Host Intro (Gene Kim)

A couple of months ago, I had the pleasure of meeting Sree Balakrishnan. He heads up engineering and product at Travelopia, a global experiential travel company with over 2,500 employees operating under 14 brands.

I am so excited that he is presenting to you because, to put it mildly, he shared some of the most mind-blowing and radical stories of using AI that I have heard. He provided some concrete evidence of things that Steve Yegge and I speculated about in our vibe coding book. If his experiences and bets are correct, agentic coding could reshape technology organizations just like DevOps changed how infrastructure, operations, and QA shifted left. We might be looking at how development has to shift further left into the business.

What I so much appreciate about Sree is his methodical approach to experimenting with AI, but also his courage in experimenting with how we work. Like the joke, some of the things can only be discovered if you ask what happens if we turn it up to 11. He and Chris Condo, client partner at Equal Experts, will present a mind-expanding and inspirational experience report that I know should resonate with many tech leaders everywhere. Here is Sree and Chris.

Chris Condo

Thank you. Good. All righty. Thanks everybody for showing up. Let's just get right started here. Hey, Sree, we had some great conversations over this past week, and we are going to talk today about not only the journey of Travelopia, but also a personal journey you have been on.

Sree Balakrishnan

Absolutely.

Chris Condo

Why don't you give us a little overview of Travelopia first?

Sree Balakrishnan

All right. So excited to be here. You heard about booking.com. In our world of Travelopia, we see booking.com as a tech platform and Travelopia as a travel house. What that means is we own ships and planes and yachts and boats, and we give experience to the customer.

In my role managing the engineering and product, we build bespoke software for these brands that we have. My role is technology transformation for them, and of course using AI to transform them.

We have roughly about 15 experiential brands worldwide. Lots of legacy, lots of cloud, lots of data center, lots of modern stack. The biggest difference is, if you look at a platform like booking.com as a reference, we call it do it yourself. The customers can do it. The software that I build is do it for you. For example, a customer reaches out and our responsibility is to build and respond back in the shortest possible time to put together the best possible bespoke itinerary. We will get into the details, but that is the gist.

Chris Condo

So not just a technology platform, but human in the loop while you are traveling.

Sree Balakrishnan

Absolutely. That is key. In any GenAI discussions that we have, you always say there is a human in the loop. In our case, all the experience that we curate, there is a human in the loop.

Chris Condo

Excellent. All right, well, let's get started. Everything was great, right? Everything was going wonderful, but then GenAI arrived. What happens?

Sree Balakrishnan

Life was really good. I was enjoying with my kids, with my family. I was thinking I would retire quite well without learning something new. Then suddenly ChatGPT came. I was a bit of a skeptic and I said, yeah, ignore it. Big deal.

Then every LinkedIn post started saying, we have done this in GenAI, we have done that in GenAI, every conversation we have done this in GenAI. I was pretty excited when I saw the LinkedIn posts initially, but it turned out to be quite an anxious journey. I pretty much hit my rock bottom in my personal confidence, saying: am I going to be redundant? Am I needed? Can I even manage the developers that I have? I just do not know what is happening. Do I need to learn LLM skills now? I was really planning for my retirement very soon. I think all the plans went away. So, yes, it turned out to be very anxious.

Chris Condo

So a bunch of emotions all mixed together, right? Anxious.

Sree Balakrishnan

Absolutely. Eager, excited.

Chris Condo

But something happened on this little journey, right? A couple bumps on the road.

Sree Balakrishnan

We started, like everyone else, with a bit of experimenting in the early days. I am talking about two, two and a half years back. We said, let's try on the customer support side, let's try this, let's try that. We quickly realized whatever we did went live but failed miserably because the quality of support was bad. The LLMs were not that great. We were wondering what was going on.

Luckily I was invited to a McKinsey conference.

Chris Condo

So this is a trough of disillusionment? Is that what you are talking about?

Sree Balakrishnan

Absolutely. They presented this hype cycle. Of course, all of us have seen it many times. Me being quite naive about this hype cycle, I thought, okay, GenAI is going to go through the hype cycle and when it reaches the enlightenment stage, I will also pick it up. Then I was proved completely wrong. I was personally going through that hype cycle.

Chris Condo

So not just a technology journey, but a personal journey.

Sree Balakrishnan

Absolutely. I immediately thought, oh, I do not need my 200-person team. I can run it with a 20-person team, suddenly to realize none of my code can be gone to production, and suddenly to realize I do not think I understand how to use this. I was in a stuck phase in my life saying, what the hell is going on?

Chris Condo

So we need to get back on track. Somehow we have to forge our way through this journey. Tell us about the first experiments.

Sree Balakrishnan

Completely false start. The definition of vibe coding is different for everyone. The way I define vibe coding is you do not check the code and you just deploy to production.

Chris Condo

Let's not debate this. What could possibly go wrong?

Sree Balakrishnan

We did a vibe coding. I sent all my top engineers and product managers. I said, go vibe code. Let's take it to production. They came back after spending a lot of my credit card money and then they said, I cannot take this live. I am like, okay, hold on. What is going on? False start.

Then we realized, again thanks to the session that I attended with the McKinsey team, they said 95% of the projects fail. There is a bit of a consoling feeling: oh, okay, I am not the only one. But then what next? You need to be curious about what is next.

So we said, can we pick up the most important project in my company? If that fails, my CEO will call me.

Chris Condo

You did not pick the easiest projects?

Sree Balakrishnan

Not at all. That is the first mistake I did. I cherry-picked the most important one that can actually go live. We said, okay, if I can take the most complex, most important, take it live using AI-enabled development, I will consider that I have understood this. Until then, it is a hype for me.

Chris Condo

Then Equal Experts enters the scene.

Sree Balakrishnan

I think it is a tricky thing. If I look at my LinkedIn, if I look at my learnings, Claude Code keeps changing every day 10 times. Cursor keeps changing. Just a reminder, we are not a tech company. We are a tech-enabled company. There is a big difference when you have a tech company, when you have smart engineers surrounding you, versus a tech-enabled company where you have limited R&D capability.

I was a bit stuck in terms of which tool should I go with. If I go to Amazon, they will say this is the best tool to use. If I go to Cursor, they will say this is the best tool to use. Selecting the tool itself was tough. Second, the tools kept changing. Every tool started shifting left. The design tool suddenly started saying, I can do front end. Cursor started saying, I can do production code. Where do we believe? Where do we start?

I needed a partner, not just a partner. I needed a cohort of customers who I could learn from. Equal Experts reached out to me and said, if you are setting up a cohort of nine customers, would you want to be the tenth customer and be part of the experiment? I said, why not? I am anyhow in a miserable stuck state. Help me out. I did not say that.

Chris Condo

So we were sort of your AI travel guide, helping you every step of the way, getting to your destination. This breakthrough led to a revelation, right?

Sree Balakrishnan

This is first of all personal to me and personal to my team. I need to go tell this to my team. What the hell am I doing? Adopt AI is not a phrase that I could tell my team. So I said, how about this? I do not think any of you are going to be replaced by AI, but if you do not adopt, someone who knows AI will replace you. That was a good starting point conversation, and mostly from a vulnerable space, not from "let's use AI," not from an authority, not from power, more from the heart.

Chris Condo

Let's talk about the fact that it is not just a coding change. It is not just engineers who are being disrupted. It is your whole entire organization, your whole change management system. Why don't you tell us about that?

Sree Balakrishnan

This is no secret. A lot of us do this intuitively. During this time I was talking to Gene and mentioning to him that we are a tech-enabled organization. I always get the chance to point a finger at my brands and say, hey, you have not adopted technology. I never knew three are pointing back at me, and it is my team who needs to now buckle up and adopt AI. It was a realization that it is no more telling them to change. It is telling me and my team to change. It was a complete shift.

As an engineer, I never thought I would have to learn change management skills. I am now reading Kotter's change management techniques and whatnot. From coding to Kotter's change, it is like a sure change.

Chris Condo

We have got a couple of key indicators here: champions drive adoption.

Sree Balakrishnan

When we did the SDLC, what we successfully did was redefining SDLC using AI. But then we quickly realized if there is no human in the loop reviewing the code, committing it to Git, this is not going to work. Purely because we are quite regulated. We are in multiple markets. We have insurance, payments. We cannot afford to take something live without taking care of data privacy and whatnot.

Chris Condo

That brings us to the human element is key.

Sree Balakrishnan

Absolutely. We defined champions who could understand the system better. We had human in the loop to review before we commit code. I think we naturally moved to people first. I will be honest here: I initially thought I do not need people in my team. Two years back, over a period of time, I said I need a certain type of people. It was a big shift in my own thinking about the need of the people in the system.

Chris Condo

That brings us to this other challenges and change management revelation.

Sree Balakrishnan

Exactly. Technology, when I started the experiment, in about eight weeks we were fairly clear what LLMs were capable of and how to manage the tools. We were completely caught by surprise by the need to understand people. I was mentioning that it is one person at a time, one team at a time, one project at a time, one brand at a time. There is no secret sauce. There is no magic bullet. There are no books out there. There is no playbook. It has to start with the senior leader, one person at a time, and people at the center.

Chris Condo

We have learned a lot from this journey. Let's talk about some of the scaling and learnings that you have had. How about speed? How important is that?

Sree Balakrishnan

I saw it in Slack. I think the two to three days seems to be a normal now. That seems to be a standard. What we did differently in the experiment, we did an A/B test. I was a bit of a skeptic when I started AI. I was not sure. If I cannot take it to production, I need to ditch this.

So we did a parallel A/B test. We had the human traditional SDLC running the same problem statement. Then we gave the same problem statement to the second team, which is AI-enabled. The first one was a two-person team. The second one was again a two-person team, but AI-enabled. For about six weeks they had feature parity. Exactly at the same point, the teams were learning, readjusting, recalibrating. We did not measure time. We said, let's explore, open a field for them to play around.

But exactly the sixth week, there was one more Jira ticket that we had to give to both teams. I was part of it. We said, hey, what is the complexity here? How long do you think it will take? My first team, which is a manual team, said this is going to take roughly about two weeks. The second team, in a different room, said, give us two hours, we will be done.

That is quite remarkable. We did not anticipate this. Sometimes when you approach it with wonder, not with an assumption that there is productivity, I think you are a lot more joyful. I was like, are you sure? They said, okay, lock yourself in for two hours and come back. They said, we are done. That was a moment. It was internalized for me. What I am reading is real. This is happening and this code is going to production and it is done. It was a crazy mindset change personally for me, which I had never experienced until then.

Chris Condo

This is where Equal Experts helps again, providing a repeatable framework with human-in-the-loop discipline and engineering rigor.

Sree Balakrishnan

I would add one point. As a technology leader, I stopped thinking about metrics and data and efficiency gain. I started wondering more what is the art of the possible, and then started challenging the teams rather than worrying about is it really a 20%, 30%, 40% gain. We do have metrics. We do have DORA metrics. The DORA metrics told us that our average ideation to execution was six weeks. If we can reduce that, that is amazing. We got it reduced, and thanks to Equal Experts' partnership, we were able to get that model in place.

Chris Condo

Here we are from squads to super lean teams.

Sree Balakrishnan

This is crazy. Gene and I spoke about quite a bit on this. There is a shift, and shift left happening. The engineers, I believe, are going to shift left toward more product engineers. Just a standard intellectual engineer who knows coding will struggle with AI. I am seeing that a lot of my craftspeople who are pixel perfect, who want absolute great code to be written, are struggling with AI. Those who are in the medium and above average are able to really embrace it, but they are shifting left toward the product org. Engineering and product are merging.

The product people, interestingly, are shifting right. They need to learn a bit more technical skills. My designer is owning the front-end stack now, and my designer is saying, I do not need front-end developers. I will design it. I will build the front end.

Chris Condo

So what do you see the future of this going?

Sree Balakrishnan

I experimented with smaller. It is not a two-pizza team anymore. It is going to be maybe a single-pizza team.

Chris Condo

Less on our pocket. They will just share one pizza.

Sree Balakrishnan

Exactly. Four to five members, smaller, tighter team is what we are seeing in the future.

Chris Condo

Smaller teams, tighter loops, faster outcomes.

Sree Balakrishnan

Absolutely.

Chris Condo

There are a lot of leaders out there who are in the middle of adopting AI. They might actually be in their own trough of disillusionment. Maybe you have some reflections for these leaders.

Sree Balakrishnan

If I bring it like in a movie style, to me the story revolves around senior leaders like myself, yourself, speaking for all of you who are 10, 20, 30 years, 40 years, a couple of decades in the industry. The villain is the fast pace of adoption of these technologies. AI is changing every day. There are releases by all these tools pretty much every other day. We cannot catch up.

The struggle, at least for me, is that we have family, children, and a lot of other things to take care of. The calendar is full. We still need to find time to learn. The action that I took was, okay, let's go step by step. Let's go systemically figuring out how do I do this over six to seven projects. Do not define early metrics. Do not define a lot of constraints during that time. Allow yourself to immerse in that journey. Hopefully at the end of it, like me, you will have self-reflection on what makes sense to you, your team, your product, your brands. I think that is super important. There is no magic bullet in my opinion.

Chris Condo

You told me something really remarkable. I have never heard this in my entire life as a software developer.

Sree Balakrishnan

Evil grin when I say this. Every time business comes and tells me, you are late, when are you giving me this feature?

Chris Condo

For the first time, you are not in the scrum meeting making up excuses. You are not doing planning poker and throwing out cards at people.

Sree Balakrishnan

Exactly. We had a moment in time where the traditional SDLC had predicted this development will happen in six weeks. My AI-enabled team got it delivered in three days. I was standing there and saying, can you take this live tomorrow? And the business said, no. Such an evil feeling.

Chris Condo

How about aspirations for the future? Where do you want to take Travelopia?

Sree Balakrishnan

Like I said, we are roughly about 200 people in the engineering org, and we are anticipating smaller product engineers running it. Of course, these product engineers are not in the market. I would imagine the team that we have will go through a metamorphosis of training internally, with the support of external and internal help. Through live projects, there is nothing better than doing things together.

Chris Condo

We talked a lot about this two-speed architecture that you think is a winning model.

Sree Balakrishnan

This is something that I have learned. McKinsey has something called the twin-speed architecture. They came up during the mainframe time saying the backend mainframe becomes a slow engine, and there is a front engine which you can build and wrap up and do more things faster.

I believe my current agile team is going to become the backend team, which takes care of insurance, payment, legal, compliance. That is what I call the stable engine or the stable team. The AI-powered team is going to be the one which is going to rapid iterate.

Chris Condo

The one that is going to meet the customer where they are. More volatile, more changing, faster pace.

Sree Balakrishnan

Absolutely. Keep up with the competition.

Chris Condo

You told me this amazing aspiration.

Sree Balakrishnan

We always had aspired to grow big, and this is one of our brands. Like I said, we have 14 brands. Each brand has different aspirations. One of the brands, with the consistent work that we have done in the last couple quarters, was stagnating at $100 million. Now with all the experiments that we have done, all of a sudden the leadership turned around and said, can we grow to $500 million without adding headcount? That was a defining moment where we saw it emerging internally. This is not a boardroom setting saying, go get $100 million revenue. It is an internal realization that we could do this together, knowing what we know with AI.

Chris Condo

That is quite a goal. How about let's share some personal perspectives from your own journey.

Sree Balakrishnan

I am in midlife. This is a once-in-a-lifetime opportunity. Like I said, I thought I would retire without learning any of this, but I am now super energized. Completely once in a lifetime. Let's see where this goes.

Chris Condo

At 47, you have seen this as a once-in-a-lifetime chance.

Sree Balakrishnan

Totally. I totally believe, I think all of us here, if you are like me figuring out your midlife career challenge and figuring out should I jump in or not jump in, just jump in. But do not start with metrics. Do not define success criteria. Just internalize this for a while, and hopefully in about a quarter you will know what exactly is to be done.

Chris Condo

Let's do a quick run-through of your technology aha moments.

Sree Balakrishnan

I think the SDLC is redefined already. No more six, eight weeks, two months delivery cycles. It is days, weeks. Same modules done in hours, days. It is no more incremental for me. It is exponential.

Chris Condo

How about the people realization?

Sree Balakrishnan

If any one of you think that I am going to just give a bunch of tools and expect success, forget it. It is not happening. It is a process change from the business towards production.

Chris Condo

We have some excellent testimonials, but let's move on. Let's talk about the kind of help you are looking for.

Sree Balakrishnan

The tools are changing. There is absolutely no way anyone can catch up on the tools. We do need peers to learn together. We do need to share notes. There is absolutely no way that we can be alone in this journey.

Chris Condo

Equal Experts is a consultancy and we want you to bring us your biggest challenges. We will be right over at the expo hall. Tell us what is going on. We want to learn from you so we can understand how to bring AI technology and adoption to your company. Thank you very much.