LaunchPod - Zac Hays
===

[00:00:00] 

Zac: Every bug is triaged by an AI tool that we built first, and it wants clarifying questions when it needs it. It has context to everything we've ever done at the company, and it's had such a meaningful impact on resolution rate, like more than 80% savings and cycle time, we call it from when the bug was filed to, and it's resolved, but in product, like just not having to deal with those distractions.

To know that customers being heard is gonna be addressed, like an engineer will have everything they need and we can focus on the more meaningful things that we do. Welcome to Launch Pod, the show from Log Rocket, where we sit down with top product and digital leaders. Today we're talking with Zach Hayes, CPO at Luxury Presence, a fast growing real estate tech platform that made a huge bet on AI adoption across the company.

Zach shares their 30 x value principle and how they rethought goals and projects to have exponential level impact. The AI design sprint process that allowed the company to 20 x product velocity and how their mandatory AI bug triage policy has cut resolution time by 80%. So here's our episode with Zach [00:01:00] Hay.

~ Every bug is triaged by an AI tool that we've built first, and it will ask clarifying questions when it needs it.~

~It has context to everything we've ever done at the company. And it's had , such a meaningful impact on resolution rate, like more than 80% savings and cycle time, we call it from when the bug was filed to when it's resolved. But in product, like just not having to deal with those distractions to know that, customer is being heard, it's gonna be addressed.~

~Like an engineer will have everything they need And we can focus on, the more meaningful things that we do.~

Jeff: ~Welcome to Launch Pod, the show from Log Rocket, where we sit down with top product and digital leaders. Today we're talking with Zach Hayes, CPO at Luxury Presence, a fast-growing real estate tech platform that made a huge bet on AI adoption across the company.~

~Zach shares their 30 x value principle and how they rethought goals and projects to have exponential level impact. The AI design sprint process that allowed the company to 20 x product velocity and how their mandatory AI bug triage policy has cut resolution time by 80%. So here's our episode with Zach Hayes.~

Alright, Zach, welcome to the show, man. Good to have you on.

Zac: Thanks for having me. I'm excited.

Jeff: I'm really excited about this one. You got the credentials from background, big company, you know, Microsoft, Autodesk you're at a company called Luxury Presence now, which, basically builds software for real estate agents and maybe not one that people would think is, you know, leading the way in innovation on ai.

But. You guys are doing some stuff, I think in 70 plus episodes here, you're ahead of, almost every single company we've had on and really interested to dig in and kind of go through how you guys are driving just such growth and innovation using AI to your advantage. Let's kind of start out just with, we talked earlier and you brought up this concept of, it's not incremental gain.

It's not. We wanna be a couple times better. the goal is 30 x improvement which is absurd even in the age of ai. Right. So how did that come about? And how are you looking at that? Kind of, let's start from the beginning of like, how are you, how are you looking to get there? 

Zac: So this is a mantra that I have internally when we're looking at ways to improve how we deliver product. [00:02:00] there's so many ways of AI to make yourself. A little bit more efficient. You can use chat JPT to help you write a PRD and you're gonna get incremental gains. And we realize like you really have to think a lot bigger to get the huge unlock.

So we ask ourselves yeah, you can make yourself a little bit more efficient by writing a PRD faster, but what would it look like if you added 30 x more value with some workflow change and it just changes the entire mindset. So what would bug fixing look like if you could do it 30 x better?

What would a product look like if it was 30 x better than the old workflows? And it just opens up people's minds too. Crazy things.

Jeff: I like the thought there because one of the things I've gotten to do is go around the country and we have this series of dinners for, product leaders, CPOs, VPs, and we have everything from fpos up to, high level people at really significant, you know, large Fortune 500 organizations.

And , that's a question that I hear a lot is. How are we getting these big gains? Because at some level. The amount of work and value the society's put into AI is not justified if we get, a 1.5 x increase. But at those dinners, you [00:03:00] know, we have a guy named Aji Dewe, who's been, all over.

He was at, he was an executive at Twitter, at Calendly, at Typeform, at, Microsoft. And his point is. You can't think about it from a 1.52 X way. He would love your take on this because at heart, for every company that's gonna be meaningful, likely, right now, there's some kids at YC who are building that company from scratch, and they are not constrained by anything we've done before.

And they're going to build it in the world of AI first, and they're targeting 30 x or more. So if we're not looking at, how do we. Do that ourselves. There's high risk there that you get innovated on. And I love that, you know, that was just the native way that you dug in.

So, let's just go right into it how did you start there? And what does that look like for a product team, at a company of your guys' size?

Zac: we actually got tinkering with AI pretty early on because we saw an actual opportunity to make. Ourselves a little bit more efficient in delivering services for our customers. What our company does is we started with making gorgeous websites for real estate agents and there's a lot of manual processes that go into that.

And then we [00:04:00] layered on marketing on their behalf pulling us into CRM. And eventually we were gonna be, you know, the full stack SaaS. Vertical SaaS play. But some of the things that we used to do was we had a team that would write blogs on behalf of our clients so they could add it to their website and put it in their marketing.

We do social media on behalf of our clients. And when Chad GBT came out, it was like, oh, clearly this thing's gonna help us write blogs better, faster. And so we actually started tinkering with that workflow. We used a tool called Air Ops, which is. Is another tool that we use just these like kind of low code, no code tools to stitch together workflows.

And turns out that our team could write really great blogs at extreme scale and it would actually start out performing some of the content that we used to produce manually. So we decided we could kind of offer that to almost all of our clients because it was so efficient. But you start to hit these micro bottlenecks as you go.

They're really fun to unlock. Like one of 'em was, oh, well, we're cranking all these blogs, like we don't have enough Shutterstock images to make every blog look unique anymore. It's well, could AI generate the images for all these blogs? Yes. I mean, it took some [00:05:00] tinkering and , the technology's improved a lot, so we just kind of chipped away at these little problems, starting with a customer problem and learned all these tips and tricks for ourselves.

So now most of my PMs are. Capable to do these workflows themselves using innate end tools. They're all excited to build workflows internally for themselves, for other teams. Our engineers are fully in on cloud code and they just, they don't write code anymore. They just review prs, tell the code what to do.

So we, we started the journey a little early, but it's been amazing some of the unlocks that we've gotten along the way.

Jeff: When you say writing blogs, is this 'cause for real estate agents? Is this kind of the posts about about the properties and things like that,

Zac: Yeah, it is mainly like SEO content and we could do a whole nother podcast about what's changing in SEO world

right 

Jeff: Oh yeah. God, we know about

that. 

Zac: things like seven ways to prepare Your San Francisco home to be sold, like that type of content. Just to fill out their blog and get some SEO juice. it turns out that AI is exceptional at writing this type of stuff.

And we originally were creating it just for chat DBT, and it would hallucinate a [00:06:00] bit, but also it would just talk about restaurants that were no longer open and like these little details that you wouldn't think about until you're in it. And I think a lot of us are learning as you go, building, these types of products and workflows internally.

You could never write a PRD upfront that says handle all these edge cases. 'cause you just can't think of them until you're in it. Like you just have to try this stuff and start from somewhere. So obviously we think big, that 30 x thing, but my advice, everyone else is think big and find a way to start small.

There is something you can do that is small, that will unlock the next problem you wanna solve. so that's our mindset.

Jeff: right not to dig into like blog writing 'cause that's not super applicable to everyone. But as a process kind of thought experiment, right? When GPT three came out, you know, if you tried to just have a auto write blogs, it was probably gonna be. Not that great.

Like it's pretty easy to tell human from, not human at that point. But like you said, it also hallucinates, it pulls in, fake restaurants or restaurants that are closed and all that. But as these malls have progressed over time, you start to solve, okay, now it can actually search real time and not pull in closed restaurants.

It doesn't need to hallucinate 'cause it has access to the [00:07:00] world and can understand what the restaurants actually are. You can build the data set it needs for that. 

Zac: You can solve those problems yourself too. So before it knew how to research what restaurants were open. We just hooked up to Yelp's API. It's

like I think you gotta get in there and find these problems. Yes, the tech is probably gonna catch up, but if you wait, then you're like, that story said some startup is gonna displace you 'cause they're not waiting for it to be solved by open ai.

Jeff: And the thing is you don't start from, let's just put a little AI chat bot in here, or Hey, let's. Write some blog posts from using AI and move on. It's, imagining a world where we have people writing blog posts. Now what if we just never needed to, and we could do it for every customer.

We didn't need people to do it. if you start to look at those questions, it doesn't matter if it's, you know, in your guys' case it was, blog production in our case, we do session replay and analytics as the foundational kinda ingestion technology, but we ask the question.

What if you could watch every single session, and pull out the salient information and correlate it, have, you know, infinite context in your head and be able to tell what the important bits were that you needed to look at as, you know, like you a product person.

[00:08:00] it's a good model to look at and maybe you said something and then we kinda moved on, but the think big, but start small. What did that look like over at Luxury Presence?

Zac: the bog writing one, like that was a good example of you. We just started super small. But one that, still blows my mind daily is. We were thinking, okay, how could we use dev and AI and some of these new agents , to solve bug tickets faster? So we actually get a lot of issues that come up because we're dealing with MLS data, which is the multiple listing service.

There's 400 of them across the us. It's all user inputted data. So we get things like, why are my images not showing up on this one property? So it's not like a system wide bug, it's just this tiny little thing that requires a lot of investigation. was the URL formatted appropriately at the source?

Did we have something in transmission? Is it some permissions issue? We used to have a team that would have to go through and scrub through all that, and we said, okay, well, what if AI just did this automatically? And we said, okay, well, we can't just flip some switch and make that happen, but let's just go through what a human does.

Triage, they check if it's in the database. Oh, well that's like a [00:09:00] small step we could probably automate. And we just had an AI workflow that did that. Okay, well what's the next step that a persona do? And you just kind of go down this list and it's okay, well, that person's expert, they have all this context.

And it's okay, well how does the AI get the context? Well, it's actually in our code base. You know, you have, okay, well how do we cook? So you just start to chip away at these little things. And then now we have. Every bug is triaged by an AI tool that we've built first, and it will ask clarifying questions when it needs it.

It has context to everything we've ever done at the company. And it's had a, such a meaningful impact on resolution rate, like more than 80% savings and cycle time, we call it from when the bug was filed to when it's resolved. And that's gonna go to 99% at some point. But in product, like just not having to deal with those distractions to just know that, customer is being heard, it's gonna be addressed.

Like an engineer will have everything they need if they need to get involved. And we can focus on, the more meaningful things that we do.

Jeff: I bet the big payoff even beyond that is engineers don't have to context switch as much. They can focus on flow state more. They [00:10:00] can focus on the big hairy problems that are interesting in going to drive, growth of the company.

Not just solving some small problem for one customer. 

Zac: you hear these stories about the headlines of, oh, Zach told me like every bug is triaged by it. It's you look around for, okay, where's the magic wand to go from where I'm at now to this magical thing that he just described.

It's like it doesn't exist. That's not how anyone built this stuff. They all just iterated small, and can have a huge outcomes, but you have to find somewhere to start.

Jeff: that was not like you went from no automation to that overnight. It was incremental gains. They're gonna have big impact when you compound to compound and compound them.

Zac: Can easily, with this stuff, get 1% improvement every single day, easily. And so if you just do that math, like it doesn't take you that long to get 30 x improvements.

Jeff: So thinking about, this idea of incremental improvement one thing you kind of talked about when we caught up a little earlier, that. Caught my eye was this idea of AI design sprints does that look like ? Like in, in practice it's easy to say oh, we have AI design sprints, but what does that actually mean? Like, how is the [00:11:00] team operating and what are the tools enabled in the process enable them to, do this kind of, doing 20 times more bets than they used to be able to.

Zac: Yeah, so this actually this is when my eyes were first opened to new way of thinking, and we talk about taking the red pill, like the matrix reference. So I took the red pill about a year ago. First time I tried V zero, you know, the prototyping tools, they're all getting pretty similar. I know, it's just wow, like this is amazing that this tool not only took what was in my head, but also added things that I didn't even think about.

And at the same time, we had scheduled a design sprint. So this used to be big part of our playbook, and I had this thing on the schedule, this five dates, design sprint, and I actually had lunch with our CTO and I sat down and I was like, I use this new tool V zero. It is like like all these playbooks we used to have. We've been doing this together for a long time. Like just, we have to throw all this out the window.

Like I don't know how to do product anymore. It's all drastically changing. And I said, ' cause I was going through the design sprint schedule and I was like, well, day one is supposed to be high level problem defined. Oh yeah, we'll do that. day two is like sketching. I was like, do we need to [00:12:00] do that?

And day three is storyboarding. I was like, well, and day four is a prototype. I was like, I just made a prototype yesterday and 20 minutes. Could we just skip to prototype? And then it's the last day is testing with users like, well, how much could we do this? So we shortened it to three days.

We still did the first day of learning, and then the second day we all picked different tools, V zero bolt lovable, and said, you got an hour, come back with a prototype and show it off. , I mean these used to be things that like we'd swarm, the whole team would work on one prototype for the whole day.

and it would be, basically a Figma prototype. Like it wouldn't even be something that was very clickable. 

And 

so they were all a little shocked back then. And I think we've all normalized how easy building a prototype can be. So we had 12 prototypes. We show 'em off. Oh yeah, that's cool.

That's cool. I like that. Then we sort of called down to, we teamed up and got down to six prototypes and we were, our goal was to get down to one because we thought that's what the outcome should be, and we had four different ones that were all like really interesting and so we said, well. Can we just test all these?

It's well, we don't have enough users. So we threw them up on [00:13:00] unmoderated user testing tool. We use user brain. There's a lot of cool ones. And the next day we had three prototypes, four prototypes that each had actual user feedback and all these insights. speed was absolutely amazing.

And the surprising things is like. No one fell in love with any of their solutions because it was like barely theirs. You know, like some of this stuff you work on for so long, these designs, you really fall in love with them. it's so ephemeral now working in this speed.

So it's just oh, that didn't work. This didn't work. Oh, that crazy idea I had, it's not working. AI had this idea. Let's go with that instead. So yeah, it was really amazing.

Jeff: we had non you, he's the head of product at Linear on the show, couple weeks back. And he brought up a similar thing, which was. You have to look at it this way of innovating, , and he said 10 x. But I think that the magnitude is the same is the cost of trying things when it goes so low.

Now you can take the long shot bet that maybe had a 10% success chance, but had a pretty high payoff before you go. That's too risky. We're not gonna do that. But now you can validate, disqualify the bad ones almost [00:14:00] in zero time

Zac: Speaking of like taking more bets tech debt. Is ridiculously cheap , to start to tackle. So I think anyone that's been through like a re-platforming project, like it's, these just take years sometimes and it's like just really painful as a product person to not deliver value and like your old team sucked into this vortex.

But with AI coding and some of the things we can do now, like we actually have a project right now that we're closer wrapping up where we've. Essentially rewritten our entire code base. It's not that old, eight, nine years. Some decisions were popular seven years ago and then someone else came in, has a different pattern.

the reason why we started doing this, 'cause our AI agents were getting confused at what they should do. 'cause the code base was inconsistent. And so we said, well, what would it look like if we just rewrote almost everything? Yeah. Why you would never do that. You guys are a startup.

What would it look like if we did? And it's it's not that hard anymore to, you gonna have to test everything. You need an automated test, which you should have anyway, and then you can just go module by module and just sort of hit delete and rewrite. It's absolutely amazing.

The cost for those projects is just [00:15:00] dropping so fast from an actual, like human input and the payoff is unknown. Like how much payoff could you get if you had a super clean code base that AI agents could go through? And then they all need to be able to spin up their own versions of local environments.

And you have all these efim, you hit these really interesting bottlenecks, you start to scale this stuff up that. I think are fascinating and our team loves is to find them and tackle them. 

Jeff: With the ability to kind of like spin ideas up faster and try them. Is there anything, kind of unexpected, you've discovered something that worked maybe that you weren't sure how you kinda low comments maybe wouldn't have done before in a higher cost environment that has worked really well, has paid off, like there's some product that kinda you were able to ship that probably would've never seen the light of day or some insight you've gotten here.

Zac: We have these debates about it's really hard to not ground yourself in what's possible today. So when we were doing the blog thing, generating images that look gorgeous about real estate that don't need human oversight, like we had a lot of skeptics, so they're like no, we still are gonna need Shutterstock.

Potentially. We just need a tool that could be better at selecting these [00:16:00] images for the blogs and that we gotta have great curated images. And a few of us, including me, stuck to our guns and said, this is going to get solved. By the time we need to ship this technology will catch up.

It's moving at a speed that is just, no one has seen this level of innovation this quickly. So how do you get out of the like first impression of, oh yeah. You know, I tried it, it got six fingers on somebody. It's not ready. It's but it could be ready tomorrow. Like you have to be able to unlearn everything you just learned like in a day.

And how do you consistently the old Wayne Gretzky thing, how do you Constantly skating to where the puck's gonna be is so hard because it feels like it's light years out there, even though it's 

Jeff: Right. 

Zac: of weeks potentially.

Jeff: on that light, right, like one of the stories I heard that blew my mind was, and I had kinda a personal connection to this, so I always come back to it, is we had at one point an early employee here who was with us for a lot of years on the content team and then fantastic marketer and ended up going to this little company that, that I hadn't really heard of.

Objectively , they hadn't been doing great. Their [00:17:00] a RR growth was pretty low and stagnant, and I think they've been around for seven years and we're at something like 40,000. MRR or a r or 600,000 a RR or some like pretty low number for how long they've been around.

And they were kind of like burning money and not making it back. So they were in trouble. And we were kinda like why are you going to this company? I don't know if that's a great idea. And she had high conviction on it that, that something special was there that we weren't seeing.

Listeners might know this company better now, not as its original name, but as Bolt the AI code account tool. And the founder Eric was giving an interview and talked about, basically they had built all the pieces they needed for Bolt to be the magical thing that it is, aside from none of the code agents for AI were at the level that could execute yet, and then literally there was a.

I think it was Claude two five or something. And it was just a binary. Now it works. And they, you know, basically tested it and it worked and they, I think like one, one tweet about it and it went from zero a RR to 30 million a RR and like a month or something and 16, I don't know where it is now, but.

I think I heard something like 80 in under a year, [00:18:00] which is just utterly absurd. But that's the thing is you have to be, like you said, skate to where the puck is. They were skating to where that puck was for quite a while, and when they got there holy crap.

Zac: looking back, it seems obvious now to everyone that's, followed this, but that felt like impossible to most people to say no. Trust us In a year, AI is gonna be able to make an entire prototype that you could use and it's gonna be amazing. So you kind of have to have a leap of faith that it's gonna be ready.

'cause it probably 

Jeff: Right. I mean, I'm waiting for someone to codify the way we had Moore's law, right? About compute power and size. You could put on different ships and stuff where people were building. Devices and things like that for where they thought, you know, where it was projecting that tech was going to be.

And those were, , part of what made Apple great and several other companies was they were looking at not where are we now when we started to design this thing, but like, where's it gonna be in a year when we actually are putting the components in and let's get ready for that.

I'm curious to hear about what does this actually look like , in product, like this concept of, and as a marketer, I kind of just love this, but the autonomous ai, marketing team for real estate agents from a product [00:19:00] perspective, like how do you look at, solving the right problems and then applying the speed to that?

Zac: There's so many startups and big companies that are like, oh, we need to have AI in our product, and they slap some chat bot co-pilot somewhere and it's useless. So we took it from a purely first principle standpoint of If we were to coach a real estate agent on the best way to do their marketing, what would we tell them to do?

So let's say you had all the experts in the world. Well, you would have somebody optimizing your website for SEO. We used to do that manually all the time. We know how to do it. It's a lot of investigating. SEM rush, finding gaps in keywords that you can compete for. Like we know all the steps and then we just started chipping away at all the steps so that.

Now we have an autonomous agent that will optimize your website for SEO and for showing up in LMS and things like that because it's the same steps that a human would do. Same thing with, we have an AI agent that if a contact entry comes to your website and says, Hey, I'm interested in an open house, what would a human do?

They would text them back as soon as they could and say, great are you pre-qualified? [00:20:00] Lemme get some more information. what our texting bot does. so we just think exactly what, like what would an expert human do? And can AI start to chip away at that so that we can offer it at scale to all of our customers.

And many of our customers are very small business like solopreneurs and they can't afford to have a team of marketing experts, so we can give it to ' 

Jeff: I with some of the capabilities of AI and agents, a mildly interesting thing is can you be a little bit more productive? Right? Can you do something faster?

You know, 30%, 40%, hell even a hundred percent faster, ultimately is not. That interesting, right? Did that probably those gains probably would've happened with just workflow improvement. When you start to look at can you take the cost and time is a cost, but can you take the cost to, to basically zero, you can start to do jobs that you just couldn't do before become possible.

And that's, you know, taking that solopreneur who wasn't ever I've worked with a bunch of small businesses and stuff and, they're never, if you are just one person you're not only are you not going to do the website optimization 'cause you don't have time, you probably don't have the know how to do it.

Even if you did, you're not gonna do it 'cause you don't have time. You don't have time to do the follow up. You don't have time to do all those [00:21:00] things as fast as possible. And if you had three or four or five other people on your team there that's, I think exponential starts to even undersell sell it like logarithmic game is, can you add complete functions that they were just cost-wise, never have been able to have before.

Now they can have that full capability. So, that's interesting thing is how do you solve those problems?

Zac: Ai, air quotes is not differentiated anymore. Like I live in San Francisco, every single billboard in the whole city, so has AI on it 

somewhere. 

Jeff: I was there recently. 

It is wild. Do you wanna get something really funny actually? Billboards have gotten so popular in San Francisco, you can't even, if you want to buy one today, it's almost impossible until March. It's just completely wrapped up. Which is funny 'cause I mean, like we, we've both been around long enough.

I remember the time when, like they were discounted. Companies were having trouble giving them away at sometimes. 

Zac: It's mainly just a one upsmanship. If your company has a billboard in a specific spot, it's easier to recruit. Like I don't think they're trying to actually sell anything with those. It's all about recruiting and wooing investors and things like that.

Jeff: , It's interesting to see when you apply. The functional model of let's solve problems [00:22:00] exponentially faster, and be able to take more bets and then look at, what would the world look like if, customers could do all these things that were just impossible before, or if we removed all constraint from, time of work and stuff like that.

I think it's interesting to see the output is, you know as a marketer, what I think is a cool platform, but also how that impacts to revenue where. Growth is happening very quickly over there. And you start to see the effect of, the theory is great, but when the theory works is even better.

Zac: we've got lots of ideas for more agents that are coming, so we'll continue to add more value.

Jeff: yeah I, you know, we'll have to, we'll have to have you on again. I always used to say, you know, we'll have, we have to have you on in a year. But in AI time a year, might as well be like 10. 10 bajillion D years. So let's stay in touch and when you do some new cool stuff let's have you back on and go through how you guys are building other stuff.

If people wanna, you know, reach out to you and find about the AI design sprint process you've got or kind of how you guys are moving quickly and doing some of these things is LinkedIn the best place to reach out to your, or, 

Zac: Yep. Yeah, I'm pretty active on LinkedIn and I actually wrote a playbook for that AI design sprint that 

Jeff: Nice. 

Zac: directly from my LinkedIn page.

Jeff: Awesome. Cool. Well, we'll [00:23:00] we'll put a link to that on the show notes. And yeah, thanks for coming on and everyone listening if this was, useful and you learned from Zach and. You want to learn from more people who are doing real things with AI and building and managing and growing companies.

What we ask is subscribe. And that'll make sure you see the episodes, but the number one thing you can do, if I ask one thing tell a friend, tell a colleague, tell someone who's gonna get value out of this. You know, that's really how we get to do more and more of this and talk to more great people.

Zach, it was great to have you, man. Thank you so much for for spending the time with us this afternoon.

Zac: Yeah, this is great. I'll look forward to coming back sometime.

Jeff: Sounds good. All right, man. Have a good rest of your day.

Zac: You too.

Jeff: Thanks.