Why AI Doesn’t Replace Product Thinking: Insights from Stephanie Neill, Stripe

In this episode of Between Product and Partnerships, Cristina Flaschen sits down with Stephanie Neill, Head of Product at Stripe, to explore what’s actually changing in product development in the age of AI, and what isn’t.

Stephanie draws on experience across government, Twitch, and Stripe to explain why the pressure to “just use AI” often leads teams in the wrong direction. Instead, she makes the case for grounding every decision in the problem to be solved, not the technology being used. The conversation dives into where AI is genuinely useful today, where it still falls short, and how teams can use it to move faster without compromising quality or trust.

They also explore the operational realities behind the hype, from messy, unreliable data to the risks of deploying AI in high-stakes environments like payments and tax. Throughout, Stephanie emphasizes a consistent theme: AI can accelerate good product thinking, but it cannot replace it.

Who we sat down with

Stephanie Neill is Head of Product at Stripe, where she leads teams focused on payments infrastructure and tax-related products. Her career spans e-commerce, public sector innovation with the United States Digital Service, and creator monetization at Twitch.

Stephanie brings expertise in:

•  Applying product thinking in high-stakes environments like government and financial systems 
•  Building platforms that enable businesses and creators to earn revenue 
•  Scaling product teams across complex, data-dependent domains 
•  Navigating emerging technologies like AI within regulated, trust-sensitive systems 

Key topics

• Why “use AI” is the wrong starting point

Teams are often pushed to adopt AI without a clear problem in mind. The right approach is unchanged: start with the user problem, then evaluate whether AI meaningfully improves the outcome.

• Where AI actually works today

AI is most effective in accelerating discovery and iteration, helping teams research faster, test ideas, and explore solution spaces without heavy upfront investment.

• AI as a tool for reducing product risk

Product development is fundamentally about reducing risk. AI increases the number of iterations teams can run, allowing them to be wrong more often and converge on better solutions faster.

• The reality of data as a limiting factor

Even the most advanced models are constrained by messy, incomplete, or externally controlled data. In domains like tax, poor data quality becomes a major blocker to reliable AI systems.

• Why human judgment still defines production readiness

Despite rapid progress, AI outputs still require careful review. Especially in financial systems, the cost of being wrong is too high to remove humans from the loop.

Episode highlights

 11:00 — The industry-wide pressure to “just use AI”
 13:30 — AI’s role in product discovery and rapid iteration
 16:20 — Automating repetitive work with internal tools at Stripe
 19:00 — What happens to entry-level roles in an AI-driven world
 22:30 — Why data quality is the biggest limiter for AI systems
 25:10 — The gap between AI hype and production reality
 32:00 — How Stripe evaluates risk before shipping AI-powered features
 35:10 — Staying grounded by continuously redefining the problem

--

For more insights on partnerships, ecosystems, and integrations, visit www.pandium.com

Transcript

Sarah Elkins (00:00)
Welcome to Between Product and Partnerships, a podcast focused on bringing together product, partnership, and engineering leaders to discuss how to build, support, and scale SaaS ecosystems. This podcast is presented by Pandium, an integration platform for building native integrations.

Cristina Flaschen (00:18)
Hi everyone. And thanks for listening to our podcast between product and partnerships where we talk everything about the challenges and what it takes to build integrations, tech partnerships and SaaS platforms. And today I'm super excited to have Stephanie Neal, head of product at Stripe, join the podcast. Stephanie, why don't you give us a little bit about your background.

Stephanie J Neill (00:39)
Well, thanks for having me, Christina. It's great to be here. Yeah, so a little bit about me. So I am a product person by trade.

Cristina Flaschen (00:41)
Thank you for joining us.

Stephanie J Neill (00:48)
I've worked across, I guess, like some of the internet's biggest properties. Started my career in an e-commerce space, sort of content marketplace situation where I got to work with Lionsgate, I got to work with New York Times, just a ton of ⁓ really storied properties.

I loved it. I've always loved product management as ⁓ sort of an approach to life in a way. But I started to feel a little bit disconnected from the actual work ⁓ of just sort of working on. ⁓

making the numbers go up and to the right, if you will. So at some point in my career, I decided I wanted to look for something where I could find meaning and connection to the impact of my work, right? Because product management is all about impact. And so I wanted to know that I was changing people's lives in a more meaningful way. So I started looking around and I actually found a role

inside the United States government with United States digital service. And ⁓ for those of you who don't know, that's an outfit created by President Obama, I think.

it must have been like 2011-ish because it was around healthcare.gov when that fell over. He was like, hmm, we need to do some things differently here. Anyway, so he stood up United States Digital Service, USDS, which basically partnered private sector talent with ⁓ the public service workforce to deliver services people need. So I joined that group and I found really what I love, which is like applying the

product management, sort of like mental framework to people who have deep, meaningful need, where you can create impact that feels, that you can connect with, right? So I led the Homeland Security Group.

through a tumultuous time, but delivering services for the American people and our guests and basically the most vulnerable folks that the government serves. And then after that,

I actually ended up joining Amazon's Twitch, which that might feel, you know, I just talked a lot about, I wanted to build products that were deeply meaningful. And then you're like, Twitch, isn't that just like TV, like live streaming for ⁓ Gen Z? And the answer is like, sort of. So it is like a live streaming software where it's largely entertainment.

However, underneath that, or if you like, I guess once you spend time with the communities, you start to realize it's about, it's actually like a place where people find their people, which is deeply meaningful, right? Like it's where people sort of refine identity and like find others who share their identity, like across the world, like it's not always easy to find people you connect with. So that sort of connection was really meaningful.

⁓ And then I also, for most of my time there, was heading up the product team for monetization. And so there was like an added layer of like, these creators are building their business and we're basically making it so they can make a living doing what they love. And then...

somewhat recently, I guess like two years, well, two years ago now, I actually left and I joined Stripe to sort of carry on that thread of, ⁓ you know, helping people make a living doing what they love or doing what they're passionate about. And of course, you know, Stripe, it's more, maybe, maybe you don't know, but it's payments infrastructure, it's e-commerce infrastructure for the internet, basically. So the mission is to grow the GDP of the internet. And it's all about basically like making the, the, the,

like leveling the playing field so people can start up their own businesses and earn money and you know basically live the life they want to regardless of where they are.

Cristina Flaschen (04:34)
That government job sounds really interesting. I feel like I could do like, we could do like a whole thing about that. Like, I gotta ask, how did you find that job? Like, I didn't even know that was a thing that, I didn't know that was a department that existed at any point.

Stephanie J Neill (04:36)
Bye.

yeah.

No, so I feel like I got really lucky. And I feel like one of the lessons in life is like, the more people you know, the luckier you get. It's like someone has heard something and they can connect to it. But so I actually I grew up as a State Department kid. So government connected. And I, you know, grew up all over the world, met a lot of people. Actually, it was it was in college, though, in LA, where I met

Cristina Flaschen (04:59)
Yeah.

Stephanie J Neill (05:14)
best friend and let's see I guess I had been in the workforce for a few years maybe like six or seven years and I was complaining to her about like I just like I feel soulless and like it's fun but it's like like what am I really doing like I'm working on the whipped cream of the internet

because I think I was working on shopping sites or something. And she's like, well, you know, my friend just emailed me about this thing. And I know you have this government connection. Maybe that'll be meaningful for you. And then she forwarded me an email where her friend was like, hey, I just became an engineer at this cool place called USDS.

And so yeah, was like word of mouth, I guess. But I agree with you. I feel like it was such a stroke of luck that I found that in that time. And it was quite near the beginning. think USDS, yeah, I think it started maybe 2012 or something. So I guess it was 2016. But it took a while to start growing.

Cristina Flaschen (06:10)
Yeah. And I bet, we won't, again, I think we could do like a whole episode just about this, but like, I bet the pace maybe isn't the right word, but just like the exit, maybe the delivery cadence. I'm just picturing like when I've worked with really large companies and also work with some government entities, it's very different than working with like a small startup or even something like a Stripe. I wonder if that was like some.

culture shock or if it was actually like surprisingly nimble. I don't and to back up. Well, it's interesting because we had someone else on our podcast who worked with the FDA. So she didn't work for the FDA, but she does like ⁓ bio stuff and was talking about getting like approvals through the FDA. And she was saying it's actually like can be really fast, like, which was surprising to me. I just assumed that everything was going to be slow and crappy the entire time. I'm curious. Yeah, that we could, we could dive into other things. So I'm curious what your

sort of experience was if you're able to comment on it.

Stephanie J Neill (07:07)
So I guess there wasn't culture shock because I kind of knew what I expected and I think the way I would articulate it is more like the way government breaks down problems is much medium, like it's heftier. Like when you're in tech, you slice it down to like really the tiniest, narrowest thing that will solve a very discrete problem and then you keep kind of like layering on top. Whereas like in government, you know, like they'll have a roadmap out.

five years and they'll be very confident in it and they've just chunked out these massive problems for delivery. So that was sort of one of the key distinctions, I guess, that I observed. But at the same time, it's understandable. There's a bunch of different factors that can go into it, but I think it's hard when you're in government.

It's so regulated and the consequences of making a wrong choice are so serious, right? It can be life or death. And so people take everything really seriously. And it's hard to know all the written rules. And so a lot of times word of mouth will become, oh, well, I've heard you can't talk to users directly, it's illegal. And then you're like, really? And then you're like, let's look into that. so I think there's a lot that can build

into this less nimble culture. But I think there are some reasons that are good, which is you have to really understand the implications of your choices upfront.

Cristina Flaschen (08:38)
Yeah, move fast and break things does not apply to like literally everything, maybe, especially when it comes to like government government services. But I digress again, I feel like we could get into like a whole bunch about that. That's like a segment of

the economy that I just know so little about through direct experience. Like I feel like I'm a spectator and everybody's got an opinion when we're watching the evening news or like trying to deal with filing taxes or like whatever it is or signing up for the ADA or you know, whatever, whatever, or the ACA rather. Anyway, I just missed the fact that you were involved in that. So just wanted to get that out of the way.

Stephanie J Neill (09:13)
No,

I'll say it's the best work I've ever done in my career. I would have stayed forever, we, not me, but the people who incubated USDS intentionally made it term limited because we didn't want people coming in and basically becoming sort of like, like, because you're kind of closed off from like,

know, tech moves really fast and the government doesn't move as fast. So you come in and you kind of freeze a little bit and it's like you didn't want people staying and then becoming part of that problem. ⁓ And so it was like we wanted, we want people to go out and then come back in and bring fresh upon return, if you will.

Cristina Flaschen (09:40)
Mm-hmm.

That's such a nice, thoughtful way to think about staffing too. My God. Don't we love to see it. All right. So let's talk about something that is move fast and break things. We'll talk about machine learning and AI. Cause I know that you have a lot of experience there, a lot of thoughts there, maybe a lot of like hands on today. So maybe just starting at the very top. And this is something that I have experienced over the past like two years.

Stephanie J Neill (09:51)
you

Cristina Flaschen (10:18)
which is the impulse for companies maybe from the board, maybe not from the board, but from leadership to use AI, just use AI, use AI in something. And I feel like even just since we originally spoke maybe like a month ago, there has been like a shift already around like, it just moves so fast. Like maybe that wasn't the right thing. Maybe AI is not supposed to be like the panacea, but I'm curious.

Stephanie J Neill (10:24)
Thank you.

Cristina Flaschen (10:42)
I'm curious if you hear folks talk about that in that sort of way of like, I'm being told to use AI, don't actually know what I'm supposed to be doing and how you like to position that in your mind when you have folks that say they're being told to use AI, like they're not really entirely sure what that means or what to do next.

Stephanie J Neill (10:50)
Yeah.

Yeah.

Yeah. Yeah, so it is a broad problem or a broad sort of like state we find ourselves in. I remember, so I'm on a government listserv too. And I remember someone was like, we shouldn't tie funding to like AI as a solution. Like, you you must use AI in order to get this funding. It should be tied to the problem, right?

But I'll say also, you know, with my PMs over the past year, just working in Stripe, we work on attack software, you know, we've all been like, okay, like, let's get our hands dirty, let's start using because like, the tools are changing so much, and they're getting so much more powerful. And so it's important for us as product people to use it and understand, okay, like, how can this help me solve real problems?

⁓ But I even hear that from my PMs of like, well, it can't make production ready code. And so like, why are we messing with it? And I'm like, well, what else can it do? Like, let's explore. So I guess what I'm saying is like, it happened sort of.

broadly, like industry wide, and then maybe like less tech savvy places. But then even when you get on the ground, like there are PMs out there who are asking these types of questions too, even though it's like a huge part of our job to understand and adopt new technologies and think about how they can help us be more impactful. So yeah, the way that I think about it, I guess it doesn't really change. mean, it's like, I'm always like, what's the problem we're trying to solve?

And does it drive sustainable change, right, for like the people we're trying to solve it for? And I think the job now is thinking through like, what are good use cases for ML, like for these new technologies that can basically get us to that better end state faster, or to a better end state, maybe, and or more efficiently faster.

But at the end of the day, it's like, what is the problem we're trying to solve? can any technology help us?

Cristina Flaschen (13:01)
And how do you think about problems that AI or ML might be good at solving as it stands today versus those that maybe it's not a good fit or can be solved easier, faster, cheaper without using AI? Like where, I don't know if you have like a framework in your mind or just like broad themes of the types of problems that you feel like there's applicable AI technology that currently exists.

Stephanie J Neill (13:21)
Yeah.

Yeah, yeah, that's a good question. I think the spaces, the problem spaces where I've seen it be effective ⁓ primarily have been, I would say, discovery. helping you to basically, you you've identified a problem. Because, okay, so in product management, sort of, they're sort of like two key parts to product development. It's like,

Am I solving the right problem? Am I solving this problem right? And it's sort of like a discovery space and then a solution, oh sorry, like a problem discovery and then a solution discovery space, right? And I feel like,

problem discovery, like the research, the feedback, like a lot of the LLM tools are super good at like helping you beat up your thinking, helping you gaps, you know, helping you conduct research, helping you do analysis, like, so I think that it can be super valuable there. And then the second part, the vibe coding, this is where vibe coding, I think, comes in handy, where you can go in and you can start building out different ideas and then actually start usability testing these. And in some

cases, you can kind of hook them up and do like janky tests, you know what I mean? Or sorry, not super janky. But you can run tests with real users, right? Because the first thing, product development is all about reducing risk. So you want to reduce the risk on, is this a good problem? Is it painful for the user and is it good for my business? And then on the solution, does it actually solve their problem? Do they understand what's happening? Are we going to get them there? And I would say,

historically, like that those two pieces each take a lot of time and I think that AI really helps shrink on both ends. And especially with the solution with the vibe coding solution testing, it's like you would you would previously have to really invest a fair amount of time. And so by shrinking that time, you just have a more at bats, like more more chances to be wrong until you finally write and then you build that thing.

Um, so I think those are probably the two big ones. And then I would add a third category, um, which is like, you can get agents to do a lot of things that don't, in my opinion, that don't require, like, well, actually everything you're getting AI to do still requires human judgment. Like, like even like the research or even, you know, writing, whatever you still have to go through and like with a fine tooth comb. Um,

But it's like all those sort of like repetitive tasks, like where you have to, I guess, you know, you have to like hunt and pack, or you have to compile information and you have to like think through ⁓ all of that. can do a ton of the legwork. Like we built a, we built a tool here at Stripe that helps with deal qualification. And it basically will reach out and like pull, you know, everything we know about a certain user, everything that we know.

Cristina Flaschen (16:14)
Mm-hmm.

Stephanie J Neill (16:27)
the questions we would normally run through to see like, this a good fit for sales or should they disqualify early? And so like we built out a tool and it's significantly helped the team's productivity. And that was something that used to take them hours and hours to do per week, right? So just sort of these like repetitive types of tasks where then there's like, where then a human can make a judgment and run is like, it's just hugely helpful.

Cristina Flaschen (16:55)
Yeah, it's interesting where you were talking about like the human oversight and human in the loop. I've been following when I see pieces of content and stuff come up about the potential for a shift in entry level type of white collar work as it relates to AI and like, are we going to have entry level like in its current form, white collar work in the future or is it all going to be automated away? And I was, was reading something,

And it was making the case that like, you do need a human in the loop and that's the perfect thing for like your most junior entry level employee to do, monitor all your agents. And I was like, that feels like the junior person monitoring, like even the dumber junior person, like the dumbest junior person that you have, which is the computer. Like that really didn't round it out to me, but it does, it does like, you

It does beg the question, what is that gonna look like in the future? What you're describing, and I totally agree, you need human oversight, and everyone I'm sure that works at Stripe at this point almost, even the most junior folks are entering into an environment that is not 100 % AI, because that's not a world that we live in yet, but what happens when so much of this task work is automated? What type of human being is going to be the person that you're looking for to?

provide oversight to the agents. I don't know. don't, I don't think there is an answer yet, but I do feel pretty confident. That's not like get your interns to monitor your agents. Like that feels like really not.

I don't know, maybe I'm wrong. Maybe like you do just need a human to like, just be there to like press enter every once in a while, but it doesn't, to me, it feels like it would be the opposite, right? Like the agents are replacing so much of that work that like someone who is not really super well experienced is doing. And you need someone who has a little more experience just like, you know, with engineering now we're saying like engineers are actually code reviewers now more than they are just like keyboard jockeys. And like that makes sense to me, right? But you, I don't think you want to let loose like your most junior person to be the one.

Stephanie J Neill (18:45)
Yeah.

Cristina Flaschen (18:54)
providing the oversight for the robot. But yeah, I'm curious where you land. Like, this is just, I'm riffing.

Stephanie J Neill (19:00)
I definitely wonder because, so I have seen models, you know, of like, this is the future. And it's like basically like execs at the top and then super ICs under them and then tons of agents. And like the super IC is the one managing the agent. But I actually, my intuition tells me that there is, there is at least like a lengthy interim step in my mind where

maybe these like junior people are the ones who are managing the agents and like building the agents and

they just don't have like the experience or the intuition, but they start to build that, right? So it's like they apply their judgment, their judgment is checked and corrected and they learn. And then the models learn, right? And then, so I don't know, I think there's something interim happening there, ⁓ but I think it does make sense. Like it's just, it's like a new style of, it's like a new form of learning, right? Like for us, and it'll challenge us in ways that we haven't seen yet. ⁓

Cristina Flaschen (19:39)
Mm-hmm.

Stephanie J Neill (20:02)
if you're managing fleets of agents, right?

Cristina Flaschen (20:05)
Yeah, yeah, mean, I think this is definitely obviously like a seismic shift in the way work, digital type of work, computer work is done. And, you know, my hope from an engineering perspective, especially, is that like this will allow folks to not have to.

do the equivalent of sitting on an assembly line, just like pulling a lever every 30 seconds for their entire life, right? Like they'll be able to focus on things that are more cognitively interesting and have a higher sort of barrier to entry when it comes to like creative thought. I'm interested to see how long it takes us to get there and also like what that path looks like, if that makes sense. ⁓

Stephanie J Neill (20:24)
Yes.

I'm going to this thing.

I think there is no shortage of, so I run a PM team and I also have a sizeable ops team that includes like tax experts and I would love to apply AI there. think there's like, we're in the process of applying like AI to different use cases there. And what I intentionally make clear to the employees is like, listen, like there's so much high value, like knowledge work that I need you doing on top.

up your time so you can just have have like stronger greater impact in a way that you can uniquely provide and I firmly believe that that is there and it will be there for a while.

Cristina Flaschen (21:22)
Yeah, I mean, think, and I'm curious what your take is on this. lot of this also, especially when it comes to like doing math, right? Or like taxes, which is a bunch of rules, very complex rules. You can get creative with those rules as a human, but like there is just like math and rules involved. A lot of this relies on data, really potentially large volumes to have patterns, right? And always just like clean inputs, which I think is, I feel like is going to be a challenge for

Stephanie J Neill (21:41)
Yes.

Cristina Flaschen (21:50)
a long time. ⁓ For companies like Stripe, you guys are probably in a great position because you are getting all of that, all of those numbers more or less like from the source, right? And like you guys are the source of truth, but a lot of companies don't have, you know, clean digital access to that type of data companies and other industries and or they're too new or you don't have.

Stephanie J Neill (21:52)
Thank

Cristina Flaschen (22:14)
the like years, many months or years of consistent data to be able to feed into a model in my opinion. I'm curious like how you think about the role of data, data hygiene and like the current state of the world when it comes to enabling folks to use AI and ML to the model's full potential.

Stephanie J Neill (22:33)
Yeah, in my opinion, this is like the greatest limiters because what's the saying? garbage in, garbage out, right? I think this is the greatest limiter right now. And yes, I think some companies are certainly in better.

in a better state, I think, than others. ⁓ And I think Stripe is in pretty good shape. But like on my product, for instance, so we have to do a lot with tax rules and regulations, which change, and we have to rely upon jurisdictional government's data, which you might imagine is not as ⁓ in as tip top shape. And so, you know, but like there's, it's not, ⁓ it's not just cut and dry. ⁓

It's always complex. And again, I see it as like the bigger, one of the bigger rate limiters ⁓ because it is hard to, it is hard to clean up and leverage data, especially when you don't control it directly.

Cristina Flaschen (23:29)
Do you have challenges with that in your, maybe in your current role or maybe with folks that you've worked with outside of Stripe with this sort of the universe understanding that as a limiter? And I asked this because I, know, we integration platform, all we do is data all day, every day. Like I have been knee deep in data of all different types of industries, all different types of volumes, all different types of patterns, like for 20 years. So.

I am not a data analyst, but like I can tell you all the whack patterns people have come up with. And when I, when I talk to folks, especially like really hardcore AI evangelists, I think there's a very large disconnect between me and them sometimes around the severity of this as a limiting factor, so to speak, that like it's actually, it could be really, really hard to get some of that, just to get access to some of it and then to try to understand what it actually means in a programmatic way.

Stephanie J Neill (24:03)
you

Cristina Flaschen (24:23)
Right. And I'm wondering too with Stripe, like, are you guys actually trying to get like flat files from some of these jurisdictions to like update your tax codes? Cause that's like a world that I've lived in in, you know, a long time ago. Um, but yeah, I'm wondering, like, I'm also maybe really far on the end of like, this is a difficult problem because I'm like steeped in it all day, every day. Um, but I'm wondering if you feel like there is a perception that

this problem is real or not real or where you sort of personally even land on the severity is not the right word that makes us sound very fatalistic, but the size maybe of the issue.

Stephanie J Neill (24:57)
Yeah.

Well, I think, so, you know, we haven't, haven't actually gotten into this debate deeply with people. And I think it's because we're still kind of at the point, like, like,

the human judgment must prevail point. And they're like, yes, there are, don't forget, there are hallucinations. know, like, like we're still at the point where it's like, there is a grain of salt. And it's like, this is helpful. This is good. Look how powerful, look, look how close this is. But it's like, I don't think anyone's looking at it and being like, okay, this, you know, this is production ready. ⁓ Like we've been working on ⁓ like the tax expert bot, taxpert bot.

to help people like basically through we can't give tax advice, we want to help them like just work their way through the tax world, which is really intimidating and awful and there are a lot of common questions that we can give guidance on. ⁓

Although I think we don't call it guidance, but like, you know, we can, we can help people through. but we haven't released it yet. We haven't made it external. And actually our first plan is to, ⁓ release it for our internal support teams and stuff. Because again, it's like we, even though we're, we've spent quite a bit of time training, giving inputs and it is really actually shockingly good.

It's just, it's not, it's the grain of salt, right? So I think since we're kind of there, it's like not the problem yet. I think it'll become a bigger problem when I think ⁓ people are becoming more reliant on like the letter of the law that they're getting from AI. I think that that'll become a different story in my mind.

Cristina Flaschen (26:33)
Yeah, it's interesting. feel like there's been like this like cultural shift over the past couple of months where it was like, a lot of this stuff hit prime time, you know, two years ago ish and everyone was like, it's genius. The AI bot is amazing. It's going to replace all of our jobs. also sucks. And then there was like the actual practice of it for a few years. And there was still that like the novelty of what it was doing. And now like in this moment,

Stephanie J Neill (26:52)
Yeah.

Cristina Flaschen (26:56)
in the year of our Lord 2026 in February, I feel like there is a little bit of like that grain of salt has become a little larger, like it's more coarse grain in a way that.

I actually think it's very healthy because people have tried, you know, they had a year or so of like messing around with it and like kind of pushing the boundaries. And now we're back in a world where like the technology is evolving. It's evolving really fast. And I'm not going to say it's not. And I think it's amazing. We use all this stuff here internally. We have AI in our product, all that. But there is like, like to your point, that's not production ready code. feel like two, you know, 18 months ago.

Stephanie J Neill (27:17)
Yes.

Cristina Flaschen (27:30)
there was less of that conversation. It was like, yeah, prime time, put it in, push it to prod, we're good to go. I don't need to do anything, I'm taking a vacay. And like, actually, again, I think that's like a healthy sort of rebound effect. And it also maybe that rebound effect, I'm observing that and you maybe, it sounds like are observing that because we work in technology, right? So like we're working with people that tried it really early and had the bumps and now we're like, it's normalized a little bit. I do wonder.

Stephanie J Neill (27:49)
But.

Cristina Flaschen (27:56)
how the rest of the universe outside of our weird little tech bubbles that we operate in just by virtue of being software people, I wonder if that same feeling has kind of permeated down to your average user of some of this stuff. And I was at a breakfast a couple of weeks ago and a woman was telling me, she just graduated from college and she was saying that she uses chat GPT for literally everything. She doesn't Google anymore, she uses it constantly.

And she was saying that she could tell that it's making her feel like dumber. Like she's like, I'm not using critical thinking. And even though I know it's wrong, I just keep going to it. And I'm like, that is just like a different, like from what you're saying, like there's a grain and salt. And I'm like, I agree. But I wonder if everybody has to that grain. I don't know. I don't, again, cause my like bubble, my professional bubble is like very large tech, right?

Stephanie J Neill (28:44)
Thank you.

Yeah. Yeah.

Yeah. No, yeah. I guess. I feel for her.

Cristina Flaschen (28:57)
Yeah,

she was very honest about it. She was like, I hate it, but I keep doing it. I was like, all right.

Stephanie J Neill (29:01)
Well,

but maybe she's actually developing judgment of when not to trust it, right? Like maybe that's actually what's happening. So maybe that's kind of a good thing. Maybe she is becoming more and like I guess the scale of impact feels luckily small. Hopefully she can't like, whereas like if we are, you know, in tech, we in tech or like building it into a product, could potentially have like much greater, you know. No, that is interesting. What I will say is using sort of like

Cristina Flaschen (29:20)
Yeah.

Stephanie J Neill (29:26)
like the chat client, like LLM interfaces, like that feels like casual usage, but when you go a little bit deeper with these tools, you very quickly realize like how far they have to come because of the amount of time you spend debugging.

Right, so it's like, think anyone who really has gotten their hands dirty and gone sort of like that double click down, I think quickly gets a sense of like, so we've got time. We've got time in this phase, or there's more that needs to change here. I don't know about, it's hard to say, I mean, for casual or broad adoption, mass adoption of like chat GPT and stuff.

I do like to believe that people like your friend are actually inherently like learning the edges and like, and using their judgment. Because I know, I mean.

Like I was using it once my son had like a really sore neck and he was like screaming and screaming and I was like asking it like, what could this be? know, like doing medical diagnostic of course. And it's like take him to the hospital right away, right away. Like, yeah, yeah, he's had metagenesis. Yeah, did we talk about this already? What was? And it was like, don't move him, don't do this. And I was like.

Cristina Flaschen (30:34)
Spendin' jadus!

No, no, just, you know, I'm thinking what's the worst case scenario? Definitely, metadata.

Stephanie J Neill (30:47)
it's wrong. And so now I'm like, you when I use it, I'm like, okay, like it's an extremist. It does this, it does that. So yeah, but hopefully that's happening in her case and she maybe doesn't know it.

Cristina Flaschen (30:58)
Yeah, yeah,

just think it's like, again, I think about this stuff, like watching the Super Bowl ads and like seeing all the AI stuff in there. like, meanwhile, like, you the next day on Monday, we come in and like, we're over here fighting with some model to like, it to do exactly what we want it to do. And like, the Super Bowl would have you believe that like the ads, like every person is being replaced, we're all going to go like hang out on islands somewhere and everything is going to be this like robot utopia. And just like the difference there and like those commercials were probably began being produced.

Stephanie J Neill (31:12)
Thank

Yeah.

Cristina Flaschen (31:26)
two, three months ago, maybe longer, when like the perception is so different still, right? Like it moves so fast and know, Stripe is obviously like almost the household name. Like I wouldn't say it's a household household name like a Walmart, but you know, it's a big, you were running a company that people know, right? And so the responsibility for you guys to like get it right is so huge. And I'm wondering, and then I think we're coming up on the end of time, but how do you know? How do you, like, when do you feel

Stephanie J Neill (31:39)
Yeah. Yeah.

Yes. It's massive. Yes.

Cristina Flaschen (31:56)
confident in putting this in someone's hands, in a customer's hands, not an internal person.

Stephanie J Neill (32:03)
Well, we're definitely not there yet as I mentioned. ⁓ Yeah, I mean honestly, I don't quite know I mean because yeah, like obviously stripe as a company has to take this stuff super seriously

Cristina Flaschen (32:06)
So we don't know yet.

Stephanie J Neill (32:16)
⁓ you know, whenever you're even close, like even at Twitch, like we took it extremely seriously, because whenever you're messing with someone's money, and this is true for any type of product development, whether you're leveraging AI or not, right? Like, so even like at Twitch, making changes to our subscription, and like how we assign out gift subscriptions to people, like could have material impact on the creators who are making a life on that platform.

Cristina Flaschen (32:27)
Of

Stephanie J Neill (32:41)
are on that service, I should say. And so, yeah, whenever you're messing around with people's money, you take it super, super seriously. And, you know, I think we're, I would say, like, we're largely at the phase of, you know, ⁓ using it for internal purposes that, you know, result in, like, external change.

that but there's like the human judgment must prevail still, right? So we do have like these things called minions going through and like, you know, and ⁓ doing pull reviews and everything PRs and but then we have engineers reviewing, making sure it is it is like good and sound and going to be a net benefit and then then pushing so.

It is helping, I would say, efficiencies significantly ⁓ and actually probably quality as well. But it is not like a runaway train. I think there's going to be serious guardrails to make sure that we're making safe choices for our users.

Cristina Flaschen (33:38)
I think dog fooding slash drinking your own champagne, however people want to say it, is like the best. Like I also, in my own experience, like if you have great folks, especially like engineers and product folks that work for you, like they have a really high bar for it, not just working, but like doing what is right for the user. And I think that that...

when folks are like, yeah, we released a thing. We don't really use it in-house. like, ooh. We dog food everything, especially because we're a dev tool. And you guys are to a great extent as well. Our engineers should be able to be the power users of these things. And if it's frustrating to them, it's going to be super frustrating to our customers. And same thing with RIS stuff. We have been using it internally for way longer than external. And when we see the efficiency gains and we feel like,

Stephanie J Neill (34:10)
Yes.

Yes.

Cristina Flaschen (34:29)
we can put our own faces behind it and our own brand. That's when it goes out. But even then, there's the grain of salt of like, and make sure you look at this before. And again, luckily, our users are engineers. So I think they will look at it. But I do wonder for more consumer-focused type products. I think that threshold is even higher when you don't have folks that come at it naturally with a skeptical eye, like a lot of. ⁓

Stephanie J Neill (34:37)
Yeah.

Yes.

Cristina Flaschen (34:57)
engineers have a habit of doing, which I love. So I think we're coming up on time, but I want to ask one final question. So you have been in product forever. You have a very large team now at a very ⁓ notable company. And I'm curious if you have like one resource or habit that helps you stay grounded and sane in the insanity of like your day-to-day product role.

Stephanie J Neill (34:58)
Yes.

gosh.

I feel like I say it enough that it's a habit. And I said it a couple times early, I guess. But just reminding people, what is the problem we're really trying to solve? Because it's so easy to drift. It's so easy to lose sight and overbuild or underbuild. What is the problem that we're really trying to solve? And who are we trying to solve it for?

And what is the slimmest solution we need? What is the slimmest way we can solve that problem for them? Because getting that solution that can solve the problem for them in their hands as quick as possible is that's what's going to sustain our business. That's what's going to sustain their business, in my case.

So yeah, I think it's just like resetting, often resetting to like, what's the problem we're really trying to solve here? And that helps me stay sane when things can look really, really big.

Cristina Flaschen (36:23)
Yeah.

Yeah. Yeah. I love the, the sort of canonical user story. ⁓ one of the, the sort of templates, which is like as a blank, I should be able to blank or as a blank, can blank. Like that is it. Like let's, and we, my co-founder is, famous for saying like, let's zoom out. Like at like, what are we actually, what are we really trying to do here? Cause to your point, especially if you love the tech too, like you can get like really deep and get really crazy. I'm like, no, no, no, we gotta like, let's back.

Stephanie J Neill (36:32)
Yes. Yes.

Yes. Yes. Yes.

So, thank you.

Cristina Flaschen (36:50)
way, way, out. And I'm sure with some of the AI stuff, you could really get into ideating and go to a totally different place if you don't. Again, zoom back out. As a blank, I need to blank. And then, you know, I just leave and I'm like, you guys will figure it out. No, that's not true. You're like, I'm the boss. I don't care. Like as a blank, should blank. Yeah. Yeah. Come on. Why are we, let's go. Why isn't it done yet? You guys have AI. It should be finished.

Stephanie J Neill (36:51)
Thanks

Yes.

Yeah, you guys are good to go. Just remember, I'll come back and ask you again in two days.

Cristina Flaschen (37:16)
This has been super fun. I'm totally going to hit you up when I come out to the West Coast so we can grab a coffee. ⁓ Where can folks find out more about you? Is there anything you'd like to plug? You guys are doing hiring, anything like that?

Stephanie J Neill (37:21)
I'm

We are hiring, yes. mean, Stripe broadly is hiring, but my team is also hiring. So if you see any listings with tax, keyword tax, that's my team. And so we'd love great PMs to reach out. And then I'm at Stefaniel pretty much everywhere. I've been told that that's horrible to spell. Nobody can spell it. So I'll spell it. S-T-E-P-H-A-N-E-I-L-L. But that's keyword for me everywhere.

Cristina Flaschen (37:56)
I'm sure folks can find you on LinkedIn as well. This has been a really fun conversation. I really appreciate you taking the time for our listeners. If you want to learn more about integrations, product, spicy takes about AI and ML, worksheets, got eBooks, everything about APIs. You can check out our website, pandem.com. We also have more of our podcast episodes with some great folks with diverse backgrounds, including this one.

Stephanie J Neill (38:01)
Thanks

Cristina Flaschen (38:23)
Again, really appreciate you taking the time. It's been super fun. I feel like we could go forever. And I hope you enjoy the rest of your day and I will see you out there.

Sarah Elkins (38:32)
Thanks for listening. If you enjoyed our content, subscribe to our channel and give us a thumbs up. For more content on tech partnerships, integrations, and APIs, check out our articles, eBooks, and other resources in the description or visit Pandium's website.