to aggregating it to all the way to, you know, solving the the end customer's problem through a software tool. And then from there it was just Get something out very quickly, get real customer feedback. When we start hearing consistent feedback that uh something is working, we we lean in. When we start hearing consistent

anecdotes about these are the problems that that we're facing, uh, we go and try to solve those. And uh that was the the algorithm. So conviction in the problem space. conviction that we were gonna attack this by making it easy for the customer. But after that, it was very, very feedback loop driven. That's how we really got product market fit across

multiple products in in just a few years and and really scaled. And then as we got more and more scale, some of the the flywheels really started to emerge. I think it's it's

a bit hard to, you know, from a cold start say, here's the flywheel and we're gonna go after it, uh, because you might actually get distracted from solving real problems. But eventually these flywheels do emerge, you know, and for us, There were really two that became clear. One is is a technical enabler, which is the data in our platform, right? We have video data from operations, GPS location data, uh data from engines on vehicles and machines.

digital workflows, all of these things are unique and proprietary and you can't find them on the internet, right? You can't just go train a large language model based on some existing corpus. This data only exists when you have millions of cameras and sensors and and apps out in the world. And so that is something that we built up by solving these tactical problems. And then all of a sudden you realize.

Hey, the fact that we've solved this problem and we have tens of thousands of customers using this product means that we now have enough data to solve new problems in new ways. So that's one flywheel that emerged. The other is actually maybe a little bit more uh subtle, which is the customer relationships we developed and the customer feedback loops we developed, where now when we want to figure out what problems to solve and and if we build something, how much impact is it going to have? It's

It's much easier for us to figure out: is this a good idea or is this a bad idea? If we build something and we want to get uh feedback and test it, we have so many customers who are eager to try it out and give us that feedback. And again, if you're if you're building products for engineers, for example, this might be less valuable because you can put something up on on X.

And uh people will discover it and they'll try it and they will give you feedback. In our market, we're not selling to to technologists. They're not out on the internet constantly looking for solutions. So actually having that, you know, human relationship with them, having them living in our dashboard. So if we build something, we can have a little pop-up that says, hey, we built this, try it out. We want your feedback.

that has become its own feedback loop. So I think these two in parallel um have really become flywheels that are accelerating as as we go, but they came from actually just focusing on on solving problems. We're gonna dive into sort of d the the data evolution and how that sort of unlocked new opportunities. But I wanted to dive in a little bit deeper into your perspective on on the role of feedback loops. We were talking about the shifting of bottlenecks. We have AI, new bottlenecks.

And this demand for faster cycle times and feedback loops. Can you talk a little bit about what you're seeing there and how that's sort of shaping your approach to how the feedback loop is evolving at Simsara? Yeah, you know, if I think uh in in the early days, the bottleneck was actually writing the code to test an idea. And you'd have an idea and actually the engineer's time physically implementing it was so precious.

You'd find all the different ways to de-risk it, right? You might do some market research, you might build uh Figma mocks, etc. And you know, even if you're really, you know, boiling something down to to its its quintessence, you wanted to be very careful with how you were uh um allocating your engineers' time. Now, you know, the ability to make a a prototype is is

basically free, right? And uh the ability to make production systems using AI agents is just so much faster than than than it used to be. And it's it's only gonna get better. So that's no longer the bottleneck, right? What is? It's actually oftentimes it's it's that customer feedback, right? Are we building the right things? Are they getting traction? If they're not getting adoption, you know, why is that?

And, you know, so I I really think a lot about in our position, what are the different ways we can engage with our customer base when the ability to make if we and we have an idea that the ability to make something that is is testable and we can start that feedback loop is is so much faster. Right. And scaling customer relationships

is still as uh challenging as ever. Um, the ability for customers to absorb change is still finite, right? And so I think the ability to actually get things out to a large base of operators. Again, take ourselves out of our engineers' mindset and into the mindset of someone who's running a a project to build a

a bridge or a highway or a data center, right? Being able to to get new functionality into their workflows and get their feedback quickly is gold. Right. And it allows that That uh that iter iteration that I think is gonna be super exciting, uh, especially in the in in the year ahead.

Uh are there, I guess, maybe tactics that you all are doing that you're seeing as ways to accelerate that feedback loop? Like how are you starting to approach that? Maybe what does it look like now? And then maybe where are some of the experiments that you're seeing to try to get leverage or acceleration? Yeah, you know, the the good news is that we we started this a long time ago, right? The idea that we can communicate directly with our customers.

through the app itself, show them the new things that we're building. We have ability the ability for our customers to share feedback that actually goes directly to our product and engineering teams. It's not all filtered through customer success or support. And Those are really foundational. So our team and our customers are actually trained that if customers share feedback, it is it is heard and and and and it'll be acted on. I think actually using LLMs as a way to

synthesize the the huge amounts of of customer engagements and touch points that we get is is also game changing, right? If we think about having transcripts of all the phone conversations, the tens of thousands of conversations that are happening, you know, every day at at scale and being able to find, oh hey, here are customers who are really interested in capability X. So if I have something new that I'm working on, I can be very

targeted in who I'm I'm getting feedback from. Like these are really kind of new, new frontiers that would have been very difficult, you know, a handful of years ago, but actually are really powerful in just accelerating that that loop.

I think what was really intr particularly interesting about that was you sort of laid out all of the different tactics and the different things that Samsara has done to to solve specific problems that then sort of created this data flywheel loop. How do you use that?

expand and answer this question of what can we build now? And so what does it look like to take something like that that took 10 years to sort of generate this unique insight around? And then how is that shaping how you're approaching this next question of of what do we build next?

How do we expand what's possible to build here? So t talk to me a little bit more about like that part of the innovation engine. Yeah. Well, you know, it starts with customer problems, right? And really understanding what cup problems customers want us to need to solve. And in our markets, these These problems are often pretty evergreen. Like they're focused on driving accidents and injuries down to zero. These are still the most.

Dangerous jobs that we do in the industrialized world. And they have big safety departments just focused on that. Something they've focused on for decades, for centuries. Like, how do you make these jobs safer and safer? So we know that is always going to be a useful problem to solve.

They also really care about efficiency. These are asset intensive. They spend, you know, hundreds of millions of dollars a year buying trucks and machinery and equipment. And if they can find ways to get more utility out of those investments, they are all in. Same with fuel, right? They uh some of our large customers have a hundred plus million dollar a year fuel budget.

And they're also trying to figure out how do they bring down their carbon emissions, how do they reduce their spend. And so if you can find ways to to save fuel, you will uh you will be rewarded, et cetera. So we start with customer problems and then we think about what are what are the ways that That we could solve a problem that we couldn't have a year, two years, three years ago. And sometimes actually the answer is.

Do the old do it the old fashioned way uh because that's still the straight most straightforward and it's effective. But going through the exercise of saying if you had to use uh an LLM or an AI agent, or if you had to use the scale of data that we have, if you had to use the footprint of assets and vehicles and and mobile apps out in the field, how would you solve that differently? And that forces you to think, all right, like what are the new unlocks?

You know, it almost forces that that that beginner's mind. And I'll give you a a great example. Customers have always wanted to keep track of their assets. You want to know where your stuff is. And um we really kind of started with GPS trackers primarily for truck. Think about like The GPS in your phone, connected to a modem, connected to the cellular network. And you put it in a little black box and it needs power and so you you could

It it works great to put in, for example, a a vehicle. And there's a you know cost to that hardware, there's cost to the the connectivity. And so it's great, but you can only use it on you know relatively high value equipment like a like a truck.

uh that has power. But we actually knew our customers wanted to keep track of their containers and rail cars and dumpsters and chemical tanks and, you know, their tools and surveying equipment and big reels of fiber and stuff like that. They wanted to track all of that. And as the uh technology improved, we said, actually you could

probably take that that device and run it off of a battery because the networks are getting more efficient. Like we went from, you know, three G networks to four G LTE networks and and and five G and and new cellular standards ca came online and said

if we had to use these new cellular protocols that are much more power efficient, how would we do it? Oh, well we we would make it battery powered. And so we built a battery powered tracker. And that was was great. It could kind of check in a couple of times a day, keep people could keep track of stuff. Now, fast forward a a couple of more years and Apple AirTags have become a thing, right? And those you can track your your keychain, your your suitcase.

And it works for for two reasons. One is you got like the bear tag chip itself, which is uh Bluetooth. And those are cheap and uh the battery can last for a couple of years and you can actually, you know, make those at a low enough cost. You can stick them on your keychain. But then, you know, you've got the you know, billion or whatever iPhones out in the world that can pick up all of these air tags. And

Apple couldn't have made that product without having so many people on iPhones. And we said, well, hey. Wait a minute. We have a similar dynamic here. We don't make phones. But we have all of those little black boxes in trucks. And we have all of those battery powered uh cellular uh devices on trailers and containers and you know, collectively there's millions of these things and they're kind of everywhere that that our customers are operating. Could we do something similar?

And so then we kind of went down the path of making basically the uh Sansara version of a a Bluetooth tracker where we have a industrial grade, ruggedized, uh super high power Bluetooth uh with our own security protocols on top. We make it such that you can smash it with a hammer and it'll survive and the battery will last, you know, three, four plus years. Um, and we tried it out and it worked.

And customers are saying, this is amazing. There's stuff that I never would have been able to have visibility into. But now because it's tiny, it doesn't need a big battery, and it's it's relatively low cost. And by the way, it doesn't just check in, you know, once or twice a day. I'm actually picking this thing up every 10, 20, 30 seconds. Now I can know where my uh reel of fiber is that I know they tend to get stolen. Or now I know I have these.

specialized tools and they sit in the back of a pickup truck and I stop for gas and someone steals them out of the back. Or I have a team that goes to a job site and they're ready to do work and they're missing that one part. Now I got 25 people sitting around idle. Now I can actually see all that. The clarity in the problem the customer wanted to solve was actually very consistent, you know, we

identified the problem a decade ago, but it's this series of iteratively saying, Okay, how could I solve the problem today? How could I solve the problem today? How could I solve the problem today? And thinking about, hey, actually whether it's one of these flywheels that we have of scale or whether it's

LLMs or chip capabilities or all the different things that are are changing in the world around us, we find new ways to solve problems and it ends up kind of creating these uh these nonlinear breakthroughs, these these step functions.

Yeah, I I really appreciate laying out laying out all of the ways that this impacts that decision. The the other side of this innovation engine that I was really excited to get into, because you you mentioned, you know, one of one of the models or one of the core parts of how you all work is getting engineers out there and to understanding the operating environments of of the customers. And so I was wondering if we could talk a little bit about like what that looks like.

Like how do you actually get engineers and I and in this context, I mean like it's not just software engineers, it's hardware engineers, it's embedded systems, it's sort of this really multimodal product system. How do you get people building those products to really get deep into the context of the customers' problems and own that customer feedback loop? Like can you bring us into from an engineering organization and the different stakeholders there? Like how you do that. Yeah.

You know, of course our our product managers, our our sales engineers, they spend a lot more time out in the field and they can bring a lot of great data and context to the engineers. But there is no substitute for the engineers actually having empathy and and and having like real context for the environments that customers are operating in. And when they create that mental model, they can make decisions.

so much faster. Sometimes that's actually coming up with solutions that a PM couldn't, right? Because They're in the code, they understand the implementation. And they say, hey, I actually understand, really understand the customer's environment, what they're trying to do, everything else trying to do. And I realize that we couldn't, while we're solving problem X, we could actually solve problem Y and Z. And that's going to blow the customer's mind.

Or I really actually understand the the constraints. And so while the PRD might say this, I know that to make it work well, it's it's gonna have to Solve those those problems. in a way that actually the customer can tactically uh take advantage of. And so if I had a magic wand, I would find a way to have every engineer be able to teleport to a customer's site for half an hour, you know, every, every day or two. Haven't haven't figured that out yet.

So we say what's the next best thing, right? We do try to get engineers uh at least once in a while to be able to go on site and visit a dispatch center or a truck yard or a rail r a job site where they're building a new data center and you know actually see how the frontline teams are using our products.

the jobs they're doing, what are the things that are making them say, wow, this is amazing? What are the things where they're actually fine with it, but you know, they're having to do five clicks to do one thing and we could maybe automate it. Right. And getting that

uh uh that mental model for the customer's operating environment. But then, you know, Zoom is is incredible. Our customers are happy to get on Zoom and uh they will screen share uh their their their device or or their dashboard and show what they're doing in our system and other systems and there's no reason uh that those have to be just you know product managers and getting engineers on those calls, being able to to ask

Ask questions, being able to show their work in progress. It creates, I think, um better ideas. It creates more pride in work and and engineers, they all love it, right? It makes the work more fun. You're actually building for a customer that that you have a connection with, you know, versus building to a spec. So it is a it is I talked about all the ways that our environment is awesome, the customer feedback. Uh the opportunity for impact.

Right. This is the challenge that we have to overcome, which is that we are not the customer. Uh we have to go out of our way to kind of create that that that connection. Um, but it's really important. And um I think for a lot of the folks who who end up working here or working in companies in kind of a similar environment to Simsara, it's because they're actually excited about, you know, getting to take a little bit of their time and learning how the world works.

Sort of getting your speculation and how you might address this systems challenge or if you even think this will be a systems challenge. Because related this question of like the value of an engineer having context and understanding constraints and like when they have a much better

like almost like subconscious sense of the problems that people are facing. Like decisions are faster and the instincts are so much better. And so I'm trying to balance that with the some of the emerging research coming out. Anthropic just released research this week about when people are engaging with different AI tools in certain ways, like the cognitive load and like the

persistence of memory like tends to decrease. So what they're saying is like educators need to change how they educate because if people just rely on some of these different models, like their memory retention or ability to engage with that material later on is like, you know, a little bit reduced.

What I'm trying to figure out is how to build a system that balances some of like the efficiency and like understanding these large-scale problems and trends with then the value of this like individual personal I learned this by being in the field sense. Like how do you build a system that balances both those inputs? Um, is is sort of my speculative question. Uh is like we're starting to understand more about like

how our cognitive functions sort of change when using some of these different tools. So I don't know. That's why it's a speculative question. It's like I how would you how would you tackle that? You know, I I actually don't see these being in in tension. I actually think they're they're very complimentary, right? Because if if we think about it, LLMs cannot yet Intuit what's happening.

you know, end to end on the ground in an industrial operation, right? Because it is not data that they're trading off of, right? So they can get pretty darn far, but there is is no substitute for actually having a mental model for the customer's environment, right? And how they think about

about the people in their organization, how they think about their end customers, how they think about risk, how they think about change and adopting new tools. That's the part that the AI can't can't do for us yet. But if Agents are great at at writing code. You know, what is the the job of the builder, right? Is to be the systems architect, to be, you know, setting the direction of of what problems we're we're trying to solve. And then it's that element of of taste.

Right. And if an engineer can bring that taste in by virtue of actually having that customer context and understanding, it actually just makes Their capabilities like that much more effective. So I think this idea of customer empathy, customer context. actually amplifies the benefits of coding agents and and all of the other tools that are are changing the way that we're building.

I I think that helps me better understand some of like'cause I think like the part that I always struggle like when we're talking we we just had like a dinner yesterday with different uh executive leaders and this always sort of comes up as sort of squaring how do we create this sort of co co piloted relationship. where there's both like your first person sense of customer and then also leveraging some of these these different systems.

And so like as you're kind of describing this, like the other concern we get is like also when you have like less experienced engineers that maybe you have less of a uh understanding of like what's the technical output of of some of these different tools, how do you even begin to assess that?

For me, I'm I'm trying to uh what I'm trying to figure out is like what is that what does that future of co piloting relationships start to look like? Um and then how do you ensure it's like achieving these these different types of things?

I think the all of that is i is true. And of course, like the models are just getting better every twenty minutes at this point, right? Uh the one thing that I think we can be pretty sure of though is if you can have a uh Um you know, a swarm of agents writing code for you, a thing that's really gonna slow you down is having to attend more meetings. Right.

I mean, and so if you uh actually have that uh intuition for the customer and the problem you're solving for them, and that can mean that you can make decisions while sitting in in less meetings and having to get Uh fewer people uh connected to coordinate, you're gonna be able to build better and iterate faster. So you know, I think that's That's like ultimately where where this empathy comes into play.

And if you think about the very old school way of developing where you've got a a six month uh waterfall approach, well actually the tax of having a a a PM really understand the customer and the engineer take requirements. Maybe it wasn't so bad, uh, you know, twenty, thirty years ago, but you know, it's not gonna work now.

So to sort of recap a couple of things we've we've touched on. So we we've sort of talked about this innovation engine accelerating the feedback loops. We've talked about uh the data flywheel that can sort of open up new questions for what you ca is possible to build. And now we were talking about building the system where engineers can get deep into the experience and the customer problems.

The other side of this that I was really interested to get into was how to use your current customers as an asset and how to scale to them quickly with some of these new different things. And so I was wondering if you could talk a little bit about what that system looks like and how you sort of use that to drive the innovation engine at Samsara.

The the the customer base is is amazing. They are so, so valuable to to how we build and and what we can build. So we have a a customer advisory board, uh, which is You know, a f a few dozen of the uh executives at the largest most complex operations i in the world and they're able to give us

kind of from a strategic perspective, what's changing in their industries, what's most important to them. And we get to take that feedback. We also get to show them work that is oftentimes very early, but we can get signal from them. And then we we run what we call

Spark sessions where we've got, you know, small informal groups of everyday users and we can, you know, meet with folks who are dispatchers at food and beverage distributors or safety coaches at construction companies and we can show them what we're building and and get their

tactical feedback and they say, hey, this is what I love. This is what I'd like to see. And by the way, yes, I would love to try that as soon as it's in ready in any usable form. And then we have, you know, the the the customer data at

scale, right? Where we're getting feedback that any customer, any user of the system can submit. Now thanks to tools like uh Dong, we have data from all of the conversations that our our customer facing teams are having with customers. And so So all of that is is just gold in in figuring out, you know, what to build next.

What do we need to improve? What are the new problems to solve? And then so then engineering in the loop for this, like what what role does the engineering organization sort of play in sort of the downstream impact of all these different sort of product inputs?

Yeah, we we try to keep engineering as plugged into all of these as possible. And, you know, that can be, you know, having some of our our staff engineers or engineering leaders attend our uh advisory board sessions or IC engineer actually just show their work at a Spark session. They get to hear directly from customers.

what they like, what they'd like to to see improved. And then of course all of the data is accessible to anyone. So you don't have to go through gatekeepers or different functions if you want to understand what customers are asking the support team about. And you can just go run a query and and and and see exactly, you know, what what you're looking for and even uh see the recordings of the customer calls, et cetera. So it's it's pretty magical.

Kieran, you've taken us on a a journey of the Samsara innovation engine. Uh why I was really looking forward to this conversation and like reflecting back on like this was extremely valuable is when I think about like the future and the w the way that the role of the engineering leader is changing. Engineering leaders are sort of bleeding into all of these other disciplines. And so being able to understand all of the different components that drive an innovation engine like this.

and how to strategically involve engineering and get people close to customers and bring those into how the like what you can build and how that gets built is extremely important. And so getting a sense of like what this looks like behind the scenes at the scale of Samsara. at the real physical world uh problems that you deal with, to me, is just is just a ton of fun. We've got some rapid fire questions if you're ready to jump in. Let's do it.

All right. First question. What are you reading or listening to right now? Uh, let's see. I've been uh going down the rabbit hole of a bunch of uh Rick Rubin's interviews with uh different artists and and and musicians. I'm a guitarist as well. I heard his interview with Rick Beato, who is a a a music YouTuber, and I found that he has all of these interviews with all of these music legends that are just awesome for music nerds. So I've been listening to both of those in parallel.

Along with a bunch of tech and business stuff, but uh that that's what I wanna uh shift my mind to something different. I'm here for the Rick Rubin recommendations. My dream would be to create a recording studio in an airstream trailer, uh very much like his first you need the the beard. Uh that'll be a lifelong quest. Second question, what is a tool or methodology that's had a big impact on you?

Oh man. Um I I can't help but just say that the the tools that I'm using now that uh, you know, I don't know if I'll be using six months from now, but they're just they're they're so amazing. Uh I become addicted to whisper flow and I feel like personally insulted when I have to type. That that that's probably my my my favorite one right now'cause when I can't use it'cause I'm in public I just get irritated.

I I'm definitely on a quest to to dictate more. For me it's like a human behavior change challenge to like shift. And every time I catch myself, I'm like, I need to be doing that. And so this is this is really good validation. So so thanks for reinforcing that.

Um, it's been it's been a ton of fun. Um I think I guess, you know, to to wrap it up, Karen, I I just want to say like for me, like a lot of what I see within the our way Sumstar operates is what I hope to be the type of impact that these different AI tools have in that it's real world material impact.

Uh and so I just appreciate you giving us an insight into how y'all build and and how that then translates in especially like extreme moments. And so it's been really fun to sort of dive into that and to understand all the different systems and processes and workflows and conversations that help drive that. So thanks for an incredible conversation. Thank you for having me. It was a lot of fun.

If you're listening to this and you're wondering, how can I connect with other engineering leaders in my city? Pull up your phone right now and go to elc.community, click our chapters page. You can see that on the menu on the left. Find your local chapter and click join. We're hosting

Virtual and in-person events all the time. And this is the best way to help you get involved, expand your network in your city, and support your leadership and career growth. So pull up your phone, head to elc.community, join your local chapter, and get involved. A huge thank you to all of our local leaders. to make community happen. And thank you for listening to the Engineering Leadership Podcast.