Puneet Tangri

I work as vice president of product engineering for ESO Solutions. And at ESO, one of my core mandates is transforming the organization from the pre-AI world to now taking it into full AI in terms of how we utilize that for productivity. Prior to ESO, I worked at Arlo Technologies. I was vice president of product management there, and I utilized AI and created AI-powered services for home security.

There's two parts to this. One is delivering value to the customers. And the second one is extracting value out of AI to think of how you're going to create value for the business. As we were just talking about, we do both at ESO Solutions. We look at what we can do from a product standpoint. But at the same time, we're in the healthcare space, and in the healthcare space there are a lot of repeatable processes. There are compliance-based aspects. So how can you really create value from a private equity standpoint for the organization, where you can move much faster by doing things at a faster pace and creating time and space for strategic initiatives?

That's right.

It's been amazing. I will say about a year ago, we had almost no utilization of AI. A lot of the team members, myself included, were somewhat apprehensive as well as unsure about where it's going to take us. And today, from an engineering standpoint, 95% of our pull requests touch AI in some way or form. But we took a very methodical approach to how we were going to bring in AI, instead of just saying, "Let's go AI." We thought about it almost like a growth pyramid. Because I come from a product background, I was able to talk to my tech teams and say, "Let's think of it as a growth pyramid. Let's not go all in on day one. Let's think about foundationally what the right things to do are." So we thought about tools that have to be set up for each engineer. Are our repos in the right place? Are the tool set and the teams ready to take AI and embrace it? So we did internal trainings, we met with some of our partners externally, did trainings for the team members, and started to do POCs slowly. The proofs of concept were around teams. We took a few teams that were very interested in taking AI and going forward with it, versus some of the teams who were more apprehensive. They did repeatable work which was initially taking four months. If you step back and think about healthcare, what do we do? Of course, we innovate, we bring new feature sets, we bring the things the customers really want to solve their problems. At ESO Solutions, our product is software and data connections between EMS, hospitals and fire departments. So there's a lot of flow of data, and the software develops that. But at the same time, we have to stay compliant every year, which means there are a lot of regulatory and compliance aspects that the state and federal governments set up every year that the software teams have to do. That repeatable work has taken four to five months every year for the last 20 years. So we took one of these and we said, the team has a challenge, it's a repeatable process, you're touching 18 different repos. Why can't we do this in an agentic flow? So we created agentic flows that ultimately touched all of these, with human judgment and approvals required during the flow, to take it from four months to below four weeks. And at this point, the remaining teams are excited and they want to see what they can do and how they can replicate it. So essentially it's a benefit the organization sees in the productivity of each engineer, where AI is a coworker, and then thinking of how repeatable workflows can be automated through this.

Less than that.

At our organization, we have a lot of strategic programs and projects that the teams want to do. The product teams are hungry to get more capacity from engineering. So there is an opportunity for us to start deploying more effort in the organization towards strategic initiatives that allow for growth of the business. Today I see a slow acceleration of engineering, where product still has a roadmap ahead. I think in the future, there's a possibility where product has to become faster, because engineering can produce the output really fast.

Yeah, and you actually alluded to exactly where things are shifting. We used to have big teams delivering big roadmaps. Now we have small pods: four to five engineers, half a QA, half a product person, and they're being more accountable for their own outcome. So what we're focusing on now is going from output to outcomes: create small pods, think about how these pods become outcome-oriented, allow them to be more autonomous, and at the same time, move fast. So that's where things are starting to move and shift a little bit. And each team has its own velocity and its own challenges. It's not easily replicable from one to the other, where the problems are the exact same, but the general process of workflows, the way we are starting to think, is very similar across teams.

It's both. I think both things are still equally important. And because we're starting to work in smaller pods, it's more important for them to become co-located where possible, so that they can talk more, because judgment is key at this point, and you can't wait overnight for it. There were points in time in our operations where teams were spread across four different time zones around the globe, and now you're waiting. So the whole point of trying to make the workflow go faster actually starts to slow down, because of the artificial time barriers that are coming in. So it's important for them to become more co-located. They understand each other, they're able to work through things with each other better, and then at least one or more of them understand the domain pretty well as well. It's not just execution. It's now understanding, and almost crossing the lines of: do I understand the product really well, even if I'm on the engineering team?

Over the last five to six years since COVID hit, we've all learned how to work remotely, and that's a skill set that we don't want to give up on. So I don't necessarily think it's needed to be in the same room all the time. Of course, being in person helps. But what I meant is that a lot of organizations have their existing teams intermingled. You have two people working on project A who are in North America, two people on the same project who might be in India, another two people in Manila, and someone else sitting in Dublin. Now, this is not the most effective way of working, because now you're working across time zones and you're waiting on people to be available. So the way I was trying to articulate co-location is that the same organization has teams that can be in the same location, in the same time zone. There is no need to change people. It's just the way you operate and how you create responsibilities in each area. That simplifies things, so it doesn't necessarily need any shift in the organization, but a shift in how you functionally operate.

Exactly.

Absolutely. I have moved to a place where I feel like this is going to be truly game changing for the industry, and for the right reasons. A few months ago, we wanted to double down on our efforts on customer success and make sure that we have an opportunity to go faster and reduce our MTTR, so mean time to respond to anything that is happening in production. And one of my engineering leaders and I sat down and started to discuss why we even need to create these four meetings, where you start from Salesforce, which has these tickets, and now you have to explain that to the engineering team, open it in Jira, then you go through prioritization processes again, and it takes two to four weeks before you've actually even acted on it or opened the code to look at it. And we decided to come up with a process where our focus was, again, outcome-based: how do we reduce mean time to resolution, really to a place where it improves the NPS scores for the organization? The way we did that is we started a batch process that runs every night. Every night it picks up a certain area that we are very comfortable with, because the domain expertise is there with the system now. The batch process runs, it picks up any production issues that are open in Jira, and it takes them through the engineering workflows. The next morning when the developer comes in, the code is actually ready for review, and the code can be pushed to production in 24 hours. So our objective is to get there, and I feel very confident that multiple product applications that we have, in about a year, will be in a situation where a lot of these flows will be automated, and it provides a higher value for the customer. Customers are happy, the business is happy. So I think that was the moment where I recognized that this is a big change. And the one thing that is going to impact all of that is how we can do things not just faster, but at higher confidence, with the right human judgment and the right workflows.

North of 125 in my team. The organization probably has slightly over 150.

If you think about any SaaS business, engineering has always been the bottleneck: delivering, and with high quality. That has always been the bottleneck. I think that bottleneck is starting to shift left a little bit. If you think about it, you've got sales and customer success, then you've got the product management teams, and then you've got the engineering teams. That bottleneck is shifting, which means there's a lot more opportunity for any business to grow, as long as they invest in the right tools and the right people on the left side of the business that can actually start to drive growth, and technology teams are going to be able to stay on pace and deliver on the needs. So as I think about the organization internally, externally, and more broadly in the industry, I believe that the opportunity is for all businesses to grow, if they can focus on where growth is going to be, and the teams are going to be ready to deliver on it.

So am I seeing an increase in headcount immediately? In certain areas, probably not. What I have seen is that unless you grow the business and you have the right growth ideas, whether or not we were able to automate workflows with agentic flows and such, organizations were cutting employees and layoffs were happening. So layoffs are not a new thing. They've been there since I started at my first company, my first year, two decades ago. There were layoffs and they continued, and the reason for that is because business wasn't growing and there were not enough ideas. So I think in any business that grows, where there is an opportunity, I do believe that the engineering workforce will continue to grow along with it in the right places. They may not be doing the same things. They may not be doing the same repeatable workflows. The skill sets are going to be increasingly more complex, but at the same time, the growth will be there.

It's different for different companies and different stages, but when it comes to engineering organizations, there's hiring definitely going on for folks who are able to transform the organization to AI-first, from an operational standpoint, from a leadership standpoint, as well as from an engineering skill set standpoint, in terms of staff engineers who are skilled at taking the existing brownfield code, because that's different than creating a greenfield app. You can go into AI today and say, I want to start something new, and it can create an application in 24 hours. But think about brownfield applications, where there's a lot of context and a lot of domain expertise required. You still need the right people to help transform. So there's a lot of value for anyone who can transform existing brownfield code, as well as the team and the workflows, into the new place.

I feel value was always something that CFOs asked for, but you rightly said it's a lot more, you can use the word scrutinized, or looked at closely, when it comes to spending in the private equity world. So, back to our question of value creation and value extraction. Value creation is you create P&Ls, you show the growth, the double-digit growth or more that you're able to bring into the business. In terms of how it's looked at by CFOs in the engineering organizations, they're looking at it from a value creation standpoint on the number of hours reduced, and then how do you translate those hours into faster roadmap delivery for the strategic value and the growth initiatives that are getting delivered. If I were to give you an example, let's say a project is going to be delivered, in your roadmap planning, 18 months from now, which means the revenue associated with that doesn't come in 2027; that comes in 2028. Now, if you're spending more tokens, if you're spending more time, and you have initiatives that are increasingly bringing AI into the organization, how does that reduce the time? We're looking at time, hours spent, man-days that are saved, and quantifying that from a value perspective. And at the same time, compound that value with how we bring that 2028 roadmap item into, call it, 2027, early or second half of next year, where you can now put your bookings target and your growth target on it. So essentially looking at both sides of the coin, growth as well as productivity, and combining them, but making sure the ROI on the investments is there.

In some places, yes, and in some places it's still questionable. It's not a straight answer, because in a lot of organizations, and in many places within every organization, you're still in experimentation mode. And in some cases, there is a belief that if something is working in a different organization, it must work in our organization, so let's just do it. If you look at initiatives, organizations have AI champions and AI champion initiatives going on right now, and if they have a list of 20 enterprise AI projects, I would say probably half of them are creating true value and the other half are experimental in nature. But in terms of the investment being spent by organizations, probably 80% of that is actually creating true value, and the other 20% is where they're trying to figure out if it's value. And we've actually had, and I'm trying to remember some examples, certain places where we said it's not working, and we're not going to use certain tools, and we stopped doing that as well.

I am pretty optimistic about where technology is evolving and the value it is bringing in. I am an engineer by education, I did a computer science master's, and I know engineers are builders and they are lifelong learners. So I'm not concerned about engineers evolving wherever they go, within their organizations and outside. And I think the value that AI is going to continue to bring in all areas is going to be great. My biggest concern actually is when I think about relationships and people, and that's not just at work, but beyond work. When my kids were really young and they sneezed, I picked up the phone and called my best friend or called my mom and asked what to do. And now I go to ChatGPT. So that phone call is getting reduced, and our dependency on each other as friends and partners and family is starting to reduce. So my biggest apprehension when I think about AI is: what is it going to do to human relations 20 years from now?

I do feel like there's value it can provide in making more meaningful conversations. I sometimes think of it as my brain has a reflection of my brain in there. And because we were talking about my kids, I'll give you an example. My older one just went into high school, and he's trying to figure out what he wants to be. I've been trying to have this conversation with him for six months, and it has not been going well. Probably not surprising for most parents. And I started discussing with ChatGPT how to actually lead this conversation and what I'm doing wrong versus right. It's been two weeks, and it's been amazing. My son and I have had great conversations. We've taken it one step at a time, and I haven't pushed anything, but somehow it has opened his mind up, without going into the details, and the conversation became meaningful and helpful. But at the same time, I don't think I've yet found it to create more conversations with people that I would have necessarily reached out to. So I think it depends on how you use it and how you maximize the value of it to create and enhance the relationships that you have.

I'm an optimist in life in general, so I always look at the glass half full. And so I'm not concerned. Tough times in people's lives make them resilient. My kids play games and they lose every day, and I think that's the best part of playing the game, rather than winning every day. So I feel humans are made to get over the situations that they have to deal with, and when there is a motivation to go do something, they will. So I'm not concerned in that sense. I think the usage, the governance, how these tools are being used, those have to be monitored. So I definitely want to make sure that the opportunities to misuse any of these, whether it's in the real world or in an organization, are minimal. But at the same time, in general, I feel like the opportunity it creates for folks to learn, have clear thinking and come up with new growth opportunities for themselves. I'm not really concerned about that. There will be a transition phase, which is usually tough. When AI had started off, I gave some of the folks the example of a movie, and I forget the name of the movie every time, but it's at NASA, where there's a room full of people, and they're human computers, or human calculators actually. And all of them are doing math on the rockets, but they're doing it manually. And then the calculator comes out. Unfortunately, they all probably lost their jobs, but then they figured out what's next. So I feel like humankind is made strong enough and resilient enough that when motivated, they'll find their way. Thank you very much. Thank you so much for having me.