1. Podcasts
  2. Dev Interrupted
  3. You can’t blame AI for the code you ship
Dev Interrupted

17 Nov 2026 · 37m · 14 chapters

Episode artwork

You can’t blame AI for the code you ship | Freshworks’ Murali Swaminathan

In short

Murali Swaminathan (CTO, Freshworks) explains how AI coding tools changed engineering at Freshworks, why “more generated code” didn’t speed shipping, and how they reintroduced discipline via an “AI PDLC” (AI product development lifecycle). He argues humans remain accountable for code quality, so Freshworks focuses on review/testing automation, guardrails, and end-to-end ownership (design/architecture/testing/security checks). He also covers how Freshworks earns trust with non-engineer customers by measuring impact/outcomes (e.g., ticket deflection, first contact resolution), not just adoption.

Guest background

Murali has 30+ years building enterprise software; previously ran engineering at ServiceNow; now leads global engineering and architecture at Freshworks. He’s on the Forbes Technology Council.

Key claims

AI can’t be blamed for shipped code; code proliferation created review/test bottlenecks; engineers must think like product builders and systems thinkers; adoption ≠ impact.

Notable examples

Seagate migrated ~30,000 employees in ~3 months; after deploying Freshworks AI agents, ~30% of tickets were deflected and first contact resolution improved.

Written by AI. May contain mistakes.

Chapters

0:45

Murali's Background and Freshworks Overview

Murali shares his background and the innovations at Freshworks.

“Joining me is Murali Swaminathan, and he's the CTO at Freshworks.”
1:50

Challenges of Engineering with AI

Discussion on the dual challenges of using AI internally and for customers.

“of all this technology disruption, another technology disruption, as I would say it.”
3:50

Engineering Organization and AI Integration

Murali discusses how the engineering organization operates and its changes.

“But then you also have to imagine the art of the possible of how are all of these industries we serve and all of these types of clients and very unique businesses, which are completely different from each other.”
6:10

AI Product Development Lifecycle

Introduction of the AI PDLC to streamline coding processes.

“Because, you know, your customers, they're actually going to take AI as like a bludgeon to everything.”
9:40

Navigating Code Proliferation Challenges

Challenges with code proliferation and the need for maintaining quality.

“We are still working through all the kinks, but that's been fascinating to see how we are transforming from just generating code to building code that really works.”
10:40

Empowering Engineers with a Product Mindset

The evolving role of engineers to adopt a product-focused approach.

“What are the new ways that we like, you know, think about our process?”
14:00

Becoming a Systems Thinker

Learn the importance of systems thinking in effective teamwork and collaboration.

“Everyone has to become a systems thinker.”
15:47

Trust in Non-Tech Products

Explore how trust affects product adoption among non-technical users.

“And, you know, this has been a really great view of how like your own engineering org has transformed during this really like rapid time of like innovation.”
19:46

Adoption and Outcomes in Software

Understand the relationship between software adoption and measurable outcomes.

“and bringing them to the table, keeping them at the table is a completely different operation when in terms of trust, you know, you are talking about a very different audience.”
23:28

Challenges in Technology Transformation

Discuss the evolving role of CTOs in managing rapid technological change.

“But also, what I love about the story with Seagate is that that's a huge shift.”
Show all 14 chapters
28:06

Evolving Engineering Challenges

Learn how rapid cycles in engineering demand new skills and communication.

“And in this whole process, like we're dealing with the bottlenecks that make those fast cycles possible.”
29:50

Customer Engagement in Engineering

Discover the importance of direct customer interaction for engineers.

“We talked about there's no longer tribal knowledge.”
33:08

The Role of Engineers in Customer Communication

Understand how engineers translate customer needs into technical solutions.

“But no matter what, there's all these different lanes by which feedback trickles back to the engineering org, by which engineers get closer to customers that are having really complex or really interesting problems.”
34:36

Learning More About Freshworks

Find out how to follow Freshworks and their engineering practices.

“And so it's been really fascinating to dig into your head and think about some of the more emerging topics around your own engineering org, but then also how you guide your customers through AI adoption.”
A man walking through an airport, asking his phone a question

Ask about this episode

Get the gist, explore an idea or ask about something said.

Claude

On Dev Interrupted, what are the key points of this episode?

Searching your listening

Transcript

Automatic transcript. May contain errors.

0:02Welcome back to Dev Interrupted, brought to you by Linear B. My guest today is Murali Swaminathan, the CTO at Freshworks. Like so many engineering orgs, Freshworks opened the gates to AI coding tools, and being agentic quickly became the norm. He explains that more code did not mean shipping faster. The cost moved downstream to whoever had the review and test it. So Freshworks is putting the discipline back in place. How do you do that without slowing people down? What does an engineer own when writing code is the easy part? And if code is no longer the bottleneck, then what is? We get into how Freshworks earns trust from customers who are not engineers and why Murali measures impact over adoption.

0:45Now here's my conversation with Murali. Joining me is Murali Swaminathan, and he's the CTO at Freshworks. And he's a member of the Forbes Technology Council. Morali has a really impressive background at some of the largest and most recognizable companies. And he makes at Freshworks the customer service, IT experience, CRM, software that thousands of companies run their front lines on. And it's all made possible by Freddie AI agents and innovations on their team, which we're going to dive into today. And here on the show, we've had some amazing opportunities to look at huge companies and how they've made these really bold bets on agentic assistance and providing these really premier customer experiences.

1:27And we're going to look at how it's transformed Freshworks. And so Murali runs the global engineering and architecture teams behind all of this. And Murali, we're really excited to have you here. So welcome to Dev Interrupted. Thank you, Andrew. That was a great introduction. I just want to add a little more color to everything that you said. So it's a great time to be here and great time to be in the middle of all this technology disruption, another technology disruption, as I would say it. I've spent 30 plus years in the industry building enterprise software, scalable, that's being used by many customers around the world.

2:04Prior to service, prior to Freshworks, I spent some time with ServiceNow running engineering for many of their products and then came into Freshworks because I felt the opportunity here is much multifold because we are the AI-powered unified service operations platform used by 75 ,000 customers and counting, which is fast to deploy, intuitive use, and then enables every employee to be more productive. Everything that you do offers powerful governance and scale without the operational drag of legacy platforms. And then we have customers like Databricks, TaylorMade, McLaren, who we are a partner of, trust, fresh service to deliver quality employee service and then manage their technology operations efficiently.

2:58So I'm really glad to be on your podcast. Thank you for inviting me. Thank you for providing all that extra context there too, because this is really what I'm trying to, the nut I really want to crack into today, because you face a really big engineering challenge. You create the engineering world that is approachable by so many other businesses that, you know, they aren't a tech company. They're a company that does something and they have to plug into the rest of the world. They have to be able to leverage modern technology. That way they can have the same kind of edge on their competition that we're always seeking to leverage just as engineers with AI itself.

3:38And so you have this like dual mandate and it's a double challenge we're going to talk about today. where you're under this immense pressure to use AI internally to transform how Freshworks does Freshworks. But then you also have to imagine the art of the possible of how are all of these industries we serve and all of these types of clients and very unique businesses, which are completely different from each other. How do we create a platform that's unified, that's secure, that meets them where they need and is ultimately like composable in their hands, flexible so that it meets them on demand, but then also backed by all of the engineering prowess and technology that Freshworks is bringing to make it stable and something you can build your business on.

4:27That's a really hard double challenge to kind of juggle. And when we've talked with leaders on the show that have been in that position, We've learned a lot of really interesting ways that they even think about AI adoption and impact within their orgs. And one of the first things I would really love is just to get a sense of maybe the engineering org that you're a part of at Freshworks and how it operates and how it's changed this year, like so many engineering orgs have. Yes, you are, Andrea, absolutely attracted there. I think that's exactly the challenge that we are in right now. It's great to have 75 ,000 customers and counting and more.

5:09That means we have the responsibility to support those customers with their business needs and solve their business problems. And also innovate in parallel and then leverage the power of AI to deliver more to them so that they can run their businesses efficiently. So from an engineering problem perspective, yes, our core engineering foundation is key. Like, what are we doing? We have to do it better. And then we have to leverage AI in many ways. The way we are building it, the way we are deploying our features, the way we are enabling our customers through features. We don't want it to be a bolt-on.

5:46So we want it to be where customers are seeing it as part of their business flows. And it's implicit in many ways, right? It doesn't have to be explicit. So that's why it's important for us to not just bring technology into the product, but find a way to bring it into their business process and make it easy for them to adapt it. And so that's one thing that we have to think from a product engineering perspective when we're trying to solve this problem for our customers. Yeah, absolutely. Because, you know, your customers, they're actually going to take AI as like a bludgeon to everything. I like how you said it's like implicit and explicit and so many of the interactions that they then start to do because that's really the new normal that we have to settle those kinds of customers into.

6:31This is the way that we'd work down. This is the way you're going to get the most out of our platform, out of what you're doing. This is what your competitors are doing. Getting other people in this mindset just even outside of the tech world is a challenge in and of itself. And, you know, all of that's made possible by having really clear, I think, language and internal process about how you think about like from beginning to end within your SELC, how you leverage the tools you do, what AI agents that you use, you know, how you run your code. What are your bottlenecks that pop up now in this new world?

7:04And then critically, how do you as an engineering leader rise to that challenge to fix that bottleneck? So like a big obvious bottleneck that's like slammed a lot of teams this year are bottlenecks around like review, right? It can also be things for teams around having the right kind of boundaries and places for agents to get the right context. I'm curious, like with the scope that for the things that y 'all are building at Freshworks, like what emerged as early bottlenecks and what were some ways that y 'all as a team started to get past them so you could continue to deliver value? It was easy, right?

7:40When we got an influx of all these coding tools and AI ways of working, it was easy to just pick up any tool and work with it. But then we had this plethora of tools that everybody was using. Everybody was using it. Everybody was using white coding and generating code and generating a lot of code. Two people writing a similar code, that lot of duplicacy there are lots of code. Anything that can be done in 10 lines is not being done in 100 lines because the code generation is much, the code that the AI tools generate is much more than what a human would write. And then it gets into the next stage of, oh, who's going to review all that code and who's going to test all that code?

8:21So we went through the same cycle where when we introduced it, I would say more than 12 months back, it was all like, oh, everybody go use it. Let's try it. And that's how we opened the gates. But then in a few months, we realized that everybody was using it. But then it was so hard for us to say, yeah, just because you produce a lot of code doesn't mean it's the right code that's helping us ship products faster or whatnot. So we had to streamline it. And so we came up with this whole AI PDLC, AI product development lifecycle, to streamline how we do things. How do I build a little more discipline into how you build code?

8:57Because just because you're using AI tools doesn't change the way you're building things. You still have to follow a certain software development lifecycle because you are accountable to your customers for the software that you share. I cannot blame AI for the software, the code that was shared. At the end of the day, the human is responsible. The company is responsible. So we had to put those practices in place. So we did put a lot of stress into how we build, what we build. So it's not just the code. Code is just a piece of the thing. I think it's all about the thinking that goes into building that code.

9:28And then after the code is built, do you have the right automation in place to test the code, validate the code, review the code, and then deploy them? So it is definitely a journey we are in. We are still working through all the kinks, but that's been fascinating to see how we are transforming from just generating code to building code that really works. Yeah, you're describing the journey. So the landmarks so vividly and clearly and they may line up so so well to what we hear from other leaders as well about how first is about to get a lot of tools. Let's let's try everything. Let's just get at the stick.

10:04Let's see what's going to work for us. And then you get to this stage of code proliferation of we're getting this verbosity that we weren't doing before. We're getting code that's harder to maintain. Then you get runaway costs. As an industry, we went through the whole token maxing and token minimizing and everything in between. And really, it's just been about trying to understand the ROI. You know, we all made this bold bet a year ago to, like, pick up these tools and change the way we work. And for many teams, it's paying off. And, like, teams are just doing so much more. I mean, achieving really amazing stuff.

10:39And because of that, now we have to look at like, how do we optimize that for the future? What are the new ways that we like, you know, think about our process? Y 'all approach this in a very like distinct way of acknowledging that the SDLC and the product development world and, you know, code and owning code and owning product, like these ideas smashed together. And maybe y 'all experience that. But what I've learned from talking with a lot of like product leaders on the show is that they're more engineering driven than ever. They're driving more tickets, more stuff across the line. And they're people who would typically operate in like a team of three with like an engineering lead.

11:21And they felt like they're now encapsulated almost as one. And it's not that there's less people there, but then maybe those other two people are also specialized. They can all three come up with a product. Like they can all three get the stuff, you know, developed and then out the door. And then they can all three turn around the marketing and explain what they need to do with it. So you get this like really change in skills. What have been like skill, what has skill growth been like at Freshworks for your team? What have been like the new norms for engineers day to day as like they work in this new kind of agentic world?

11:54Yes, I think it's definitely all of this that you mentioned is exactly what's going on in our minds as well. while we're building this stuff. And then naturally, we are no longer expecting the engineer to just go. We are expecting them to do design thinking, architecture. We are expecting them to do a lot more experimentation. We are experimenting to build a lot more prototypes. We are expecting them to think like a product manager, think like a designer, because all of that can be done. It's no longer that I'm a back-end engineer or front-end engineer or I only am an architect. Those lines are blurring up right now.

12:29So there is no longer, so an engineer who knows the product should be able to do a lot more because coding should not be the bottleneck anymore because they can apply the knowledge of the product to build better products. That means that they can double up as a product manager because if they know the domain, they know the product, they know as good as the product manager. They shouldn't have to wait for a product match. So that's exactly what we are trying to encourage our engineers to be a lot more, I would say, have that product builder mindset where they're thinking about building the product end to end.

13:03It doesn't mean they're taking the entire product end to end, but whatever they own, own it end to end. Don't rely on a tester to test your product. Don't rely on a security person to make sure that your product, your code is secure. A lot of the stuff you can do it yourself and you're putting those checks in place. That's why we have created skills and reusable things for our engineers to be guided while they're building. And then we're putting those checks in place as part of review and all of that. That way, if they miss something, the reviewer kind of points it out. All right. And if it's, if it misses some of that stuff, there's a check somewhere.

13:39So there's always a check in the system at some point for them to catch those things and, and learn from that experience and do it better next time. It's not like the, we want, don't want to repeat the same mistakes over and over again. So we are giving them guidance to make sure that, yeah, do it right the next time that way, and then get it better and make it better. Yeah, iterate on the system. Everyone has to become a systems thinker. Those become the leaders, the most productive folks that stand out, or the systems thinkers who can take the problem, break it down into pieces, compose them into loops, find the collaborators, the other DRIs, the people that need to be a part of it, but also to identifying the domain experts within their company.

14:23Who would make, I can set up this loop, but who's going to make this loop perfect? Who's going to iterate with it and catch every problem and really make this something that we can carry forward? Because that's what then sets the difference between a disposable practice and a durable one. Because otherwise you end up in this phenomenon, which you so succinctly hit on that I loved is that it's been over a year of teams adopting these tools and using them. And what you have are a bunch of individually highly agentic folks that are scattered around, or maybe it's like one person on a team and then their team is super agentic.

15:00Or someone is, there's like a few pockets of people and they've all three or four redeveloped, rebuilt, basically the same way of working or have created a complete unique castle, right? About like how they do everything. And because of that, it's like, you know, wasted work. It's duplicity. There's not interoperability. But then it wasn't designed in a team uplifting way in a way that's going to make it is like a durable process with the carry forward. So it's also too about like identifying when those pockets happen of like, OK, what can we learn to make the whole team, the whole process more agentic?

15:38And then bit by bit, it becomes a durable part of your SDLC, your your AIP DLC, however you want to envision how something comes from idea to to execution. And, you know, this has been a really great view of how like your own engineering org has transformed during this really like rapid time of like innovation. And during that same time, and, you know, we talked about this at the beginning of our chat here, like this transformation is also coming for all of these other industries is also heavily impacting all of them right now. So from your perspective, when you're creating tools and a platform that's consumed by users who are not engineers, you know, let's talk about how you think about adoption, how you think about getting them in the right mindset to like be successful in a gentic world and build products that are useful for them.

16:28So like, I'm curious about like how you get trust and keep trust and how you think about trust, because I think trust is completely different when you build products for a non-tech world. Us techies are a little more forgiving and, you know, to each other and to our status pages and whatnot. But when, you know, a major like provider for like how a business stays open is impacted by something or dramatically changes, these pains are real. And so, you know, having trust is so important. I'm curious to know how you think about that. Yeah, that's a great point. I think just because, yeah, engineers definitely are more forgiving and you want to be and stay on top of technology and leverage the best technology.

17:12It could be my platform that goes down tomorrow. I can't blow it up. Exactly. It could be me. You know, we are a little traumatized by the nines, right? And so other businesses are not. They're not. Exactly. And then developers have this, it works on my system, right? It's always that phrase you see, oh, it works for me, right? But when you push your code into the Git and when you test it and deploy it with other code, you realize that it breaks in integration. When you test with complex use cases, it breaks. So that's where the trust is important. It's like you're measuring it based on the outcomes you're delivering.

17:45Is it the customer use case, right? Our customers, we serve a lot of industries, some tech, non-tech, right? And the users of our system are, again, different kinds of employees. Some technology for some in other departments who are using technology, but they're not the most technology savvy. So you are delivering functionality for them to use the product. And you trust if the system is helping them get their job done, they're going to trust you. If the system is not going to help with their job, they're going to start saying, OK, I'm not going to use this product. They'll find reasons to not use the product.

18:21So that's why it's important for us to gain that trust. And it has to be earned, right? It cannot be just granted and said, okay, I'm going to, we want to let you trust the system. They have to tell you. And it's up to the customer satisfaction or employee experience scores. I think that's where determines whether the system is working or not working. And that's all we have to, when you build software, you have to think about, is it, if you're incorporating AI into the workflow, it's not that, just putting AI is not going to help them trust the system. They have to look at, okay, is the outcome that they're trying to get out of leveraging AI, are they getting the same outcome on a better outcome?

19:04And that's how the trust is going to be. This episode is brought to you by Linear B, the engineering productivity platform. Your agents now ship more code in a week than your team has hours to review. And on November 18th, we go live with a walkthrough of Linear B's new agentic studio, built to deliver outcomes when human attention can't keep up. Co-founder Dan Lyons opens with why we rebuilt Linear B around context. Then the product's team shows it working. I'm hosting with Jamie Burse, and we're saving the last 10 minutes for your questions. So be sure to save your seat. And if you can't make it live, register anyways, and we'll send you the recording and the assets afterwards.

19:44Link is in the show notes. and bringing them to the table, keeping them at the table is a completely different operation when in terms of trust, you know, you are talking about a very different audience. And I think also to something that's different is how they, um, the stickiness and the adoption of tools, because really realistically, like if a business owner can, can get their business scaffolded and built up on top of a system that they can rely on, they didn't go into business of opening whatever store or doing what they're doing so that they could sit on top of their tech all day and make it perfect.

20:17Maybe a really nerdy one did, but most of them don't. They want to set it and forget it. So there's like a little bit of like a push in a pool where you want to put the most innovative, the most like exciting, the newest stuff right at their fingertips. But you also have to make sure that like what they're used to and what they're expecting is still delivered. And this requires, I think, stepping customers, especially in the last year, through a lot of little steps of adoption changes. Let's like slightly, you know, wiggle you this way, wiggle you that way until we're all kind of like, the customers are understanding what the tools are supposed to be using for and they have the intuition to adopt them and be sticky with them.

20:56And you're able to get that really good feedback now about the tools they are using and be able to make them better and more effective. And I'm curious, like in that journey, like what about adoption in the AI world and adding AI features to your platform has surprised you and, you know, in a, in, in any kind of direction, like, is there, uh, things or, or parts of it that really stood out as part of like that transformation? Definitely. Yes. Uh, uh, just because we built a feature of the product doesn't mean that a customer has been adopted. Right. Uh, and then we have to ask the reasons for why is the customer trying to use that feature because they are trying to deliver outcomes.

21:36And it's all about, so there is adoption, there is impact and outcomes, right? So adoption is something that we all measure to make sure that, okay, yes, are they using the product and making sure that they're able to discover the product? Are they able to, we look at usage, we look at some of those metrics, we look at token usage, all of that. But end of the day, customer doesn't care about any of those. the customer care about the outcomes that they want. Is it helping them do their work faster? Is it better quality? Is it reliable? Is it trustworthy? So let me give an example of one customer of ours, Seagate.

22:14Seagate is in the whole Describe business. And they became a customer a year or so back, and they migrated 30 ,000 employees from their existing system to our Freshworks, fresh service product in about three months. So they took a big bet on us to get it done. But the whole intent was, how is it going to be better in the system they had? And are they able to improve the satisfaction for their employees? And what we are seeing is once they started deploying our AI agents, you see that at least 30 % of our tickets are being deflected. their first contact resolution has improved. And that's the kind of operational metrics that we want our customers to start measuring to say, oh, I did change the tool.

23:03I put a lot of effort into changing my tool set and re-educating my employees to work with a new tool. But if it doesn't deliver impact, the cost of change is not going to be justifiable. So that's why I think we need, the adoption is important, but then bigger than that, the impact is the most important one that people have to look at when you're rolling out software. Yeah, you have to be able to illustrate it, make it really obvious, and also to just make it self-instructive. But also, what I love about the story with Seagate is that that's a huge shift. That's a large amount of seeds. And that is a really big bet on where the problems are and what can make it better.

23:49And ultimately, it's like having the right kinds of systems that can support those big bets and be able to be that like landing place where a company can make that change and then experience the transformation, that internal transformation that they're trying to get. It's like, I think teams are willing to make big bets right now, especially with agentically enabled tools to out-transform their competitors. And that becomes like the real engineering challenge is bringing those people to the table. And, you know, as I listen to you talk about it and what I really think is must be the truth is like with so much of your world that you build must rise out of user stories and getting super close to how all of these different industries and companies use your product, hearing their pain, live on calls, surfacing them up, turning them into real things that can be creative.

24:43there's a true potential in how the tools exist and how you get feedback from your so many different customers to truly create a unified platform that can meet their needs but in order for that to happen they have to be using the tools because you can't just build the tool you have to get everyone enabled everyone excited about it maybe this person is it has a behind a feature flag and don't even get me started on how feature flags make all i know i know right and so it's like then you start thinking about how you market that. And really what happens is your user base and how you talk about your product becomes a whole bunch of different segments of like, these are the really, really AI enabled, the super agentically enabled cohort of our customers.

25:25And you'd be able to like learn new things from them versus maybe how like newcomers are starting to adopt things. But even they have their valuable user stories you dig out to make your experience better. And I'm curious, like how you think about your role as a CTO and overseeing that kind of technology transformation. What things have changed for you in the last year in your perspective? And, you know, do you have any other advice for folks in your same kind of position? Definitely our jobs have become a lot more challenging. I think, again, it's no longer steady state anymore, right? So every day there is some other company that disrupting what you're trying to do and that the technology is getting better.

26:07the barrage entry is becoming lower. So naturally you have to continuously innovate, differentiate, keep up with the customers. That's become a bigger thing for us. So it's not that, okay, yeah, I'm just not focusing just on technology, right? So we are thinking about how do we bring everybody along? How do we bring our, change the ways of working? How do we bring our engineers along? How do we bring our customers along? So a lot of these things are, shifting to the cultural shift that needs to happen on how we build software, how we ship software, how fast we ship. There used to be a time we always say, I'm going to ship it in three months.

26:46And now everybody's like, okay, what can you push at the end of the week? Right. So the appetite for new things is a lot more. People are expecting change to happen much quickly. Nobody's waiting for a yearly cycle to push software. So a lot of the expectations have change so much, right? So it's all about how do we, it's not about deploying more AI or deploying the best technology there, but what technology will make the organization meaningfully better. I think that, or what solutions can you bring to the problems that will make the companies operate at a much higher efficiency is where my thought process is changing.

27:27And that's how all the engineers are thinking about it, because it's not about just building the software, because that seems to be the easiest part of it. It's about how do you optimize yourself and how do you make sure that you are improving the efficiency of your developers, changing the perception of how do you innovate. At the same time, not be reckless, but be more disciplined in how you build and ship and make sure that your employees, the quality and the, I would say, the scale and other performance metrics are not compromised. Yeah, absolutely. You aptly called out how the pressure to ship more often and constantly has compressed.

28:05And there's definitely more rapid cycles. And in this whole process, like we're dealing with the bottlenecks that make those fast cycles possible. And those bottlenecks are not solved. So then we learn about the real state of production in production. Observability becomes really important. And like the way that we think about how do we maintain this, make it durable, completely changes. I loved how you also called out that like the the timelines like shrink. So everyone's involvement gets like a lot more hands on. And engineers just have to have a lot more context about the problems that they're solving.

28:45Because before it was, you know, maybe they get the nice ticket. They have their sprint. They're going to work on this part. Eventually, we're going to add all of the pieces to this jigsaw puzzle and it's going to be solved. But now they have a jigsaw solving machine that they can just like fire on demand. But if they don't know what the picture needs to look like, they're not going to be able to do it. And so now the challenge is cultural, which you pointed out. And it's about getting people aligned around how do we work, what are our expectations, and how do we communicate clearly? Where is communication breaking down?

29:18those are often creating then those bottlenecks that show up as code that doesn't need to exist or something that should have been a refactor or 1 ,000 plus lines when it could have been 10. And so that's where those things start to pop up from. And so since AI makes code easy to produce, that's not the problem anymore. What becomes valuable now in an engineer? Yeah, exactly. So I think these are all great skills for engineers to develop because I think they need to think about, like we talked about systemic thinking. We talked about there's no longer tribal knowledge. A lot of the knowledge now sits in the context that's built when you build new code, right?

30:02Right. So the boundaries are kind of, there is not much things about, yes, there's not much differentiation between a senior person and a junior. It's all about how you think. and a lot of the stuff is not about, okay, only this person knows, experienced people know about something. And so a lot of the knowledge is being transferred into the context that's being built while you're building new code. And we need to take advantage of all of that while applying, taking that context to build newer features, newer products. Does Freshworks take like a forward deployed engineering approach? Does it bring engineers right into customer conversations so they can hear and get close to customer pain.

30:42I'm curious how you make sure that that context about like how to deliver that impact and those outcomes gets to the engineer since now that seems to be the true now cultural bottleneck. That's right. I think, so we have 75 ,000 customers and growing. So it's very hard for us to, and we have customers of all sizes, right? Customers who are original customers we don't even see. They just pay through a credit card. They use the product. And these high value customers, who have customer service support people engaging with them, services people engaging with them. They have a CSM, they have an account manager.

31:20So we have a variety of customers. So our ethos is all about intuitive to use, easy to deploy, all of that. So most of our stuff, there is no touch from engineers to meet for them. But there are complex customers, right? The higher the revenue they are generating for us, Naturally, there are complex systems that need to be demystified and we need to work with them to help them solve this problem. So we engage our engineers in those cases. And then that's the more helping them be successful. And then once in a while, there's always these escalations that come up, software bugs and stuff like that.

31:58And those are times where if a customer support or frontline support engineer is not able to solve an issue, we engage a developer to work with the customer. And we also create these opportunities for customers to come and meet us as part of our conferences that we do. And then that way we have a dialogue with the customer. That way, if customers have something that they can guide them through engineering. But sometimes they want to talk to engineering because we supposedly are the truth tellers. So it's very hard for engineers to say something that is not there to say it's there. So customers want to talk to engineers to get the right facts.

32:36So that's why we keep our engineers engaged. But it's not like every customer gets an engineer engagement. We can choose which customers we engage with. Yeah, it makes perfect sense. And just like how what we've talked about in our conversation, you have to think of this really wide scope, this really wide surface area. And in addition to all of the different industries and scopes and sizes of the businesses, there's also different levels of investment in each other that is obviously going to bring an engineer closer or further away from whatever's going on in their world. But no matter what, there's all these different lanes by which feedback trickles back to the engineering org, by which engineers get closer to customers that are having really complex or really interesting problems.

33:20And then this becomes the new place where the engineers can provide like the utmost value. This isn't something you can just pull out of an LLM or pivot the claw to ask. You need a domain expert with a system mind who understands how this stuff is built. A code owner who maybe runs the fleet of agents and run and reviews their code. I can get closer to the problem and understand how to translate that back through prompts and context into the engineering reality. And so that expression of customer pain into code is just like a new dance, I think, that engineers are learning to do. And as long as there's ways for them to get close enough to start doing it, I think they can build the muscle.

34:03So that's really exciting to hear about Freshworks. It sounds like a really great way to keep engineers kind of like close to how customers are using the product. And I just am so impressed, Murali, by the surface area by which you cover every day in your role. And it's very rare that I speak with somebody who operates such a high-level engineering context, but then also has to think about how non-technical, non-engineering users at scale and all across the world are going to think about and adopt and use the technology that your organization's driving. And that's a really big challenge. And so it's been really fascinating to dig into your head and think about some of the more emerging topics around your own engineering org, but then also how you guide your customers through AI adoption.

34:52It's been really, really informative. And I think it really speaks to the engineering practices there. And just for our listeners, where can they go to maybe keep up with you or learn more about what you or your engineering team is doing at Freshworks? Thank you for kind of giving me the opportunity to be here. You can definitely find us, find more about Freshworks on freshworks.com. And if you really want to know how our engineers build and how we think, we have a Freshworks engineering blogs on Medium. We share deep dives on our architecture, infra, AI, and operating at scale. We also have developer.freshworks.com, where you can find all the developer-facing tools that we share for our customers, There's APIs and developer toolkits and all that.

35:39And you can also find us on LinkedIn, Freshworks, or you can Google me on LinkedIn and then you get information about us. Amazing. We'll be sure to include those notes or those links in the show notes. And I'll include links to your socials as well because to our listeners, if you've made it this far, then you clearly loved our episode. You clearly loved what we dove into today. And so wherever you're listening to it, be sure to give us a like or a comment, but also please come find Murali and I on socials. We're both on LinkedIn. I'm going to be posting this episode there along with a newsletter on Linear B.

36:12And we would love to hear your thoughts about our discussion today, maybe your own perspectives. And if you have a bone to pick about something we brought up today, then that's the perfect place to do it. And so we'll see you online. And thanks again for listening. Murali, thanks again for coming on Dev Interrupted. It was a pleasure to host you. Thank you, Andrew. Thanks for the opportunity.

36:37FunVestback

From the publisher

AI can write the code, but a human still answers for it once it ships. Freshworks CTO Murali Swaminathan built his engineering org’s approach around that idea after opening the gates to AI tools brought a wave of duplicate code, often 100 lines where 10 would do. This week on Dev Interrupted, he joins Andrew to explain how an AI product development life cycle brought discipline back, and why engineers now own their work end to end like product builders. Murali also shares why trust in AI features has to be earned through outcomes, like one customer deflecting at least 30% of its tickets after rolling out Freshworks AI agents. 

Save your seat: Deliver outcomes when human attention can't keep up

Follow the show:

Follow the hosts:

Follow today's guest:

  • Freshworks: freshworks.com Murali’s company, which makes AI-powered service software for IT, customer service and CRM
  • Freshworks Engineering Blog: Medium Posts from Freshworks engineers on architecture, infrastructure, AI and operating at scale
  • Freshworks Developer Platform: developers.freshworks.com APIs and developer toolkits for building on Freshworks
  • Connect with Murali: LinkedIn

OFFERS

  • Start Free Trial: Get started with LinearB's AI productivity platform for free.
  • Book a Demo: Learn how you can ship faster, improve DevEx, and lead with confidence in the AI era.

LEARN ABOUT LINEARB

  • AI Code Reviews: Automate reviews to catch bugs, security risks, and performance issues before they hit production.
  • AI & Productivity Insights: Go beyond DORA with AI-powered recommendations and dashboards to measure and improve performance.
  • AI-Powered Workflow Automations: Use AI-generated PR descriptions, smart routing, and other automations to reduce developer toil.
  • MCP Server: Interact with your engineering data using natural language to build custom reports and get answers on the fly.

OFFERS

  • Start Free Trial: Get started with LinearB's AI productivity platform for free.
  • Book a Demo: Learn how you can ship faster, improve DevEx, and lead with confidence in the AI era.

LEARN ABOUT LINEARB

  • AI Code Reviews: Automate reviews to catch bugs, security risks, and performance issues before they hit production.
  • AI & Productivity Insights: Go beyond DORA with AI-powered recommendations and dashboards to measure and improve performance.
  • AI-Powered Workflow Automations: Use AI-generated PR descriptions, smart routing, and other automations to reduce developer toil.
  • MCP Server: Interact with your engineering data using natural language to build custom reports and get answers on the fly.

More from Dev Interrupted

All episodes