Chad Holdorf | How I Tested Pull Requests

May 13

Have a Listen

Summary

In this episode I’m joined by Chad Holdorf, longtime product and technology leader whose career spans John Deere, Salesforce, Pendo, and now Demandbase, where he leads AI initiatives across the company.

We explore how AI is fundamentally reshaping the way modern product teams test, ship, and learn, from debugging customer issues directly against live codebases to product managers and support teams submitting pull requests themselves. Chad shares how tools like Cursor and Claude are collapsing traditional handoffs between product, engineering, and support, creating a much faster feedback loop between customer problems, experimentation, and shipped solutions.

We also get into the messy reality behind enterprise AI adoption, including data quality, hallucinations, trust, evals, and why testing AI products inside real customer environments is much harder than most demos make it look. Chad gives us a peek into how his own workflow has changed, how his teams are learning by building in real time, and why this moment reminds him of the early days of Lean Startup, where he and I first met.

If you’ve been wondering what AI-native product development actually looks and feels like inside a real company, this episode is for you.

Takeaways

AI is collapsing traditional handoffs between product, engineering, and support teams. Chad described customer support teams going directly into code repositories with AI tools to investigate issues, understand root causes, and eventually submit merge requests themselves.
Most enterprise AI demos fall apart when connected to messy real-world customer data. Chad emphasized that “just putting Claude on top of the data” failed quickly without extensive labeling, validation, testing, and human feedback loops. Customers could detect hallucinations within a few prompts.
AI systems expose hidden data inconsistencies inside organizations. One example showed AI selecting a custom CRM field that technically produced better targeting results than the field support teams were trained to use, creating confusion about which “truth” was actually correct.
Trust has become the critical success factor for enterprise AI adoption. Chad explained that once customers encounter inaccurate outputs, confidence in the system drops immediately, which forces teams to spend enormous time improving prompts, SQL queries, evals, and validation workflows before broader rollout.
Product managers are increasingly becoming hands-on builders again. Instead of relying entirely on engineering handoffs, Chad now spends large portions of his week inside Cursor and AI coding agents investigating bugs, generating tickets, reviewing repos, and shaping product direction directly through code conversations.
AI-native workflows dramatically compress feedback loops. Problems that previously took days of back-and-forth between support, product, and engineering can now move from customer issue to deployed fix in under an hour through AI-assisted workflows and automated merge requests.
The biggest organizational bottleneck is shifting away from engineering speed toward enablement and adaptation. Chad compared this moment to early Agile adoption, where downstream teams like sales, support, and training struggled to keep pace with accelerated shipping cycles. AI is now amplifying that challenge even further.
Continuous learning and experimentation matter more than formal process mastery right now. Chad repeatedly compared the current AI moment to the early Agile movement: the people progressing fastest are the ones willing to try tools, build things, stay curious, and learn in public rather than waiting for established best practices or certifications.

Guest Links

LinkedIn: https://www.linkedin.com/in/chadholdorf/

Demandbase: https://www.demandbase.com/

Transcript

David J Bland (00:01.242)

Welcome to the podcast, Chad.

Chad Holdorf (00:03.022)

David, great to meet you. Happy Friday.

David J Bland (00:05.872)

Happy Friday. I am so happy you're here. We first met, oh my gosh, it was like 20 years ago or something crazy like that. And we were at a conference and you were at John Deere doing all kinds of crazy stuff. And you just, you just stuck with me. I was like, wow, this guy and the way he thinks, I want to keep tabs on him as he goes through his career. And then I followed your work at Salesforce and Pendo and now you're a demand base. And I just wanted to catch up with you and just chat with you about

Chad Holdorf (00:18.076)

You

David J Bland (00:35.898)

you know, leadership and testing and how you're approaching things and all this crazy AI stuff. But maybe before we jump in all that, if you could just give people just a little background on yourself and how you got here.

Chad Holdorf (00:48.032)

Yeah, I mean, like you said, we met at some conference. I followed you in the early days of Agile. I thought you were one of the early thought leaders and I still remember your blog. And I was like, this guy is spot on. Like totally, totally gets it. But again, yeah, met John Deere building GPS software to make tractors drive straight anywhere in the world. like, we went all in on Agile. had a

I still remember the day we had like 800 people in a conference room getting all trained at the same day. Like we were all in Des Moines, Iowa. you know, it's just wild times, you know? And I spent most of my career at John Deere going out to the farm, testing tractor software every two weeks, summer, winter, right? That was crazy times, you know? And then had an opportunity. Somebody heard from me at one of the conferences and offered me a job at San

Salesforce in San Francisco and I was like, you know, like I will always be there. So yeah, I spent a good seven years at Salesforce working everything from like agile development to infrastructure to app development, you know, deployed Hyperforce in like many different countries, which was just, you know, wild times. And then I left there to join a smaller company, Appendo, you know, really product-like growth.

It was awesome to work there because you're working at a company that's focused on product managers and you're a product manager, so it's like you're building for yourself, which is just amazing. And then I joined DemandBase because I...

wanted to learn more about go-to-market and how marketers and sellers work, you know, and it's just, again, it's a different persona and just a different world. And now I'm, for the last year and a half, I've been leading all AI things at demand base. And it's just, it's just, it's, I can't say, it's just fun. It is apps, it is.

Chad Holdorf (02:52.83)

fun every day to take three steps forward and five back and 10 forward and three foot, you know, every different direction. It's, I can't get enough of it. I feels like the early days of Agile of just like trying to understand how things work and you know, learning from others as quickly as possible. And you'll never learn at all, but if you don't keep trying, it's just, you know, you'll.

You won't learn what the latest is. And again, you learn something every day and it's exciting to hear from what other people are thinking about it too.

David J Bland (03:25.444)

Yeah. And you all just had a lunch at DevanBase. Tell us a little bit about that.

Chad Holdorf (03:29.317)

Yeah, yeah, I just got back from London. I was there for seven days. We had launched Demandbase AI and I'm happy to show you what that is and talk a little bit more about what it is. like Demandbase is like, supports both go-to-market teams. So sales to, or for marketing to sales to operations. there's just a lot of data that people have to look at to make decisions and.

looking at dashboards and 40 different screens and trying to make decisions, it's just too hard. And those days are probably gone.

the capability where you can just interact, have a chat with all of your data from ads to firmographic to technographics. And you can just ask like, are my ads doing? Which one should I invest in? Which accounts are progressing? Which ones are stalled? You know, it's just like, you're just for conversation and like that then leads you to like what accounts you should focus on and which ones are like top of mind for people to like focus on.

It's just now a simple chat UI that leads you to different parts of the product. And I heard from customers all week that they're just like, thanks. Thank God. I could have to like spend all day digging through data and like trying to understand which accounts to focus on because, you know, there's just so many of them and you only have so much time of the day. And so customers were just like, just absolutely ecstatic that we're just making their life so much easier. And

I love going to these kickoffs to just, work the demo booth like all day long. You know, I'm just like, bring another customer, bring another customer. Let's log into your tenant and just like use your data. like, you know, I don't have any problems of showing the good, the bad, the ugly. And it's just, it was amazing. Just one little story about going to London. So.

Chad Holdorf (05:26.956)

My trip to London, was like, let's use Claude. Let's have Claude build me an entire agenda, entire itinerary. I've been there for eight days. Let's just let it rip. And it built me the entire itinerary. Put it in a spreadsheet. I fed it into Claude code. It built me a, it's called a PWA, like a personal website, but basically a mobile app. It's all, it's working. I love it. It's telling me how awesome I did planning.

Day two we go all the way up to like nodding hill from down We're like 30 minutes away. We we do a bunch of things. We had an awesome time. I wake up the next day I'm looking at my itinerary. It's like sending me back up to nodding hill and I'm like The stupid thing like why don't we do like? Location based area planning to do all the things in the same day. Like why am I going back up there? You know, it's just it's just AI is

Even Claude is not that good about like spatial, what would you call it? Like spatial mapping of different places all in the same area. And I am just like, so then every day I'm like, okay, I gotta go back and look at my itinerary. You know, and I just, it's awesome. It feels magical. But then like when you get down to the root of it, like it's still not great in certain areas. And how I was talking to customers about this, like if you and I were, let's say you lived in London and I...

was visiting and we had to go like drive somewhere and the Google Maps says turn left, but you've been at London your entire life. Let's just hypothetically say this and you say no turn right. Like who's right? Google Maps for saying go left or or like you saying go right because you're local, right? And like I met with a customer talking about our AI and

they wanted to run a campaign that said, like, let's advertise to all customers that have a headquarters in Germany. Okay, sounds pretty easy, right? Like any AI should be able to do this. And what's crazy for me is our AI picked up a field in their CRM called HQ underscore locale, and it found...

Chad Holdorf (07:51.299)

the Germany companies that had headquarters in there because that was their custom field that they've like built that their entire like existence, right? So, but my customer support person didn't know that field existed and they were trained to use a field called billing code equals DE. Okay, and that generates a different list to focus on. Okay, so our AIPipe picks up this field that the customer was like, wow.

but our customer support person's like asking me like, hey, I think it's wrong because this is my list. you know, just, who's right in those things. like, you know, so you would think like the location to like driving directions to like just advertising and fields, like it really opens up a Pandora's box on how do you measure

accuracy of these things. I mean, I heard about, I'm deep into evals, but like that, it's just, that stuff is just really, really interesting and really makes my head scratch on like how we should do it. But anyways, yeah, that was London talking about AI and trying to explain to customers like sometimes AI is smarter than somebody that doesn't know and what somebody else may think. I think that's just.

It's just hard to explain to a customer. That's been my last few weeks.

David J Bland (09:26.799)

It's only got Whirlwind. I I love that you're all in. think something that you said though really struck a chord with me. So when it comes to data, and this keeps coming up time and time again, we talk about testing and how we make decisions. I see a lot of companies feel as if they have to put AI into their workflows and they're really trying to grapple with that. But they're layering it on top of data sets that are just like not connected. And what we're seeing

Chad Holdorf (09:39.224)

Hmm.

Chad Holdorf (09:54.702)

Hmm.

David J Bland (09:56.368)

at least what I'm seeing, and I'm curious what you're seeing. I've seen people make decisions off different sets of data, and then when they get bubbled up, people are like, well, why'd you make that decision? Oh, because similar to your customer support story, I was looking at this, and this is what the data told me. And so how much of this are you focused on when you think of, how do we pull this data together in a way where, if we're gonna hit, we're gonna hook Claude up to this or something, it has to...

Chad Holdorf (10:08.267)

Yeah.

David J Bland (10:22.657)

improve, know, improve our productivity. Like how are you thinking about data and testing and just that single source of truth?

Chad Holdorf (10:27.63)

Oh man, it's like, yeah, it's like, we too like thought like, oh, let's just put Claude on top of this data and it's going to be like easy, you know? And we tried that. And again, another story, I showed it to a really large Fortune 20 company, how you can put AI on top of this data. And...

We didn't do any training, we didn't do any labeling, we didn't do any human in the loop, we didn't do anything, just were like, let's just try it. And the story, I showed it to a customer and again, they...

I typed in the system like, hey, tell me about this company that I'm meeting. And they are on the line. They're on the phone with me. They're on Zoom. And it said that the person that I'm talking to has visited our website lately. And they've been visiting our job page. And it says, this person is likely to leave their company and join Demandbase. And the person is seeing it on the call with me. And I'm like, showing it to them. And he's like,

I've never visited your webpage. I've never visited your job description page. And we don't even label our, we don't even tag our job description pages. So it's like, then I type in there like, are you sure? And it's like, sorry, my fault. And the guy's just laughing. And I was like, you know, I'm like, I'm telling him, listen, you can get a.

easy demo all day long of any company that's going to make it look like it's absolutely amazing. But when you use it with your data, you are going to find it's a hundred percent. I wouldn't say a hundred, but like 99 % wrong. And we just spend a lot of time labeling, testing, validating, improving our system prompts, looking at the SQL queries, you know, doing everything like thumbs up to thumbs down, you know, and we just

Chad Holdorf (12:35.604)

It's just we've spent a lot of time in the last six months doing everything we can think of to make our data accurate underneath of a non-deterministic system. And I think that is just really hard. And it is really easy to fool somebody giving them a demo with

whatever data you think and make it look amazing, but as soon as you give it to a customer and they sniff something out that's not right, like they're just like, I'm gone, I don't need this thing. And there are some businesses, I mean, ours is just like advertising and revenue, but that's still like, if you're off on which accounts you're telling your sellers to focus on, it's lost revenue, lost pipeline. that's just damn important is what I'm experiencing.

Every hour. So...

David J Bland (13:33.039)

Yeah, it feels like trust.

Chad Holdorf (13:36.131)

Yeah, I thought it would be easy, it's like, it is every customer can sniff it out within a few prompts. And we just spend a lot of time in beta testing it, validating it, checking with the customers. Thank God we have great customers that have been giving us feedback for the last two months. And you just can't, I just wish there was a way to like fast forward it, but I haven't been able to find them.

David J Bland (14:05.699)

Yeah, no, I appreciate you being transparent like that. When I talk to companies and they're trying to add in, know, AI to make user experience better or anything, I find I hear stories of, but we had to put so many guardrails on this thing that we even wonder if it's still viable by the time we were done with it. And so it's not as trivial as it sounds. And I'm wondering, has this changed?

Chad Holdorf (14:25.719)

Yeah.

David J Bland (14:34.031)

how you view this stuff in a leadership role. So when it comes to your process of getting tests out and getting things out quickly, what kind of impacts is having on your leadership, on your style of working, anything you could share as insights of what you see changing.

Chad Holdorf (14:51.566)

Yeah, I I have always been in the ship fast. What's the quote? Like, if you're not embarrassed on your first... What's that quote? Remind me. Yeah, right? know? So I'm like, it's like tattooed on my arm. Like anybody that knows me, like, I'm like, if you're not embarrassed, you you're ship too late, right? And...

David J Bland (15:03.743)

I think that was Reid Hoffman, wasn't it? Like you were not embarrassed. You waited too long. Yeah, it was too late. Yeah.

Chad Holdorf (15:19.628)

And I still do that a lot. And I think you just need to spend more. It's still like, it doesn't matter if it's AI or the old way, like you're still spending time in that beta moment or pre-release or whatever it is, trying to get that customer feedback, you know, and making changes. we still have immense amount of pressure to ship it GA and ship it GA.

We just try to set really strict criteria about retention, activation. And if we don't have those numbers for the set of beta customers we have, we just don't ship it. And I think our sales team is glad that we're not just like, just ship it to everybody and hope that it lands.

So that's still like doesn't for me it's like doesn't matter if it's AI or not you still have to go at least I feel like I haven't been able to like get around that. But in terms of like day to day it's I was telling you before this call like every Friday I go through our customer support feedback requests and every week I'm I'm probably doing anywhere from like 10 to 50 merge request.

And we have cursor, cursor is great. have clogged code. We use that. It's great too. But I spend 80 % of my time now in cursor, just like hooking it up to repos. And we have like 20 of them. I should show you, you should try to explain to users what we're looking at here. like, we have our own coding agent built into,

How do I share? Okay. I'll show my screen so you can like You can see what my my screen looks like and try to explain to people Yeah, like let's try to try to explain what the heck it is. Let's see. Okay entire screen Okay, here it is. So like here's we have this thing called Dev dev bot one of our engineers built it and I don't I don't know what's under the what's under the hood. So but

David J Bland (17:05.709)

I will narrate. I will narrate what I'm saying to you listeners.

David J Bland (17:27.375)

All right, I'm looking at a Slack channel at the moment.

Chad Holdorf (17:28.942)

And like how many repos do you you see you see me looking at?

David J Bland (17:32.469)

What?

That is crazy. You're just, it's just repo after repo. Wow.

Chad Holdorf (17:39.631)

And each of these threads are me investigating a problem that a customer has submitted. So this one is our core platform one right here that I'm showing you. And how many threads do you see? Do you tell the group that we're at? 56, OK? And here's an agent that I'm working on. And how many are in here?

David J Bland (17:57.252)

I see 56? Wow.

David J Bland (18:06.415)

That's 67. This is insane.

Chad Holdorf (18:09.004)

You know, and this is my day and I'm, you know, like here's, I'm filing, like, look here, let's go up a little bit to what I was working on. I'm trying to fix this agent to get it to respond correctly. And, you know, it's given me the exact places in the code where it's not joining data correctly.

You know, and it tells me what I should tell customer support. You know, and then it gives me like I type in some more questions that I got fed from Claude. And again, it's telling me all the places in our schematic semantic layer that is not correct. And, you know, go down here, it's still looking through our code and.

eventually I get to the point where I'm like, let's file a ticket. You know, and like, I can't tell you, I can't, let me show you if I can show you the ticket. Like you should read what this ticket is. And I'm not even a hundred percent sure what it is, but I am sure engineering will love all of this context that they can understand about what I'm trying to do. And here's all the prompts. Here's all the prompts that I was trying.

You know, and it lists the acceptance criteria that they need to meet and it talks about what's out of scope here. But this is probably like one of probably 50 that I did this week. And that's my day and I don't know.

David J Bland (19:36.047)

Wow.

David J Bland (19:51.545)

That's crazy.

Chad Holdorf (19:59.031)

I've worked on Salesforce and Demand and even Pendo and everybody has a large platform of like code and everybody expects product just to know the answer to everything, right? And traditionally like you would.

I would go ask an engineer, would send him a Slack and be like, hey, engineer, like, customer said this. And then guess what engineer says, have them send me a screen recording. Then you call back the customer, you get a screen recording, you know, three days have gone by. Then you like show the developer what they've seen. And then they're like, remind me what is this problem related to? You know, and then it's like, they're like, yeah, we have a support article already on that. So then you like go find the support article. And then you're like, but it feels out of.

I think it's out of date, you know, and then you go back to the engine. yeah, yeah. Let me check the code. Three more days go by. Right. And it's like this like ping pong that the PM spends all their goddamn time like looking through. And I, I now just go straight to the code and I just ask whatever somebody's asking me and I just am like, here's, here's what it says. And it's like 99 %

I haven't found a time where it's not wrong. It tells me the code and it gives me the example. Now I'm training our customer support team to just use it. So they don't even have to ask me. It's just great. They log in here, you would laugh. You would come into our customer support team and they would show the repos now. They're just customerized to this and it goes literally straight to the code.

David J Bland (21:16.591)

well.

Chad Holdorf (21:40.899)

gives them the answers and they're just, I think the next phase is they're gonna start submitting merge requests. It's gonna happen, I can guarantee it. They're gonna find little problems, text changes, and I am 100 % confident by the end of the year they will be submitting MRs for changes to our team and that frees me up, that frees lots of people up.

I don't know how engineering is gonna keep up with all the merge requests, right? That's another challenge. I'm sure there's a way, but like, that's what, that's what, it's just fun. I mean, it's, I am learning so much about our code base. I'm learning how agents behave and I'm literally just in my Slack channel asking questions anytime somebody has it. As you can see, just 67, 14, 41, 50. mean, these are, these are just.

101 like this, you know, like I was, and you can see this is our analytics agent, which is like part of my story today. And you can see 101. And I'm, doing this on my flight home from London. I am doing this on the walk in the morning. I, I'm doing it laying in bed. If I'm, something comes up, you know, it's just, it's, it's, it's so much fun. So much fun.

David J Bland (23:06.954)

It seems like it just fundamentally reshaped how you approach your work. I mean, compared to what you described to me before, I mean, that was going into weeks and now you're just having a conversation with the code and decentralizing. It's completely reshaping things, it seems.

Chad Holdorf (23:18.677)

Yeah!

Chad Holdorf (23:25.366)

Yeah, I should show you like, I submitted a merge request and just some engineer you can see here like, yeah, somebody tagged him and he just submitted, he just made the change. He's like, changes in dev, take a look. You know, and tickets there and I went into dev and it's working. And he's like, we'll ship it. We'll push it with all the other changes tomorrow.

And he's like, do more. Let's fix this thing. Let's make our customers happy. Let's kill this market, you know? it's, the engineers are like, do more, keep learning. And they're like, what do you want to learn, Chad? You know?

David J Bland (24:11.671)

Okay, I have to ask. I have to ask. How do you do this on your walk? Like what are you doing? Like what does that look like? Just explain to us.

Chad Holdorf (24:18.222)

Yeah, so it's like this morning I somebody from customer support, let me see if I can remember who messaged me this morning No, I can't. yeah, I think it was Anna I think Anna sent me a message of like she sent me some problem a customer was having and I just I copy the problem Okay, so I think she filed it in our we got a slack feedback channel. Okay, so let me just show you

just like any team, here's, what do you see David? Like same old, same old, like a list of problems, right? And Anna submitted some issue. Let's see if I can find the one. It doesn't matter which one it is, cause you can see that I'm working on all these. she submitted this issue here called enable UTM scoring in engagement points. I don't know what that means, you know? And she's like,

David J Bland (24:55.213)

Yep, yep.

Chad Holdorf (25:17.516)

She's talking about, she gives me all of this information. So she types it all into this ticket, right? So I just copy, copy everything. And again, I'm on my phone. I'm on my dog walk at 7 a.m. She's based on the East Coast, I think. She files a ticket, signs it to me. I copy it and then guess what? I head over to my DevBot in a different channel and I literally just.

I go in there and I literally, know it's in our platform and I just go in there and I literally just, I just paste it and it comes back with like, here's what's happening. Here's the problem, here's the code. And then it's like, would you like to do a root cause analysis? I'm like, sure, let's do one. And it spits back like, here's the problems, here's what's happening. And then it's like, would you like me to plan a change? What do you say?

David J Bland (25:59.875)

Wow.

Chad Holdorf (26:17.204)

Sure, why not? Right? And I'm just like, again, walking dog, 6 a.m., still dark, know, walking around downtown San Francisco. My dog's enjoying life. I'm outside, she's outside walking. And then it's like, you know, it just responds with, here's the bug, would you like me to submit an, or here's the issue, would you like me to submit an MR and assign it to the team? And what do you say? Sure.

Like, you know, like let it rip. And then it's like, here's the MR. It's a sign to the team. I get home. One of my team's like, hey, yep, this looks good. You know, and they just changed the, they merged, they do a merge request, I think. I don't know, they do their engineering thing. Like I don't ask questions. And I ping Anna back, like less than an hour, maybe two. Depends on how long the walk was.

David J Bland (26:46.328)

You

Chad Holdorf (27:16.558)

I come in here and I'm like, Anna, here's the changes on dev. Do you wanna take a look? And she's like, yeah, take it a look now. And she's like, cool, looks like it works. And then I just tell engineering and Slack, let it rip. And like, it's deployed. we could do hundreds of those a day if we wanted to.

David J Bland (27:44.943)

It just seems so different. mean, you're in a leadership role. You've been in leadership role for long time. What does this do? mean, okay, maybe a little more specific with that question. You are bringing, let's say product people on and you're trying to really embrace this and see all the ramifications of it. What are you looking for? What do you see yourself looking for when you're hiring people now?

Chad Holdorf (27:48.83)

I what would you- what would you- Yeah!

David J Bland (28:14.636)

What, does this change even the hiring process, how teams work with each other, like roles and responsibilities. It just feels like this has ripple effects across the entire org.

Chad Holdorf (28:18.369)

Yeah.

Chad Holdorf (28:24.398)

Yeah, I've been down just like my Agile days. I go down these rabbit holes and try to understand how things work. And you and I saw it, I was ahead of the curve. could...

You know, like when we were doing Agile, like we didn't have all the tools. So what did we do? We had a whiteboard. We're just, you know, find a blank wall, put some tape down the middle, right? Like put the sticky notes on the board, take down the cubes, like move engineering product and testing closer together, right? We're just, you know, we don't need all the tools, but like we will, we will figure out all the, all the permutations. And then like, when you tell, when you teach like two or three people, they're like, I get it.

You know, like I get the agile way, right? And I think every product person in engineering, design, they all wanna stay relevant and they want to...

All product people that I've ever worked with have a growth mindset and they just want to learn. They're gonna learn and they're gonna like bootstrap and try things out and they have no, most people have no like, are we following the process? Are we checking with blah, blah? It's just like, yeah, this is so obvious.

And when I work with my team, my team or other peers or even like during interview process, like I think there, I just look for are people willing to continuously learn and try things. And I think most, most people I have and like, I just remember like one of our, again, I think of somebody else on one of my other teams. He's like one of our, name's Tom. He, he does,

Chad Holdorf (30:21.602)

He's on the expert team, okay? So he helps our customers become expert on how to use our product. And he's super wild and knowledgeable about how our product works. And I showed him how he can look at the code and make his own changes.

And he's never done development. He's never like tried to figure out how to fix something for engineering. And he would always try to explain to engineering how things should be, you know? And there's like two ships passing. can't, they're not saying the right words. And he's now in cursor like all day long. He submitted changes and that's like less than two months.

He built a whole entire cursor mock-up for our London Go conference that they showed live on stage. And I was sitting next to him and he's like, my hands are sweating. Is this what product feels like when they ship things? He's like, I don't know if I like this. And he's just like, this is amazing. And he's watching them click on things. again, it's it's a bootstrapped, somewhat fake product. he's...

He's like living on cloud nine. mean, and again, he's like never been a software engineer and he's on cursor, Versal, Versal, Versal. What's it called? Versal, Versal, Versal and Git like all day now. And he's now making huge changes and giving me great ideas. And we...

We used to spend time together writing stuff in Google Docs and trying to explain it and blah, blah, blah. Now we're just like, open it up, open up Cursor. Let's just make it, let's just do it. Let's just get the MR out there. And I think it's just like the old days of Agile where you just work with people all day long rather than having these handoffs back and forth. And I just, people are...

Chad Holdorf (32:22.126)

My design team, our engineers, our product people, they're definitely trying to grasp it as quickly as they can. And when they see it, I see some teams now, the number of merge requests they're doing, I am just like, God, how fast can we possibly go?

David J Bland (32:46.574)

It's, I mean, I was talking to a product leader. They used to coach at Realtor, Dave Masters. And he said, I'm shipping code. He's framing it like zones of work. Like they're different zones of work and you have to adhere to a zone. I'm really hoping he writes more on it publicly, but it's, I keep talking to product people. I keep talking to people who never shipped code and this has unlocked something for them. And I don't.

Chad Holdorf (32:58.349)

Yeah.

David J Bland (33:13.014)

I think we're just scratching the surface on this. Like when I first met you, I'm pretty sure I was trying to convince the Azure community that like Lean Startup was coming and you guys should listen. And you were like one of the few, because you're just like, yes, yeah, I know. This feels like that a little bit. Like back in the day, you and I hanging out and you were talking about like, hey, we got to have assumptions on this stuff. We got to...

Chad Holdorf (33:15.104)

Wow.

Chad Holdorf (33:24.246)

Yeah, I'm like, I get it, I get it, I get it, I get it. Don't tell me twice, where do I get it, right?

It does.

David J Bland (33:39.116)

validate this stuff. It's not just about iterative delivery. It's about the outcomes and the impact and this, I mean, it's different, but it feels similar.

Chad Holdorf (33:48.889)

feels the very same. I, and you and I know where this is gonna go. Like the days of Agile, cause you can imagine now there's gonna be like certifications. There's gonna be like, you know, conferences of like how to set this up. Like I can see it a mile away and you know, it's.

There are some people that learn that way by having a session and recording and learning it every year. you and I, back in the day, we're like, somebody comes out with some new agile tool and they were like, well, let's try it. Why not? What's it gonna hurt? And I just think that that's what the new...

new way is I was gonna show you like what my cursor looks like and just have you be like, what is this guy doing? Like this, this is my, this is.

David J Bland (34:35.48)

David J Bland (34:39.598)

Okay, I'll be honest. I'll be honest with our listeners here. I have still not gone all in on cursor. Now, Claude code has pulled me in, but cursor, maybe if I had started there, all right, pitch me on cursor. Like what is it you're doing here with cursor?

Chad Holdorf (34:46.446)

Yep.

Chad Holdorf (34:55.586)

Yeah, mean, so my, my clog code that I use personally, like it's usually for a few repos. Just another story. I had to do my taxes this last week or whatever it was. And I made a bunch, I did a bunch of like a goodwill donations of clothes, right? Right. And you got to keep track of that. And

There used to be this website called It's Deductible. Did you ever remember It's Deductible? It like an Intuit. It got shut down. It got shut down like three years ago. And it was the simplest thing. would just type in like men's shirt and it would find it and you would just be like, yes, 10, you know, or five. And it would just build a spreadsheet and like take the fair market value. And then you could just ship that to your tax person and have them like enter it. Well, they shut it down and I've always, I never could find one.

David J Bland (35:22.784)

no.

Chad Holdorf (35:46.831)

So I built it on my flight. Okay, but it's just one repo and you're just like, you're just going back and forth and like, it's pretty like isolated, you know, and you're doing a deployment. There's not any other repo dependencies. I don't have to like think about six or seven repos, but this is my cursor and you can see like on the left-hand side, like I've just like in my Slack channel, I have a bunch of different repos and sometimes I have to.

ask questions across multiple repos. You know, and I haven't, I haven't tried to figure out how to do that in Claude, right? To ask a question across many repos to be able to like make a change to three or four or five or 10 repos at one time. So what I like, what I like about cursor is like, I can see all of my repos over here on the left. Okay.

David J Bland (36:36.654)

I see.

Chad Holdorf (36:45.26)

And then again, it's very similar to Claude that like here is all of my, you know, I'm just having a conversation and on the right, best case I can do is I usually can have like six or seven parallel tasks going on at the same time. Right? So these are all my, they call them agents, but they're just like make changes to many different repos. Right? So if I'm, if I, I've seen some of our engineers here,

you know, at demand base and they have like 10 to 15 of these going on at the same time. I don't know how to keep track of them. But again, like when you're in the groove, like, you know, you're firing them off, you know, and I'm not advanced. I'm not that advanced on using it, but I know how to use it. like, you know, there's one repo over here that's a new team that's doing some new innovative stuff. And I'll just type in a simple example of it. I'm not going to say the name of the project, but it's...

It's over here on the left. And I would just simply say, give me a summary of this repo and explain the product vision. And I'll just type in, I'll just say this name of this project, right? And I'll just ask it. Normally I would go ask the PM, I would schedule a meeting, right? They would give me a Google Doc.

It might be updated, it might not be. It not include all the questions that I have. They might not know what's got latest checked in. It's just a big delay, not getting right to the source. But now, just with cursor or anything, here's, it's given me a summary of everything. Here's the vision.

David J Bland (38:35.15)

wow, that was fast.

Chad Holdorf (38:41.44)

I could ask all kinds of questions about this. You know, and it's given me the architecture plan. I'd even ask about the architecture. It's given me use cases. I guess I should have asked. But I might have had to like do that on the fly with the person, right? And these use cases change, you know, every month, every week. And it just tells me what it is. And that's just amazing.

David J Bland (38:54.285)

Ha ha.

Chad Holdorf (39:11.49)

You know, and what I do with my team most weeks is I'm just going to show you like this. This is the future where I want to take the product. And this is is this is built in cursor. And I have links that go directly to like, do you want to read more about the vision? I just put it all all here. They can read it. You know, and if you scroll down here to the bottom, like if I want to put this in a Google doc, I just click here and it puts it in a Google doc.

David J Bland (39:40.846)

Yeah, I see that.

Chad Holdorf (39:42.103)

Right? And my team and I go through flows, like use cases and how things will work. And it's just, I could build one of these in like 10 minutes. You know, and like this, as you could imagine here, like these things across the top are just gonna be agents working with humans on a list of objects. And you're just handing work off. Right? And guess what? Like that.

that probably will also show up in something like Slack. And I just, again, I just prompted him and said like, make it look like Slack. And I show this to customers all the time and they're like, I get it. I get where you're going. But like, you know how much time this would have taken with the design team and back and forth and like building a mock-up and creating the steel thread and writing down the steel thread and like on and on and on. Now, my designer and I, we just are like, let's...

We got an hour, let's let it rip. You know, and she gives me great ideas, good suggestions. I meet with some of our users, they have questions and I just, guess what? I just rip them into cursor and like, just let it rip. And then David, what's wild is, if I show you back here in cursor, like, let's see if there's an example, like, I will just tell cursor to make.

make this an epic in JIRA so our engineers team can tie the work to. And I just have an MCP to JIRA and I don't ever have to go into JIRA. just am like, put it in there, let them hook up the tickets, let it go, like we're off and going. So my engineers are like, I get it, I see it, I understand it, I can see your repo, I might use some of it. And they're...

They're thankful that it's so clear to them where we're going and what they need to do first and why we're doing it. And they can go through the visuals. They can read the description. You know, I can send it to our CEO. I can send it to my boss. I can send it to anybody and they can click on it. They can send me questions. You know, and I just, had one today. She's like, our CMO, our CMO. She's like, I want to get into Cursor. I'm like, my God, our CMO.

David J Bland (41:59.181)

Ha ha.

Chad Holdorf (42:00.137)

Can you imagine that David like when have you ever heard of a CMO? Maybe I don't know but a CMO wanting to get into Cursor to like build their workflow? You know? I can guarantee it will happen by the end of the year that she will be in Cursor having a conversation, we'll be in the same room and we'll be just we'll be building out her workflow that she thinks that we need to do for a specific customer. mean she sent me this one and I was

I was starting to build in. When I say build in, like just think through and she was like, yep, no, change this. this doesn't make sense. Let's not do that one. And that's crazy.

David J Bland (42:31.629)

well.

David J Bland (42:39.479)

You know, there was like a time, yeah, there was a time in my career where I just came out of my first startup I joined and we were like software as a service, but as before, I'm dating myself with this, we were called the application service provider like back in the day. we literally, like it was a point in my career, it was early on, and we were like, crap, we don't want people logging in, like we want them to authenticate and how do we do that? Nothing's built. And we just literally like rolled our own.

Chad Holdorf (42:54.658)

Hahaha.

David J Bland (43:08.141)

single sign on and all this stuff. And there was a time when I came out of startup, I joined Nerd Startup. was like, you know what, maybe I wanna be an API developer. Maybe I just wanna dig into that. It was not for me. I gave it my shot. But this is the first time probably since then I come back and I go, oh, huh, I kinda get it. Like MCP, maybe I'm not logging into all these things anymore. And it reminds me a little bit of how disruptive that was back in the day where we were kinda like, how do we connect this to this?

Chad Holdorf (43:18.286)

No, no, no, no, no,

David J Bland (43:36.769)

it doesn't exist. Let's build it. Let's do it. And so with you, and just, mean, I really appreciate you being transparent and just showing me all this stuff today. Where do you think this is going? Like, where's the constraint moving around inside orgs that go in on this? Like, where's this headed? I mean, I you can't predict the future, but you've seen a lot and you're in this. Where do you see this going?

Chad Holdorf (43:49.347)

yeah.

Chad Holdorf (43:53.495)

Yeah.

Chad Holdorf (43:57.623)

I think what I'm noticing again at this moment, it probably will move around as always. Like I think...

Again, I hate to say this or hate to think about it this way, maybe it's fine, but like back in the days when we started Agile, we did two week sprints, right? And, you know, engineering got fast and then guess what? Like the training department, the enablement team, the sales team, they're all just like, whoa, wait a minute. We used to ship every six months, you know, we had like long enablement sessions and blah, blah, blah. We could predict and you know, and.

And I think, you know, most, most go-to-market teams got familiar with the two-week sprints and incrementally, blah, blah, blah. But, but now it's coming at them 10 times faster and 10 times more. And...

You know, like we're having AI pricing conversations and token utilization followed by like 11 new major features coming out. You know, and I see our customer support team and our sales team like, what is all of these things? On one side they're like, yeah, more selling. You know, but on the other side it's like, I gotta explain this to our customers.

But I've been training, like I said earlier, training our customer support team that they can now use this Slack channel to talk directly to the code. And they can ask questions, like, I'm like, I'm not writing any more enablement documents for you. I'm, that these are, those days are over. And once you show it to them, most of them, their light bulb goes off. Like, I get it. get, yep, I get it. I don't need enablement. You gave me the repo. I...

Chad Holdorf (45:35.277)

I'm good. Whenever a question comes up in my head, straight to Slack, ask a question, get an answer. And the really, really forward thinking customer support team is just like, this is awesome. Like our technical account management team, they're just like, you know, I hate to say it this way. Like they're like, we will buy more cursor. We'll do, we'll get, we'll pay for more for cursor because that will scale us.

David J Bland (45:49.206)

Chad Holdorf (46:03.502)

You know, and I wish cursor was funding this pod. Can we do a cursor pod? Is this where you do like this episode sponsor? Why? Yeah, right. Like go to David.com slash, you know, cursor for your 20%. We should totally have that. Do you think we can get that sponsor? Anyways, you know.

David J Bland (46:07.307)

Hahaha

David J Bland (46:12.141)

Chad Holdorf | How I Tested Pull Requests

Have a Listen

Summary

Guest Links

Transcript

Büşra Coşkuner | How I Test B2B vs B2C