"In Q1 in 2025, the first approach was like, 'You know what? Let's not measure, let's get adoption'... We trust our engineers. They work hard. So like if they tell me yes, like I think I'm winning like 20% productivity gain... we trust them.”
Forget top-down mandates. How do you foster organic AI adoption on a skeptical, high-performing engineering team? Loic Houssier, Head of Engineering at Superhuman, joins us to share how he did just that. He explains his strategy for overcoming cynicism, which involved leveraging a highly respected internal champion, the Chief Architect, to re-evaluate the tools and prove their potential was no longer just buzz.
Discover his team's biggest, unexpected productivity gains: dramatically faster ramp-up times on new codebases and the rapid creation of valuable internal tools, rather than just raw code generation speed. Loic details Superhuman's pragmatic, high-trust approach to measurement, blending qualitative feedback with simple signals like PR labels. He explains how they fostered this adoption with a bottom-up "AI Guild," empowering engineers of all seniorities without a heavy-handed mandate. He also reveals a stunning real-world example where AI turned a potentially days-long compliance task into a 90-minute win. This episode is a practical playbook for building genuine, bottom-up AI adoption.
Show Notes
- Connect with Matt: Connect with Loïc Houssier
- Learn more about Superhuman: superhuman.com
- Superhuman was acquired by Grammarly: Read about the acquisition
Transcript
(Disclaimer: may contain unintentionally confusing, inaccurate and/or amusing transcription errors)
[00:00:00] Ben Lloyd Pearson: My guest today is Loic, an engineering executive with 20 years of leadership experience at startups, scale-ups, and global enterprises. He currently leads engineering at Superhuman, the most productive email app ever made. He specializes in helping high growth companies scale with speed and discipline, bringing a culture of excellence and rigor to how teams ship, scale, and make decisions.
[00:00:23] Ben Lloyd Pearson: Loic has also held senior engineering roles at Product board first base and DocuSign, where he led global teams through platform expansion and merger and acquisitions. Loic, welcome to the show.
[00:00:35] Loic Houssier: For having me excited.
[00:00:37] Ben Lloyd Pearson: Yeah. Awesome. I, I've heard that you, you have a team that maybe.
[00:00:41] Ben Lloyd Pearson: Skeptical of ai, so on. If you have a team that's skeptical of ai, like what's the first like small scale experiment or or step that you would recommend for a team to sort of like build trust and demonstrate the value of ai, um, without disrupting like core [00:01:00] workflows?
[00:01:00] Loic Houssier: That's a good question. Um, we were in that situation like, uh, beginning of 2025 at Superman. And I think the thing that clicked, uh, with, uh, my team, mostly senior people, long tenure, they have their flows, they know how to work. They look like they're pretty cynical with everything that is like, uh,
[00:01:18] Loic Houssier: a bit, a bit budy
[00:01:19] Ben Lloyd Pearson: Very typical engineers from the sound of
[00:01:21] Loic Houssier: you know what I'm talking about, right? Um, and uh, I think where we've been smart, we've used our most, uh, senior and respect engineer, which is like a chief architect, uh, can go like with stack everywhere, like the typical crazy guy that everyone looks up to. Anyone's one of those cynical and, uh, we discussed together and said, you know what?
[00:01:43] Loic Houssier: Like a deep dive because did a deep dive six months ago and it was not ready, it was not there. It was not good enough for the level of quality that we were looking for, and we move fast. So it was more like something that was slowing people down than anything. And in, uh, it did another try, like, oh my God, [00:02:00] that's different. At least I see the potential, at least with the right guard rails that can work. And he started to use it on a daily basis, uh, being intentional, putting that into his flow. And is like a, I mean, I don't like this term like 10 x engineer, but like it's typically someone that you would qualify like this, uh, if you have to.
[00:02:21] Loic Houssier: Uh, so he slowed down a bit to learn a new flow, to learn like how this is impacted his own ways and everything, and basically did a quick talk internally to say, Hey people, that's crazy. And I think that creating this sense of, oh damn, if Mike. Is looking at it seriously now there's something to it. So that was the first thing, like creating this sense of like, oh damn, it's not like just buzz anymore.
[00:02:48] Loic Houssier: It's not just like coding stories that you hear on Twitter and anything and everything. Like it's, it's real. It's real. Sure it's not perfect and everything, but uh, that was a kick.
[00:02:58] Ben Lloyd Pearson: we've seen, [00:03:00] you know, both here, Dev Interrupted, but LinearB be our, our sponsor company. Like we've seen a very similar effect where we're. Product decisions that we maybe tried to make like a year ago, um, with the given technology, like it, it just didn't, it either would've taken too much effort to build, like something with AI or it just didn't work the way we wanted it to.
[00:03:18] Ben Lloyd Pearson: And then, um, you know, and then stuff we're doing today, it's like it actually takes a fraction of the effort that we expected because the technology is advanced so much that suddenly now things that we didn't think would be possible, like we're suddenly doing, like with relative ease, you know.
[00:03:33] Loic Houssier: Yeah. One thing that is interesting on that topic is like, uh, we try to measure the impact over the time and, uh, between the qualitative and quantitative, uh, I would say metrics. The one thing that everyone is aligned on is that if there's a velocity gain, it's basically on two things.
[00:03:50] Loic Houssier: One is like the time to ramp up on the new stack. You're not in that domain. You need to collaborate with a new team. Uh, you don't know the stack, you don't know the libraries, you don't know the [00:04:00] structure and everything. It's like brand new. Like the, the ramp up now is so easy. You can just like, talk to your code, understand the entry point, but why they do it this way, not that way.
[00:04:09] Loic Houssier: Like all like code, code and everything, like they basically bring you up to speed, uh, super fast. So like this domain expertise that you don't have that take couple of days to be fully set up, fully understand the way the new team is working one hour, and you get like a, a nice picture. So that's the first one that we've seen.
[00:04:26] Loic Houssier: Um, the other piece, uh, you were talking about like the, the time to build becomes like, different based on the stacks and everything. What we've seen is like all the tools, the side tools that, um, you always like, oh damn, if I had like one day I would build that tool that would save me like hours and hours and hours.
[00:04:43] Loic Houssier: It's not production code, but like something that you would do on the side. Maybe it's uh, I some sort like CSV on the queue. You don't have the time to do and everything. And now. 20 minutes and you have something that is working, you never had the time to do. So all those small additions [00:05:00] overall contribute to a great like.
[00:05:03] Ben Lloyd Pearson: Yeah. And, and a and a phrase I'm hearing more and more is like disposable code. So even if, even if that problem is disposable, you only solve it now and then you just throw the code away for the rest of time. It's like it'll only take me two prompts, so why not do it anyways,
[00:05:15] Loic Houssier: Yeah, yeah, yeah. And it's saving like, uh, so much time, uh, globally and you don't care.
[00:05:19] Loic Houssier: I mean, yeah, it's like, uh, kind of like a drawing on the back of the napkin and it's like code on the, on the back of the napkin and just works.
[00:05:25] Ben Lloyd Pearson: Yeah, and I'm, and I'm wondering, because another thing that I hear a lot as I, as I speak to a lot of engineering leaders is there's, there's almost this relationship between internal AI initiatives and building AI products. So if you don't have that sort of like internal built competency with ai, like you don't know how, if you, how.
[00:05:47] Ben Lloyd Pearson: In, in your day-to-day work, like to be more productive yourself, like whether it's writing code or onboarding to a new stack or, or you know, just you, you haven't, like, internalized a practice, you may struggle to actually like, [00:06:00] adopt AI as, uh, a product. Like if you were trying to build AI into like, your product and, and that's something that, you know, superhuman has been doing.
[00:06:08] Ben Lloyd Pearson: A recently, do you feel that that sense as well,
[00:06:11] Loic Houssier: that's an interesting one. Like are there inter interconnected to some extent? Like, uh, especially in the, uh.
[00:06:18] Loic Houssier: That's a good question. I'm, I'm not sure. I'm not sure because the approach is like slightly different as as you said, like, um, on one end it's like a productivity tools and how you improve your flow, uh, which is slightly different that okay, I use an LLM and an open, open AI API or based and like inference, whatever, uh, to do the job that I need for the code.
[00:06:41] Loic Houssier: I, I think you can be very cynical for your own tool because. I would say our people craft mens or crafts, so they love the craft. So they're like very, I would say very specific in like the tools and tweaking like their own workflows compared to like, Hey, I.
[00:06:58] Loic Houssier: [00:07:00] I know like the way to surface, like that type of information I used to, I need to use an LLM.
[00:07:04] Loic Houssier: So I feel it's slightly correlated. Um, and I don't know if one is influencing, uh, the other, I guess seeing the result of an LLM in production for our users getting better and better is an incentive to maybe a reconsider your own flow and how you can leverage some of that in your flow. But the problem is like, it's different tools.
[00:07:25] Loic Houssier: Even if like the, the backbone, it's still an LLM. your flow depends on the way it's you faced, uh, like in the way you work. So I, it's not because, uh, OpenAI or whatever APIs are working well for your use case implemented in your code, uh, that your IDE use that I would say the right way. And uh, yeah, I don't know what you think.
[00:07:46] Ben Lloyd Pearson: Hey, I'm not the one being interviewed. Um, well, so, so you touched briefly on, uh, measuring the impact of ai. So I, I wonder if maybe we can dive into that, uh, a little bit more,
[00:07:59] Loic Houssier: Yeah.
[00:07:59] Ben Lloyd Pearson: [00:08:00] you know, that's something we're hearing a lot.
[00:08:02] Ben Lloyd Pearson: like every engineering leader we talk through, this comes up like, how do we make sure that we're investing in the right tools and that.
[00:08:09] Ben Lloyd Pearson: Those tools are actually having a positive impact on our organization, and then we can prove that that positive impact has happened. So how does Superhuman approach that
[00:08:18] Loic Houssier: that's an interesting one. Uh, in Q1 in 2025, uh, the, the first approach was like, you know what? Let's not measure, let's get adoption. Yeah. Because like there's, like, everything is changing. Uh, everything is different.
[00:08:32] Loic Houssier: Like if you're a mobile engineer working on cot.
[00:08:34] Loic Houssier: Your frontend engineer working on React, the models might be different. Your ideas are different. So the tools are different. Yeah. Uh, so, and the maturity of such tools might also be different. So you cannot expect to have the same level of adoption or level of, um, productivity gain based on the different teams.
[00:08:54] Loic Houssier: So that's the, the first point. But, so we opened the flow. We said, you know what? Free four. You take three subscription to [00:09:00] whatever tools you want and everything we'll pay. And uh, the only thing that we were, uh, asking for is to consolidate, like the use cases where it was working. So we compiled a list of like, Hey, here, not mature, doesn't work.
[00:09:12] Loic Houssier: That, damn, that was cool here promising, maybe we should try again in three months. Yes. Um, so that was like the, the aspect to drive the adoption. Uh, the adoption at that time was mostly measured in terms of how many subscription people have, do they consume their tokens, whatever. Uh, and it was great adoption was there, and then like you, you mentioned like, how do you prove this is working?
[00:09:36] Loic Houssier: Because at some point my CFO was like, Hey. I say you guys like they're spending like quite some money. Like is it useful? Um, so we started to be like, uh, a bit more precise, but it's still juggling between qualitative and quantitative approach. Uh, we trust our engineers. They work hard. So like if they tell me yes, like I I'm, I think I'm winning like 20%, uh, like productivity gain [00:10:00] over like the course of a month,
[00:10:02] Loic Houssier: we trust them.
[00:10:03] Loic Houssier: Very qualitative, uh, but we wanted some quantitative data. Yeah. So, uh, basically we started to flag, uh, our prs, uh, with some specific labels. Yeah. Like, yes, I used ai and in that case no impact. Or like, I spent so much time trying to get, I would say the PR with the help of AI If I had done it myself, it would've been like so much better.
[00:10:26] Loic Houssier: Uh, so we were starting to like categorize the work pretty fast. Uh, we're seeing that 90% of the PRS were AI helpful, uh, qualified, and with another adoption, uh, roughly said around like, uh, 70%. Uh, so 70% of the PR were flagged. Sometimes you just don't need, like, uh, you fix the CSS or fix, like, you don't need ai.
[00:10:47] Loic Houssier: So like it's, I I think it's,
[00:10:49] Ben Lloyd Pearson: Yeah. Yeah.
[00:10:50] Loic Houssier: you don't need to to measure more. So like all of a sudden we had some sort of like a proof that one it's used. And two, it's useful. And again, we are [00:11:00] relying on the trust of our people. Uh, if they say it's useful. And they're pretty like vocal. So if they say, no, no, it's shit.
[00:11:05] Loic Houssier: Pardon my French. Uh, if they say it's not great, uh, we'll trust them as well. So clearly that was like a, a very solid signal. Um, now in terms of productivity. It's multifactorial. So AI is one piece, but uh, there's also like all the change you do in an organization, you structure your team and everything.
[00:11:23] Loic Houssier: And, uh, the one KPI we're looking at and superhuman is the number of PR per engineer per week on average, not at the individual level. So that's why we normalize, uh, per per engineer and we see the trend. We are getting acquired, uh, over the, the summer we see this trend of like the, basically the raw throughput being better while the, I would say not the average, but like the, the major PR size, staying the same. So there's a lot of like, um, things we can say about this, uh, this metric. Yeah. But at least it's one that is the [00:12:00] most, um, directionally right about like the performance of a team in terms of throughput.
[00:12:05] Loic Houssier: We've seen the progress and we do believe that AI was like a
[00:12:10] Loic Houssier: part of that.
[00:12:10] Ben Lloyd Pearson: Yeah. The PR throughput, it's one of those, it, it's a, it's a tricky one because it's, it, it's probably not the best one to set goals around because if you tell people you have ai, make more prs, you'll get more
[00:12:22] Loic Houssier: I'm transparent on it. I'm transparent on it. So like I have some, I'm gonna say I have my OKRs and I say, guys, ideally we want to be there.
[00:12:29] Loic Houssier: Do, I will use that as a performance metric. Like for the performance cycles and everything. No, but just know that this is my way to, I would say show to the board, show to the rest of the exec team how we work better. Because one, I don't want to measure too much, uh, things because then you start like more measuring than working.
[00:12:51] Loic Houssier: So like let's have like one thing that is simple, everyone understands you understand that it is okay. That's rough you said. LA indicator of our throughput, [00:13:00] uh, and our performance. Uh, but at the end of the day, I don't care. You are, I dunno, you're backend engineer and you're working on like removing, like being, like doing like a huge migration from one framework to the other.
[00:13:11] Loic Houssier: You will have like a massive PR and you spend like two weeks on it and everything. I mean, I won't complain. I won't complain. So, uh, it's still a metric. It's still an objective. Yeah. Uh, but like loosely held, uh, so that they understand the.
[00:13:25] Loic Houssier: uh,
[00:13:26] Loic Houssier: what's behind it, which is much more
[00:13:28] Ben Lloyd Pearson: And if, and if you're in a, if you're in a high trust environment, you know, and, and you, you know, that it's, it's, it, it can at least tell you something has changed, you know, within your organization.
[00:13:36] Ben Lloyd Pearson: It, it, and, and it may not be something you track forever, you know, but it, but it could be something that as you adopt a new tool, you can say, we can see that something has changed about our throughput. Understand the, the ramifications of that and try to understand if there's any sort of negative
[00:13:53] Loic Houssier: Exactly. And, and as with every metric, the metric in itself is useless.
[00:13:57] Loic Houssier: It's like the discussion that the metric is [00:14:00] triggering. That would be interesting. So like, if you see something like maybe one team not taking off. As part of this like constant, uh, I would say improvement. Then you can ask questions, you can begin, you can talk with the tech lead and like, what do you think, what's your perspective?
[00:14:14] Loic Houssier: Like, everyone is like beneficiating from that. Why not here? What's, what's the problem? And, and maybe it's contextual. Maybe it's uh oh, because like this month we're working on this freaking project. It was tricky. So a lot of thoughts, small amount of. Makes sense, but you have these discussions, so, uh, and, um, that's the most valuable part of, uh, any metric.
[00:14:37] Ben Lloyd Pearson: here at Dev Interrupted, we really evangelize this idea of like, you know, metrics, metrics lead to action. You know, if, if you're just measuring and you're not doing anything with that information, there's really no point in measuring. So the goal should always be like, have it lead to something that makes improvement.
[00:14:52] Ben Lloyd Pearson: You know? So
[00:14:54] Ben Lloyd Pearson: as
[00:14:54] Ben Lloyd Pearson: long as that's, that's the outcome, like that's what
[00:14:56] Loic Houssier: No. Exactly. And you mentioned something I important like that the culture of [00:15:00] trust. Yeah. Like I was like direct. I said, Hey guys, I need a fucking K, PI. I'm freaking KPI. Um, for my, uh, I would say for my exec team, um, come up like I discuss with my leads. We think this one is like, okay ish, nothing is perfect.
[00:15:16] Loic Houssier: Understand that this is what it is. Basically don't game it. Uh, I don't need to have like amazing results if it's not related to what we do. Okay? So let's be honest, each and everything, and when you trust people, they're like okay we understand we do our best and boom that works. But trust.
[00:15:34] Ben Lloyd Pearson: Yeah. So I wanna talk about some of the, the barriers to adopting AI within an organization. So just beyond technical implementation, you know, 'cause that's, that's something that, that we kind of discussed endlessly on this show. Um, so what do you think are some of the like, effective communication tactics that you've used to like, address particularly like psychological barriers around AI and, and around like.
[00:15:57] Ben Lloyd Pearson: Fear in particular. Like, you know, you [00:16:00] mentioned that like, you know, you had that lead developer that sort of changed the perception around AI within the organization. Um, be beyond that. Like, like what have you done just to help communicate the value of AI within your organization and help like, overcome like the non-technical side of ai.
[00:16:17] Loic Houssier: One of the aspect we've decided very early.
[00:16:22] Loic Houssier: Separate
[00:16:22] Loic Houssier: Separate, the different stacks. We understand things are different based on where you work because different workflows, different like ideas and everything. So just saying like, Hey, don't compare yourself. Don't whatever. Like we won't do that. So we built guild.
[00:16:35] Loic Houssier: Uh, so we have this concept of guild. We have a frontend guild we have a backend guild, blah, blah, blah, uh, and we decided to create like an AI guild. The purpose of the guild was to gather like the best practices of each pod, understanding like how they were, what's working, what's not working, and also like being a evangelist.
[00:16:53] Loic Houssier: Uh, at the same time, uh, it proved to be like, uh, a really good, uh, way to [00:17:00] have like every pod knowing that this guild existed. One person on their part was the, so like everyone was aware of everything that we would do. So for example. I had to modify a bitta where like compliance approval process for new tools because I just wanted like, uh, like you wanted to today, you need to have it today.
[00:17:17] Loic Houssier: Like you don't wanna wait like one week because compliance, because security and everything, we still want to be safe because we manipulate some, I would say pretty critical data, but we wanted this to be like the P zero for the compliance team, so they knew like new tool for AI and productivity used up everything that you do.
[00:17:34] Loic Houssier: You validate. So that those are the, the, the things that were helpful. And the Guild was basically saying, and basically like, uh, sharing this information that, hey, all the roadblocks from a compliance financial standpoint are removed. You want to try something? Try something. Oh, and you know, uh, we say yesterday I was on the, uh, guild meeting.
[00:17:55] Loic Houssier: That team is doing X, Y, and Z. Sounds cool. Oh yeah, we'll try it this week. [00:18:00] And, and I think that works a lot. So, for example, on my side, so I'm A CTO, VP eng, whatever you wanna call it, I totally stepped away. I didn't want it to be like top down like, oh, we need to use ai. I said, Hey, yeah, AI seems cool free for all.
[00:18:15] Loic Houssier: Like my KPI is like throughput. If it's not helpful, don't do it. If it's helpful, please, uh, would say, don't be smart. And I totally delegated the, uh, influence to, um, that guild. Led by this, um, led chief architect that was highly respected. Uh, and I think that, uh, it, it worked a lot. Like I was focusing on what I do best, which is like working with the rest of the organization, making sure that there's like no, like red tape and everything, and they had like the in depth understanding of what's working, what's not working.
[00:18:46] Loic Houssier: While I would've been maybe a bit more like a. From the code, uh, and using some like
[00:18:54] Loic Houssier: tweet yesterday
[00:18:57] Ben Lloyd Pearson: Yeah. Yeah. We, we actually have a similar [00:19:00] concept. We call it the AI Council, which I love because it's almost like a Jedi council, you know, like, and I think what I really love about this approach is it kind of reflects like what's. It's becoming like a traditional approach at this point of like the Dev X team, where it's, it's like a, a broadly focused team that, um, sort of has to overarch the entire organization in many ways.
[00:19:20] Ben Lloyd Pearson: And it's more focused on like broad developer challenges, like tooling and best practices, and. And, uh, training, support, enablement. It's like seeing, seeing the forest, but then also being able to like, when they have to like, dive in and like, evaluate the individual trees to like, help, help like the front end team.
[00:19:40] Ben Lloyd Pearson: Like understand why this model is hallucinating like crazy when they use certain libraries and stuff. 'cause you know, I think this is, this is something that a lot of organizations, like 2025 has been the year that a lot of organizations have really learned. One really powerful lesson, and that is an AI tool.
[00:19:56] Ben Lloyd Pearson: If you just throw an AI tool out there, it's, it's, it's probably gonna [00:20:00] fall flat on its face in many situations. It'll work sometimes, but there'll be a lot of times where it just doesn't have the context that it needs to be successful. And, and a lot of what it, like an AI guild or an AI council can help is identifying those situations where it, it does have the context out of the box, and you can just deploy it and just throw it out there and be like, here it is, developers go use it now.
[00:20:22] Ben Lloyd Pearson: But then also more importantly, finding the situations where it doesn't have all of the context and tools and resources that it needs. And then helping those teams build those resources so that you can then deploy AI in that situation, you know?
[00:20:37] Loic Houssier: Yeah. And um, I would say we've seen it like this, so our AI console equivalent, so for one, it's, it's not a Jedi console in a sense that, uh, we didn't like.
[00:20:47] Loic Houssier: Selected, like the most senior Jedi, like, uh, to, to work on it. But,
[00:20:51] Ben Lloyd Pearson: But, uh, well, what I actually love about that team is it, it's all ranks. Like Yes, exactly. It, it's like we have junior people who really got into AI and were like, please join us. And [00:21:00] like, like you're helping guide
[00:21:02] Ben Lloyd Pearson: the, you know, the most senior people at our company.
[00:21:04] Loic Houssier: I, I think like the, the most senior people comes with that sarcastic curiosity, uh, about things.
[00:21:11] Loic Houssier: So they will try things like to really understand if it's working and not. I would say basically, I would say scream victory, like, uh, as soon as they have like a line of code that is, uh, generated. But uh, more junior people have still this brain plasticity. Yeah. They know the tools are changing every week.
[00:21:28] Loic Houssier: So like they will try the tool every week
[00:21:30] Loic Houssier: Uh, compared to someone that is maybe like, uh, uh, I would say more senior. You try it once, not working, eh, confirmation bias. And you, I mean, you don't look at it for like a month or two. So I think that blend was working, uh, working well. you need people also that are pretty energetic, so some, if those people are too introvert, uh, too much like.
[00:21:53] Loic Houssier: To the rest of the organization and everything might be, uh, might be a challenge. So we have a good blend of people. Uh, I was in that [00:22:00] console and it's, um, it's working well and, um, they're the ones who are facing like, oh, we have this cool, like, uh, config file for cursor and like using our context and everything, and all of a sudden, okay, can we do it?
[00:22:11] Loic Houssier: Like, yeah, it's personal. Like we put it in the report and everyone is using it and everything. So like those practices start to, uh, to rise from this. Uh, and um, yeah. Cool.
[00:22:22] Ben Lloyd Pearson: Yeah. And, and I, and I think one more point on the, you know, the levels of seniority like junior, you know, whether it's developers or any role within the company, junior level people versus senior level people adopt AI in very different ways from, from what we've seen, like a, a junior level developer, generally speaking, they often are using it for more things, more like code generation or.
[00:22:44] Ben Lloyd Pearson: Maybe learning onboarding to a new tech stack, whereas a senior developer, um, you know, they're typically, they, they already know how to write code really efficiently. So they may maybe don't use AI for that as much, but they might use it for ideating in the early planning [00:23:00] stages. You know, so it's, it's really important to get all of those perspectives on.
[00:23:04] Ben Lloyd Pearson: Yeah. On your AI council or
[00:23:06] Loic Houssier: No, totally, totally. And, and you're right, like the, the ideation part is probably like the, the way like, um, most people are using, like most senior people are using it like, Hey, I've seen, I see this. That's the problem I'm trying to fix.
[00:23:19] Loic Houssier: My strategy is to do X, Y, and Z. What do you think? Challenge that. And they use basically tools like this to make sure like, oh, okay, go. This is thinking out of the box this time. That's interesting. I might dig in and uh, and gives you also like more confidence that you're on the right track.
[00:23:36] Ben Lloyd Pearson: Yeah. So I, I just got one more question for you. So do you have any, like, specific examples of like workflows or daily tasks within the superhuman engineering team that were, were you seeing like the introduction of AI lead to like, some sort of non-obvious improvement to either speed or quality? Um,
[00:23:52] Loic Houssier: Um, I can mention one very personal, like we were, uh, we were acquired, uh, over the summer by, by [00:24:00] Grammarly.
[00:24:01] Ben Lloyd Pearson: Yeah.
[00:24:01] Loic Houssier: any due deal there's a lot of like computation of data that you need to do.
[00:24:05] Ben Lloyd Pearson: Yeah. Which, by the way, I wanna mention, it's like my two favorite tools in the world, like coming
[00:24:09] Loic Houssier: Lovely, lovely. So hopefully we get, you'll get the best of both world. Um, and during these due diligence, you need to compute like a bunch of data, like you crunch.
[00:24:18] Loic Houssier: And in one case, like it was like at least all the open source libraries that you're using, tell me like all the license there are MIT, uh, G, PL, whatever. tell me who's the, um, licenser. I was like, what the fuck is the license store? Sorry for your audience? And, and, and then like, oh, that's the name that is officially on the license and everything that holy cow.
[00:24:39] Loic Houssier: I would go like into like all the GitHub repo and find
[00:24:42] Ben Lloyd Pearson: Yeah. I said
[00:24:43] Loic Houssier: hey Claude code. I have this small problem, I have this CSV with all the list of the open source libraries. We already tracked like the, uh, the GitHub repos of, uh, of those like find, like, create like a script that we find the readme or the the, or the, like the license, [00:25:00] uh, that .txt or whatever like file that is like containing the, the license In that document, find the sensor. Oh my God, 90% of it was matched.
[00:25:09] Ben Lloyd Pearson: Wow.
[00:25:10] Loic Houssier: Something that would've, like, I would've bru force it. I mean, there's no chance, like your due diligence, you don't have the time, whatever. Like, and, and you cannot just involve all your team, like to distribute the work and everything. So like, and it took me like, uh, I guess like, uh, one hour and a half like coding through it.
[00:25:26] Loic Houssier: Yeah. And um, so that's it. That was a good example. Something that where I would've spent like two days and nights, uh, during the due diligence, as you said, disposable code.
[00:25:35] Ben Lloyd Pearson: Yeah. Wow.
[00:25:37] Loic Houssier: even know where this, I would say script is now. And, uh, and it's done. So, so that's a good example.
[00:25:42] Ben Lloyd Pearson: Believe it or not, I used to work in open source compliance, so thi this use case, like, I, I would wanna rip my eyes out rather than trying to do that manually.
[00:25:51] Ben Lloyd Pearson: So having AI for something like that is pretty incredible.
[00:25:54] Loic Houssier: Yeah. Because like all the tools that are doing those, um, I would say SaaS and like to understand like the, the open source libraries [00:26:00] that you're using and everything, they don't go to the, who is the licenser?
[00:26:03] Ben Lloyd Pearson: Right.
[00:26:04] Loic Houssier: like, you have this nice fight and everything, and you believe that you're covered for your due deal, and all of a sudden you have
[00:26:09] Ben Lloyd Pearson: I imagine that data has gotta be so messy. Like there's no way that data is clean.
[00:26:13] Loic Houssier: coming. It's, it's, it's not structured. It's not, uh, so it was fun. It was fun, but it was a good usage of ai.
[00:26:19] Ben Lloyd Pearson: Yeah. Well, awesome. Well, it, it's been really great having you on our show. Loic thank you so much
[00:26:24] Loic Houssier: Thanks. All the good time.
[00:26:25] Ben Lloyd Pearson: to all of our listeners, thank you for tuning in today. If you're not subscribed to our substack, head over to Dev Interrupted dot substack.com.
[00:26:31] Ben Lloyd Pearson: Click, click the subscribe button, give us a review on wherever you listen to us for podcast, and tune us next week. We'll see you then.
