Pascal Weinberger: An Over-The-Shoulder AI Companion

Pascal: Hundreds of years ago, most of us were working in agriculture. Now agriculture makes up a single percentage, and we do all these other cool things like, doing podcasts, building tech companies. It's not that we're like sitting around doing nothing.

Connor: Hello and welcome to go to market with AI, a podcast for sales, marketing, and customer success leaders using AI to scale their growth operations. I'm your host, Connor Jeffers. And what follows is a conversation with Pascal Weinberger, the CEO and Founder of bardeen.ai. 

Pascal, welcome. Thank you so much for joining us. We recently met because you guys recently got an investment from HubSpot ventures. And so, we share an investor and we are similar HubSpot portfolio companies and we'll get into all things HubSpot and all things Bardeen. But before we get there, I wanted to start with hello and welcome and also I don't know you super well, and I want to know your background, and how you founded an AI company, and all of that other cool stuff which I'm excited to jump into.

Pascal: Cool. Yeah. Thanks so much, Connor for having me on the show.

Connor: Where are you joining from as a starting point? Is it, are you night time?

Pascal: Evening time in Switzerland. Yeah. The sun is down already. We actually finally got snow today, which is great. Cause it's like snowless winter, which is sad. But yeah, no calling him from Switzerland and yeah, thanks so much to you and the team and everyone for making this happen. I'm super excited about being on the show.

Connor: I am extremely excited, and I, all I want to do is start digging into Bardeen stuff and why and everything else, but maybe as a starting point, what is your background? And I think whether we don't necessarily need to go play by play, but I think that maybe from the journey of what is your experience of background?  How did you find yourself in a position to say, " I'm going to found an AI company and we're going to build cool AI stuff". Which we'll spend a whole bunch of time on.

Pascal: Yeah, sure. So my background is machine learning. Originally I started with computer vision. And then lastly, I ran the AI team for telephonic cars, like the, I think still number two telco in the world by number of users. And,

Connor: Are you an engineer? Is that, are you an engineer by background?

Pascal: Yeah. I mean, I'm not the best engineer. Like, we have much, I'm like a script kiddie for the AI world, I would say…

Connor: They have AI script kiddies? That's a thing?

Pascal: I mean, with Chat GPT, it's way more of a thing.

Connor: We're all AI script kiddies.

Pascal: But yeah, no, I mean, don't spend as much time writing code as I would like to anymore, but I still do every now and then. And yeah, the first version of Bardeen especially like the machine learning parts was largely me and my co founder writing it.

So, yeah, I mean, how did Bardeen start? So at this Telefonica gig, basically we scaled a really cool team. It was a really cool mission. The idea was to like, use this telco data for like things that aren't like your typical telecommunication business, right? Trend prediction and all this boring stuff that's critical to the business, but not the most exciting. So we were looking at healthcare, electric, like energy distribution, city planning, like all sorts of different cool use cases and the job description was like amazing, you get all this data You get lots of resources amazing team that we got to build and yet I found myself spending a lot of time doing very stupid like repetitive workflows like, whether that was in recruiting for like copy pasting data from LinkedIn profiles to our recruiting system, writing outreach emails to those people then forwarding it to the recruiting people we were working with or like even project management.

We had like bug reports coming in on email, had to be logged in Jira, like sent to the engineering Slack channel, all this type of stuff. And I'm sure anyone out there, and I know, like we talked about this before there's probably like a million different things each of us does. And that, given week that we're all every now and then we're like, Oh why do I have to do this?

Connor: I mean, I think that like what you're describing for most people is just like, work. Like that's just like what work is all of the things that you're like, man, none of this is writing cool AI code. It's doing all this other bullshit that I have to deal with.

Pascal: Yeah, but you know, it's sad that's the reality. I think like the, and then like we'll get into what happened in a second, but like the, from a philosophical point of view, it's kind of, we've gone through this like bundling and unbundling of tech in the last, 20 or so years.

And recently there's this big unbundling where now you have for every single thing that you could possibly do. There's a SaaS app for that, right? And I mean, I'm sure you've talked to a bunch of people on this podcast already that are like doing an app for X specific use case. And that's amazing because they're all...

Connor: It starts like, I'm going to hyper narrow and then I'm going to platform and then I'll be like, oh wait, no, we need to go back down.

Pascal: Yeah. And nothing against it. That's amazing because there's super specialized apps that are like really good at doing this one certain thing that you might need to do. But it kind of puts us as the users in the situation where we have 50 tabs open. To get anything done and we almost become the router for our work apps, where we end up copy pasting data between different apps from your LinkedIn into a HubSpot to email to calendar, you name it.

Connor: I love this idea of like, the human tech worker is just like the OG API. That's like your old, actual job is you just move information between software systems.

Pascal: And I mean, it's crazy. Cause so I mean, it's crazy, but not that crazy, right? Because we've had this problem before with search, for example. So like search, pre-search engines, you would typically go on a specific website, wait for the website to load, click the like sixth sub menu where you know the data you're looking for is, wait for it to load, and then get or read that information that you're looking for.

Then search engines came along. Now you're basically just going to your favorite search engine, whatever that is. And you just ask the search engine to go find it for you. And it then does all this like delegation almost of the search for you and finds it. Now with Chat GPT and like other chatbots, gen AI technology, it kind of like changes the game again, where you you don't even go to the search engine and you just ask the bot to what's the thing, and it goes and does it for you.

And I think a similar shift has to happen with, this like deep web of all the logged in services and SaaS apps that we're using that I, right now, I'm still in this pre-search engine era where I have to go into, not to pick on HubSpot, but like I go into HubSpot, I wait for it to load,

I go into create new lead, like fill in the data in HubSpot. And it's an amazing system for doing that. But I just don't love this idea of having to do all these like steps manually across all different services. So what we're trying to solve for with Bardeen really, is try to like, build a search engine, but you think about it almost as a do engine in the sense that it like goes and does the work for you across those different services.

And like another way to think about it is like Excel macros across your workspace. So, in Excel, people are used to this idea of macros since a long time. You just take some keyboard shortcut or you write some function and it goes and does things that cross cells in your Excel spreadsheet.

And like with Bardeen what we're trying to do is a similar kind of idea across your workspaces. So you're on someone's, conference profile. I'm, I'm seeing Connor on the conference profile. And I'm like, Connor is a cool guy. I want to, yeah, I want to like at the GTM AI conference. And I like want to reach out to Connor.

Now I put your data into my CRM system, my HubSpot, copy paste the data in there. Now I want to write a customized outreach email that uses the data from your profile. You may be a speaker at a conference, so it says something about you. And then I want to say " Oh, Connor would be amazing to chat to you about the work you're doing with hapily and the GTM podcast.

And we have this common connection with HubSpot, let's chat". And I would write that outreach email for you. So like traditionally that takes me like 5-10 minutes to do manually, I do this. Before I go to a conference, I might want to do this with hundreds of people. So it becomes like crazy.

And with Bardeen, what we do is like you're on that conference profile and you just click a button and then we connect all the API services, the AI integrations with, GPT to write that outreach email and all that like fancy stuff. And all does that like for you with the click of the button.

And that's really kind of like the future of work as we imagine is really kind of taking all this heavy burden away from us as users and delegating it to what machines are really good at is repetitive workflow. So we can focus on this, like creative decision making of deciding who I actually want to reach out to and then leaving this busy work, kind of the long tail of busy work to Bardeen in that case to, to do that.

So that's kind of in a...

Connor: The origination of Bardeen like most great software is someone who's really smart and very lazy and says, "I don't want to do all of this anymore. How can I not do it anymore? I know what I'll do, I'll start an AI company and I'll raise money. I'll hire all these people and then I won't have to do these things anymore."

Pascal: And yeah, I mean, I think it's the very first one is actually like a really funny story. I don't want to spend too much time on it.

Connor: No, absolutely.

Pascal: Both my Co-Founder and I, we both pre-founding the company, I built this prototype called me.ai which was actually a Chrome extension that would get data from GitHub profiles.

I was recruiting a lot of engineers and was basically like getting email addresses and data from a GitHub profile, write an outreach email, put it in my Gmail draft folder. And it was using like another API, iPass platform in the backend to do that. But like, it solved for this like context problem that no one had solved.

Until Bardeen, is kind of like you just click the browser extension would do that for you and my co-founder built a thing called mini me, which was basically like the slack bot that would follow up with his engineering team who was a, VP of engineering or the big cloud computing company.

And he had like many direct reports and they just needed to follow up on like certain Jira tickets that weren't closed before, just like he would just create a Slack bot to do that. And then like, when we started talking about the idea of, but you were...

Connor: Can we just laugh for a second as this is the most German engineering problem to “ I need to manage all of these people. And it's very challenging. And what if I get, I'll build this AI and it'll just follow up with them on their Jira tickets. And then I don't even need to manage them anymore."

It's incredible.

Pascal: Yeah, you're trying to automate. I think it's so this is kind of a philosophy, my co founder and I both subscribed. was like, you want to try to automate yourself to grow, right? If you, figure out how to do something. And like a lot of people, I'm sure on this podcast have said this before it's not nothing new, but if you've tried to figure out something, like how to do something, you figure out how to do it well, then you want to either delegate or automate as soon as you can, right? Because, so you can focus on like the new things to actually grow. And traditionally you would maybe hire an EA or do this with like that delegation part of it, right? And I think now there's this like technologies there, like we have all everything's built around SaaS APIs.

Everything's built around like this API economy. Everything is a, very nicely defined like SaaS app in your browser with more or less good API documentation. And you have all this new technology or AI, generative AI that allows you to make it like super easy for people to access.

So like today, I think a lot of the stuff you shouldn't delegate, you should just like straight up automate. And that's kind of what we're trying to solve for with Bardeen. And then the problem becomes now, when you talk to almost anyone out there, they will all acknowledge that they have stuff to automate.

Yet, you look at their workflows and they haven't automated a lot. It's reserved for, engineers, big company people who they have the resources, they have an automation center of excellence or something. Or like a big company hires engineers to write those automations, but like the rest of us, like small companies, founders, entrepreneurs, or even people in big companies that just don't have the resources, access to the resources.

Connor: Big companies take forever to buy the platform, pick the thing. Yeah.

Pascal: Yeah. So like we're left with doing it. so then the problem becomes, okay, now that we acknowledge we need to automate stuff, how do you make it accessible for people to do that? And then it's like, okay, like when we started by Bardeen, the first version we launched two, two and a half years ago was like a no code platform.

And we've...

Connor: Is that, is the two, two and a half years ago, is that like the founding of the company? Like you're two and a half years in, or is there some before that?

Pascal: Almost four years. Yeah, we're almost four years. And then there was a lot of just figuring out the foundations of it. For anyone who's ever dealt with APIs in their life it's there's a lot of kind of nitty gritty, pipe building, as you would say that you need to do before you can actually do the user experience part on top of it.

So we spent some time doing that. We figured out all the engine, compiler engine, how do you make this work in the browser? There's a lot of complications that come with kind of the platform that we chose. There's building in the browser, in the context of the user, not another web app.

We didn't want to add another tab to your browser you have to open. Automate stuff. We want to bring it right into where you already are, so we chose a browser extension as the kind of medium to do that, which becomes very valuable because most of our time today is spent in the browser across those assets we just talked about.

And then building a browser extension bit almost like tabs into the operating system of the browser and allows us to do a lot more interesting things than you could do with a web app. But anyway, so did that, build the first version was a no code platform made it already a lot easier for people to build stuff that don't necessarily have engineering skills, but you still have to think like an engineer, like if you're writing no code platforms stuff, right?

Because you still have to understand what is it that I'm trying to automate. Then like dissected step by step, like first to the second, do those, use the input of that, like use the output of that as input to the other thing, that kind of thinking still makes it pretty hard for a lot of people to do. So what we did last year, in this whole Gen AI world, is we were the first ones to launch this language to automation concept where, and it's live in the product now since almost a year, and we've had many people using this successfully.

And the idea is basically just describe in natural language, what you want to do the same way you would describe it to your friend or your assistant or your coworker, or whoever you would traditionally delegate it to, just describe it in natural language. And we will then like with a machine learning model, try to build the automation for you.

So you don't have to start from scratch with a blank sheet of paper and a no code builder, but you have the skeleton at least sometimes the full automation build out. And then that makes it much, much more accessible and that's the whole angle, right? It's build an automation platform that can solve this in context, proactive automation problem, make it accessible and easy to use for the rest of us outside of engineering people and so on.

And then the last step of what we're working on is make it proactive. And like what I mean with proactive is really this idea of. What if I don't even realize that what I'm doing is  ? And I think Grammarly is a great example to this, right? Like Grammarly...

Connor: AI looking over your shoulder and being like, "Hey, by the way..."

Pascal: Yeah, but like in a non creepy way, I think if you put it like that, it almost sounds creepy, but like...

Connor: Sounds amazing! Like an AI, AI just saw you be like, "Hey you don't have to do that, man, I like, I got it. Don't worry about it."

Pascal: Yeah. So I think Grammarly for example, is an amazing product from that perspective where I install Grammarly, I buy their plan. I forget that I even have it installed and now I'm in my Gmail editor and I like write an email and I make some mistake. So, as we all do, I'm a German speaker writing English, so I make a lot of spelling mistakes in English and Grammarly just jumps in and like proactively corrects them for me.

I don't even have to think about Oh, I have to copy paste this text into Grammarly, get it corrected. It just proactively does it for me. And that kind of experience is what we want to get to with automation.

Connor: Actually. I think that's really amazing. 'cause I think that the, yeah I, I think it makes so much sense where every other business application does feel exactly if you just still Grammarly down to, you can have an exact Grammarly experience. You copy it, you paste it somewhere else, it'll correct it for you.

And then you grab that and bring that back to the email. But that's how literally everything works right now. That is how every single point solution of software...

Pascal: Yeah. So you're the router again.

Connor: Exactly. Yeah, 100%. I I love that idea of it all runs back through the router. And if you can eliminate the router itself, because you are the router, you decrease human bandwidth and therefore dramatically increase human productivity.

Pascal: Yeah. That's yeah. That's much shorter than I could say, but yes.

Connor: Solve the world's problems. We're good to go. What in terms of current state what are the things either you're seeing if there's a, if there's a particular customer that you were like, Whoa, they did something incredible, or if there's something incredible that you guys have done on top of Bardeen attached to, kind of a GTM function what is that?

Or what have you seen that either impressed you or your teams put together that you think really shows the power and capability, not just at Bardeen, but sort of, of taking this approach in general?

Pascal: Yeah, great question. I think there's a, I'll give you like three quick examples. One is on a GTM side, like we have handful of people use, like a lot of people using this in a sales motion, right? But it's and this is also where the HubSpot partnership comes from, right? Like it was basically like, you're on some sort of profile, right? In my case, it was GitHub. Some people do conference profiles, LinkedIn, like whatever the, new status or scrunch base, like wherever you see your leads, now you kind of manually take this lead data from the platform. You copy paste it into your CRM system usually and then from there, you may like there's a kind of a long tail of other steps that you may do. Like you write a custom outreach email, you create a Calendly link and attach it in there, or you forward it to your sales team and slack for them to take a look at, or you add it to some Google sheet or other spreadsheet to keep track of it. Are you, whatever, the status that you're doing, but basically this idea of like lead generation, like getting data from one place and then logging it into some sort of system of record and doing some actions around it. And that's really like what Bardeen lends itself really well to, examples we have is, for our own recruiting, we use it a lot.

Connor: So, but my setup of that workflow and Bardeen is what? So, I log in, I create an account, I connect it to my email? What are the steps that I do in order to make what you just described feasible?

Pascal: Yeah. So there's a few different angles to get there, right? Like one is, and best route to get this, we shipped a product with 1, 200 prebuilt automations at this point, a lot of them fitting this pattern that I just talked about. And you would just use one of the prebuilt existing automations.

You would then have to authenticate, like you create your Bardeen account, you download the Chrome extension, you log in, you select automation that you want to use. Let's just say GitHub profile to HubSpot, for example, and then invite an outreach email. And then you would just click that and you would now need to authenticate the services.

So you would authenticate with your HubSpot and your email and whatever other services are part of the and then you're good to go. Then you basically have the Chrome extension installed. And now when you're on a profile, say the GitHub profile, you want to click a button, like you just click the button in Bardeen.

And it then like kind of captures data from the current screen that you're on...

Connor: Is your view, and this comes back to I think the thing that I think of when I try to understand where you guys fit in versus like a a Zapier or something in that mix, which is, you are, because it's a Chrome extension and because you're looking at this as these are user initiated workflows versus system initiated?

Is that is that a good distinction?

Pascal: like We think about it as like proactive versus reactive. So like what you would traditionally automate with iPass like Zapier, trigger.

Connor: Something has to happen somewhere in order for everything downstream to exist.

Pascal: Trigger based. Exactly. So it's reactive. Like I get an email, I want to save the attachment in Google Drive or whatever. So it's like purely reactive.

It's great, but like most of what we do is not that, like that's the trivially automatable stuff. Most of what we do is there's some decision making involved, right? Like I don't want to reach out to everyone on LinkedIn. I want to reach out to Connor and whoever works at HubSpot and maybe like some other people.

But I make that decision, like I'm in the driving seat. And then, that becomes really hard to do with this trigger based like backend system automation platforms that you traditionally see, that's where you have to bring it into the browser, give it access to the context and so on. Having said that we also support trigger based stuff.

So like long term, the idea is you're not going to want to define your workflows in many different platforms. You're going to define your workflows. Like we call them playbooks. You're going to define a playbook once. And then you may want to trigger it manually, like what we just talked about, or you may have an automatic trigger that says every time I get someone who signs up for my marketing form on my email on my website, go find a LinkedIn profile and then run this playbook.

And that, that kind of stuff we also support. But really the angle and like the, how do you say, the spearhead that we chose to go to market is really like this proactive context.

Connor: What type of, who's the user that you guys find comes into Bardeen? And you guys are all, I wouldn't want to use the word all, you're primarily PLG where it's people coming to the site. They're sending, okay. What type of user are you guys finding are the ones that find the product, install the product and start doing stuff with it?

What is the profile of that user who, because I'm very rambling on this question. But I think that the thing that I find the most interesting about all of this AI technology is, everyone's talking about it and I think that people find it very difficult to get started and use. And so I'm curious because you have this insight in terms of this PLG motion of what types of people and organizations are or individuals are finding this thing and saying, oh, wow, I can solve a problem with this. I'm going to start using it?

Pascal: Yeah. So I think there's still some like early adopter bias that we see we traditionally see mostly young people who are either in startups, agencies, small, medium businesses. And they have this like productivity drive, like they're early adopters of many other tools...

Connor: Do you think that is it they have a productivity drive and maybe autonomy and like they can do something without it being like, lockdown, or what's the...

Pascal: Yeah, some sort of autonomy. So it's mostly people who have early adopters and they have the autonomy from their organization or by themselves to use it or to give it a try. Then there's a PLG motion there, right?

Where you now want to share the automation with your colleagues and your team. You may be working in a sales...

Connor: "Hey, I made this, and it's awesome, and let me..." Yeah, yeah, yeah, yeah.

Pascal: Yeah. Or I see "Hey Connor, how are you so productive? Like, how did you just do a hundred leads in a day? Show me how you did that". Or your manager coming " Hey Connor, great job doing this hundred leads today. Like, how did you do that?"

And then we see some pull emotion through that. Primarily growth driver for us today is just like SEO. So people search for, they have a very clear search intent. They're like, I want to automate. How do I get data from, inbound HubSpot conference into my HubSpot CRM system?

Connor: How are you guys managing the, how many people do you have creating that? Right? Cause it's a lot of content for you guys to crank out and then how are you managing and figuring out what you're making for, if that's like the primary driver, I assume that's something you're spending a fair amount of thought process on.

Pascal: Yeah. So we're super small team on the go to market side. We're like basically mostly engineering. We've invested a lot into like programmatic SEO. So basically it's AI to some extent generated or like just programmatically generated landing pages for all of those mentioned before 1200 plus prebuilt automations.

So, each one of them will have a landing page that there's some, depending on how much traffic we get on the keywords on the pages and so on, we invest more or less into the actual copy and content of the page, but then, like every time we build a new automation and we built like many of them per week, right?

And the way we come up with them is I would say 80 percent of the, of it at this point is like user driven. So like people in support or in the community or customers ask us. " Hey, like I want to use Bardeen for X" and then we'll help them build it, but we'll also just publish the automation of it, unless it's like a proprietary system or something that we can't just openly share.

But if it's something that's, useful to more people than just say, Connor asking for it, like we'll publish this automation. And then as we publish it, usually we start out with pretty thin content. And then as we start seeing the site rank and get like traffic on it, we'll invest more and more into making it like a better landing page experience. And then that's kind of how organically we build it out. And then there's...

Connor: For anyone listening, it's honestly very funny to me because years ago pre all AI stuff, I worked in an organization where we're doing a lot of outbound sales. We had a marketing leader who was like, very much we need to build all this inbound content. We have all these things, all these people want to do let's just put someone on writing pages and writing blogs for everything that these people look for.

And our CEO at the time was like, that's a total waste of time. It's going to take all this time. Let's just send more emails outbound to everybody. And it's just very funny to me that the most tried and true method of what do your customers want to know. Answer it.

Pascal: Yeah, publicly.

Connor: Will work all the time.

Pascal: Yeah. I mean, I think it's a, I'm a big fan of this first principles thinking approach, right? And it's if you're trying to build a product that you're fundamentally trying to scale through a product, let growth motion, but the alternative it's a different thing if you're doing outbound sales, because you have someone talking to your customer.

And then it's that sales person's job to figure out like in, in a half an hour conversation what does Connor want and how can I match my product experience to what you want? But if you don't do that, okay, you have a product that growth experience or an SEO growth experience, then you kind of have to boot force that you have to almost be like, okay, what are all the possible things Connor would want?

Let's make sure that anything he types into search engine we somewhere show up, right? And then it's also just like SEO is this great kind of like long term compounding growth method where we started doing it like three years ago and in the very beginning we had very little traffic through stuff.

And now like the more pages you get indexed and more like, valuable your content becomes and then the more valuable content you have to hire every new piece of content ranks and it just kind of becomes a self fulfilling prophecy almost, where as long as you put out good valuable content in the niche that you're trying to solve for, it's just like compounding.

So it's a very long term but we think like ultimately winning strategy for us. Now, there's this big question with what if search engine is no longer the primary interface in which people search. Hey, do you stretch up to other things? Is that still valuable or not?

Connor: I don't buy, I don't know, I think saying that ChatGPT or Chat Interface is the search killer. I think for certain things, yes. Like recipes, like no one wants to read the blog about the story of how you met your husband in France to make brussel sprouts. Like no one cares, but I think that's where you go to an AI situation and you're like, I did this.

I'm not even being an obnoxious AI podcast host. Like I did this week of looked in the fridge, saw what I had, went to chat GPT and I was like, what can I do with this? And it gave me a recipe. I was like, this is great. And I, that experience I think is something that is, I am not looking for a specific thing.

And I think the, how should I go and automate this and I'm looking at multiple different approaches and I actually think to your point you want to become I think this ties into the thing I actually wanted to close on here, which is like I find the idea of this over the shoulder AI companion extremely compelling.

And I think to your point, that's how you jump over the search gap because the user never actually looks for the solution to that problem. You are observing that user and making suggestions to that user all of the time, and if you are doing things in a reactive model, that's not really possible because there's no behavior to observe.

Pascal: Yeah. I mean, I think like one sentence on the search killer question, I think the jury is still out. Like, I honestly find myself using, not necessarily Chat GPT, but like other services like you. com, perplexity, other services like that, like a lot more like recently to the extent that...

Connor: It's bad for Google's ad revenue, to be clear. I do think that it is net bad for Google's ad revenue. Absolutely.

Pascal: I mean, I think look like, and this is kind of actually ties into what we just talked about with this action automation question, right? I think, does like blue links as an output of search isn't, is this, it's fundamentally a suboptimal solution. If I'm asking a question, I want an answer.

I don't want like an option space of different resources to read. It's kind of like it's solving part of the problem, but I still, it's still kind of...

Connor: You still have to go and filter through it. And I think that the difference is like there's certain questions, I actually think that was really an articulate way of condensing what I was kind of meandering to with my recipe concept, but it's almost the idea of discrete solutions versus like discovery based solutions and anything where you're like, there is a discrete right answer. Search sucks. And if I want an exploratory experience.

Pascal: Then it's different.

Connor: I don't think AI is very good, but I actually would call into question like search kind of also still sucks and it's optimizing. Yeah,

Pascal: I think on that subject, the jury is out a lot more. Having said that, I think a lot of the value in search these days is from answering concrete questions, and I think that they will have to reinvent fundamentally now, having said that Google obviously has a good place to start from, but I guess also not the topic of conversation here, but I think yeah, the jury is still out.

I think the interesting analogy to the kind of automation space, is that it's a similar problem space where there are existing solutions in the automation space. You mentioned some before that kind of sort of get you some way there. If you know exactly what you need to do, you navigate to a certain new website,

you like go through the builder experience, you kind of go through all these hoops, basically. Then they do a decent job at automating what you're ultimately trying to automate. But if I just have one specific thing I want to do, like why not give me like a text box, I can just describe what I'm trying to do.

It goes, figures out how to actually do this for me. And like I click a button, I'm done with it. That's ultimately kind of what I think search should be and automation should also be. And that's kind of our approach with magic box.

Connor: I almost wonder to your point, like how much of the I would say in, as a manager and the manager of managers, right? I think that a very normal management workflow is you ask someone to do a thing. You, they go and do stuff. You come back and you're like, "Hey, where's that thing?" And they're like, "Oh yeah, I don't have it."

And you're like, "wait, what? Why? Like why don't we have this yet?" And then you're like, "how are you doing this?" Then they show you how they're doing this. You're like no. Let's not do it that way. Let's do it this other way. And this should be really easy and we can make it happen.

And. I think that the magic box concept, I wonder if it also begets itself to where you subvert that entire cycle by you start with, I want to do this thing. You go to AI and then AI can maybe even come back and say, Hey, we can do some of this, but you're going to have to do these parts. And now you end up having a management.

Like you don't need, do you not need a manager or do you not need a worker in that dichotomy? I don't know which one of those two things you end up removing.

Pascal: I think so this goes into a really deep debate, like philosophical debate of will AI ultimately replace knowledge workers? And so on, right? I think our, or yeah, kind of our philosophy with this is we want to empower people to be more productive and better, right? So and then this like divide between, what you just called workers and managers, I think over time will shrink.

Like the difference between a manager today is that they like, they have the methodology down usually. And then they like delegate the execution because it's more efficient for the business to do it. And then the work I usually, in your example they may be like really good at just like execution, which I would argue long term yes, some of this should be delegated to machines because I was better at it, but then now you have more productivity kind of, unlocked and one individual person can do maybe the work of 2, 3 4 5 and number of people. Now in reality, what's going to happen is like businesses aren't gonna fire staff and like the people, it's  primarily a growth driver, right? And I think that's a very important point, especially in how today people use Bardeen.

We see this as a growth unlock and does our whole sales pitch is like a growth unlock. So if your sales team and you or your sales. Team could use Bardeen to just be 10x more productive that would give you a competitive advantage over your competition. You now would not let people go, you would use that as an advantage and increase your top line and win the market.

So I think that's, I think ultimately the power of AI. To do that, you have to make it like usable for the end user, the risk that other approaches have that are much more manager driven approaches that are like top down sales motions. Go to the CIO, CTO, CFO department sometimes and go Hey, we can save your staff costs, like kind of the traditional RPA approach that you see happening a lot.

There's so much resistance there from the actual people who should adopt it precisely because they're scared to remove themselves from the equation, which I, I understand it's relatable. But you can turn this whole thing around if you empower them to automate their own workflows. No one wants to do repetitive, automatables, stupid work, frankly.

Like everyone wants to be more creative and do more interesting things with their time and be more productive. And you can become the leader at your company, like almost become like, the automation superhero and unlock that productivity for people. And I think that's a really important aspect, which is why I think, when we talk about go to market motions for this, which is why I think, although more challenging, just like product let growth approach for automation, ultimately long term can win over top down approach, precisely because you empower the end user to do stuff.

Connor: I had this whole other thing that I wanted to say, but I thought that was so beautiful that I want to close on that note and I don't want to take it over as a closeout because I thought it was so articulate and eloquent. And I think it, it really synopsizes what I think everyone looks at this.

There's a finite amount of work to be done and we're just divvying up chips versus there's an infinite amount of work to be done and the capacity for human beings is not to your point, the function of, am I a worker? Am I a manager? But like people take roles out of necessity. And if you can remove the necessity for those roles, you make everybody that AI superhero and everyone gets to do more meaningful work.

Pascal: Yeah, I agree.

Connor: Oh, Pascal, I could talk to you for hours.

Pascal: No, I am curious what you wanted to say.

Connor: I kind of plugged it in there is the answer for me. I think that the piece for me is I think to your point, a lot of people look at this, especially that RPA process selling the CIO's as companies fundamentally are looking at cutting costs and capitalism and business and this whole thing is just like, how do we cut costs and remove people?

And I think that misses what the core of commerce and business is, which is creating things that people want. And if there is an infinite amount of work to be done, and there is a infinite number of things to do, and there's actually just a finite amount of resources, what AI tools do is increase the amount of resources that we have.

And I think one of the most scarce resource as I feel and I think every entrepreneur feels is like talented, capable human beings because they're all stuck doing stuff that sucks, that they shouldn't have to do. And whether that's like subsistence farming or cooking the same repetitive dishes or it's driving a car around these are all things that like humans have to do because they have to get done.

But if they didn't have to get done, there are a lot of things that those people could be doing that would make everybody's lives better. And if you can unlock that, I think that is a noble and wonderful mission.

Pascal: Yeah, no, I couldn't agree more. And I think like the interesting thing is like in history, we've always succeeded to do so, right? There's, it's not the first time we're kind of fundamentally shifting the toolkit available to humans, right? Like we've, I mean, there's many books written about this and I'm like not in the best position to talk about it, but it's hundreds of years ago, most of us were working in agriculture.

Like now agriculture makes up a single percentage, like a single digit percentage of people, working and we do all these other cool things like, doing podcasts, building tech companies. It's not that we're like sitting around doing nothing. In fact, like more people are employed now than they were ever before.

And it's I think like, oftentimes, especially in media, there's this like mindset, there's like scarcity mindset that gets communicated, especially when it comes to AI and automation that I think is a big problem. It's an important debate to have and I don't want to say let's not talk about it at all and I could talk about it and think about it, but I think it's only half the story.

And most of the times we find ways to do more interesting things with our time, especially as we move more after like high, high friction, low value work that, unfortunately all of us spend a lot of time doing.

Connor: I mean, I think to your point, I think human beings have this tendency towards, I'll use the word productivity and I mean that even with like art and music. And I think that those are productive endeavors. And I think that there's this human bias in that direction. And I think there's a lot of people who

want to be productive in ways that they enjoy and the necessities of economics, which I think isn't a top down like this is you are, the system has decided you are poor. And therefore, I think there's luck of the draw component there. And as a result, you can't be productive in the ways that you find stimulating and enjoyable because you have necessity.

And if technology and AI itself can increase the net output of everybody. Wealth goes up and people in general are able to spend their time in productive ways that find them joy. And I think if there's a lot of people that are like, look, I want to stay at home and I want to, eat snacks and watch TV.

If you have AI abundance, like you can do that. Like, it's because I think even today, the idea, the amount of leisure that we have today is unfathomable to people not very far in the past from us, and I think that a lot more of that technology ends up unlocking more leisure and giving more and more access for people to have that versus wake up, work, go to sleep, die. Which was like, was the human condition for most everyone.

Pascal: Still is for a lot of people like we always argue from like a US, European standpoint and we're very privileged and what we see there and we still see a large chunk, like a majority of humans alive today in fact don't have any free time. So I think there's a lot of work to be done there that we can, as humanity do 10 to a hundred x better than we're doing today.

And then at some point we may run out of things to do, but I'm pretty certain we will think of new things that we can build and come up with.

Connor: Well, I could talk philosophy of AI with you for hours and hours, but I have to be respectful of your time especially joining us from later in the evening. And so I want to thank you, Pascal, so much. Let's hang out soon. Come to New York. Let's get dinner and we'll talk for four hours. It'll be amazing. And we'll make something happen, but this was an absolute delight. Thank you so much. And I hope to see you again very soon.

Pascal: Thank you so much, Connor. It's great to be here.