Latest / Growth After Headcount / From Stop Hiring Humans to Be More Human — Jaspar Carmichael-Jack (Artisan) #011
Transcript
- Haris: Stop hiring humans. That was the famous advertising campaign from Artesian. Today I have Jack with me on my podcast, and they made the AR SDR popular. for the Stop Humans campaign they put billboards all over New York, San Francisco, London. They made even Bernie Sanders post about it and it made it to the Joe Rogan show. Now they have a new campaign called Be More Human. So this episode we're gonna see what made them transition to this more positive approach. Let's jump into it. Hello Jasper, welcome to Growth of the Headcount. How are you doing? Jaspar Carmichael-Jack: Doing great. Thanks for having me. Haris: Of pleasure. tell me maybe a bit about Artisian in your own words. Jaspar Carmichael-Jack: Yeah, so Artisan we make AI employees to work with humans to complete different job functions. So our first is Ava, an AI BDR that works with human BDRs to automate the majority of their work so they can focus on doing things that are truly uniquely possible for humans to do, like cold cooling, which is illegal for AI to do, and very strategic social in person selling. Haris: Okay, great. And how big is your company now? Like how many people are working there? Jaspar Carmichael-Jack: Yeah, we've got about fifty people on the team and we've got thousands of customers on our self serve motion and hundreds of of enterprise customers as well. Haris: Okay, yeah. I think I read in an article that you were had even more about one hundred people in a team and then you went downsize. Like what was the story there? Jaspar Carmichael-Jack: No, we so we we've actually brought we a year ago maybe had like thirty. We had plans to grow to over a hundred, but we've managed to be much Haris: I see ya. Jaspar Carmichael-Jack: more efficient than than anticipated and we we're we're growing revenue very fast, but not growing the team at the same rate and managing to increase revenue per employee. That's the the benchmark. Haris: Yeah. Is that for you the number one KPI? Jaspar Carmichael-Jack: No, I think th the the number one KPI would be ARR and and customer retention. Like those are the the two top ones for me. But there's like other other KPIs that feed into us building a a sustainable business. Haris: Yeah, I see. No, that makes a lot of sense. And so you mentioned you don't wanna scale much with people in that regard. You're providing a product which makes also SDRs, BDS, and the sales organization more efficient, that they can do much more with the existing team. And how do you run Jaspar Carmichael-Jack: Mm-hmm. Haris: your own operations then? Jaspar Carmichael-Jack: We do a similar thing. So in every team we try and automate any work that isn't uniquely human. So for example, in our sales and customer success process, we have a bunch of Claude skills that help them do different things. We have Claude skills that help them come up with campaign ideas for customers that help them scope out the optimal implementation path for an account. we obviously our our outbound sales function is is managed by Ava. Ava's the our our top performing PDR. but we Do everything we can to automate non human work. Haris: Yeah, yeah, yeah, okay. How do you decide what should an AI do or what should be with humans? Jaspar Carmichael-Jack: The the bar for us well, A, if it's illegal for AI to do it, a human does it. Like for example, cold calling. You can't legally have Haris: Yeah. Jaspar Carmichael-Jack: an AI cold call. B, if the AI can do it as well or better than the human, then we typically will route it to the AI. Unless it's cheaper to get the human to do it, which shockingly is is sometimes the case. and it's cheaper to have someone on the team do something than to have Fable Five do it on on cowork and burn thousands of dollars of tokens. Haris: do you have a recent example where you actually run into a exa such a case? Jaspar Carmichael-Jack: Yeah. We've had some like crazy cases of people on the engineering team but also on just knowledge work teams spending tens of thousands of dollars on on cord tokens and we've had to to rein it in to the could to the point that we could just hire another person for the amount that we're spending on on cord tokens for some people. So we've reined it in in in some areas and and tried to be more token efficient. Haris: Yeah, I heard before it was like the topic token maximizing and now I heard ma many people experiment as well. Okay, where do we use a top tier frontier model and where maybe can we even plug in an open source self running model? Do you also do such assessment yourself or are you only relying on the big Jaspar Carmichael-Jack: E Haris: names? Jaspar Carmichael-Jack: Yeah, we do. We we try and and figure that out. We yeah, we're we're not we we were into the token maxing phrase at the time as well, but we've decided to take a step back and analyze we're ROI maxing now. Haris: Okay, that's a good term. Error I maxing up. Yeah. yeah, I see that a lot now because it's just the cloud bills g exploded. The cloud bills Jaspar Carmichael-Jack: Mm-hmm. Haris: exploded in that regard and people are just trying to optimize it even further. I think you found when did you found a company? Two thousand twenty two or three? Jaspar Carmichael-Jack: Twenty three, end of twenty three, beginning of twenty four. Haris: Yeah, so yeah, we one could say pretty much an AI native company, like from the first version to now of Ava, like what were the biggest changes for you? And learnings maybe as well. Jaspar Carmichael-Jack: The first version was very much of an MVP. So I I didn't even really understand the workflow of a BDR well. I'd I'd done outbound sales and I'd done the agency I had before, but I was not an expert in the BDR workflow. So the first version, when I look back at it now, I'm like, wow, what what was this? it's just completely different. Our first version was just Ava running one campaign, only sending emails, not responding, not doing anything beyond that, and it was very, very simple. The current version, we viably automate most of what a BDR does apart from cold cooling and apart from anything that's in person. So we automate the vast majority of work for the BDR and we replace a bunch of tooling as well. So we replace Zoom info, we replace outreach, we replace RB to B, and who I know you had on podcast as well. We replace all of the different tooling that most people use in their GTM stack and automate the majority of the work that they're doing. Haris: Okay. And how like the early customers, like did you face then a lot of journal in the beginning to get where you're now? Like it's always with MVPs and especially the BDR function because there was so much hype about AI SDRs. And if you c and you can imagine if you don't deliver in the beginning, it's like okay, like what the first spend money on? Like how is it now compared to before? Jaspar Carmichael-Jack: Yeah, yeah. When when we first started, it it's very difficult to make a good AI BDR product. Like it's not an easy Haris: Mm. Jaspar Carmichael-Jack: it looks easy from the outside because you're like, just send some emails. But it's very difficult to actually build an orchestrate a full system that consistently generates results and pipeline for customers. and it's taken us three years to get to the point of actually having a good product and a and a really good onboarding process that we can consistently deliver value for customers. and yeah, at the beginning, of course, as with any MVP churn was high. We've managed to to move all of our metrics in the right direction. Haris: Okay, no, that is good to hear. maybe even taking one step back, so the AI BDR, you still then have I guess account executive who are closing that idea? Jaspar Carmichael-Jack: Yes, so Ava books meetings for AE's directly onto their calendar. We are launching in a few weeks an AI AE, but it's specifically geared towards low value transactional sales cycles that you would otherwise route to PLG. So for example, on our website we have routing where if a company is not qualified for our sales team to speak to them, we push them straight into a self served flow. We want to have a middle ground where they're they're not qualified to take up a human AE's time, but they are qualified Haris: Mm. Jaspar Carmichael-Jack: enough that we want them to get some kind of demo. And in that case you route them to the AI AE. so you you give leads who would otherwise be forced into self serve an option to talk to an AI account executive. Haris: See no that makes a lot of sense. I talked also the other guy to Jonathan, the VP of marketing of One Mind who Jaspar Carmichael-Jack: nice. Haris: do do and they are like also having roles like go to market engineer, customer success manager. Do you also plan expand to these type of roles? Jaspar Carmichael-Jack: We do have that. So we have a team of customer success managers who help customers with implementation. and our CSMs are are very hands on and very strategy oriented. So they'll help customers define their go to market strategy and help them implement that within Artisan. Because we cover outbound, inbound, and the sales cycle. So we really have an all in one platform that has everything that you would usually have a GTM engineer operating, but we just make it simple enough that you don't need a GTM engineer to to operate it for you. but yeah, we we have that function and we have a forward deployed function. Haris: Okay. No, I think what I understood from Jonathan, they are having AI avatars who do these roles, for example, as well. Complementary to a real customer Jaspar Carmichael-Jack: I see what you mean. Haris: exactly. So basically instead expanding the AI BDR also in other go to market functions, well fully, maybe partly. Like is that on your roadmap or you really want to focusing on the sales side? Jaspar Carmichael-Jack: Right now we're focusing on sales. We might do post sales in the next year. for example, we have chat with Ava in our product and Chatwith Ava acts as an AI C SM helping our customers set up campaigns, helping them do different things within the product. We might productize that for our customers once we've really taken ownership of the top of funnel and sales cycle. Haris: makes a lot of sense. So one conversation I have now with many people is a lot about distribution because what we are Jaspar Carmichael-Jack: Mm-hmm. Haris: seeing now with all of the AI tool, you probably see it also on your side, that developing the product and shipping, you can do it much quicker than let's say five years ago. So I guess Jaspar Carmichael-Jack: Yeah. Haris: your development team is much smaller compared to a comparable company. ten years ago, five years ago, which puts a lot of pressure of course on not pressure, but everyone can build easier. So building is getting commoditized. So how do you see the role of distribution, brand, marketing, etc. in this new world? Jaspar Carmichael-Jack: Yeah, that's interesting. So so I do think that building is getting somewhat commoditized. But there's a couple things that aren't. One, ideation and coming up with a good product idea and figuring out what problem to solve, and actually having taste and the ability to know if the output that the AI creates is high quality or not. That is not commoditized. And I I've seen especially I do a lot of product work. Like I I build prototypes for a lot of the features that we ship before they go into engineering. And AI does not have good taste. Like if you Ask AI to come up with a feature and build it, it will be complete slop. Like it's not yet to the point where AI is gonna help you come up with Haris: Mm. Jaspar Carmichael-Jack: good product ideas and execute them. So having taste and knowing what to build and how to structure the product is is not commoditized at all. So I think that's still a huge part of of the product development flow. but I do think as well distribution is emote to an extent now. You if you have 10 competitors who have a very similar product, why are you going to be chosen over them? And a lot of the reason is If you build a brand that people trust, if you have customer logos that people respect, if you create a distribution moat around your company, you can be competitors even if they have similar products. Haris: Yeah. And looking back, don't hire humans, was it Jaspar Carmichael-Jack: Yeah. Haris: a campaign that supported your visual definition of bread now? Or is it something you're maybe looking back maybe I would have done it differently? Or was it the right thing at that point? Jaspar Carmichael-Jack: I mean th there's different ways to look at it. One, it it really did put us on the map and it and it gave us brand ownership in the AI BDR category and in the AI employee category. So from that perspective, it worked very well. But the on the other hand, we created this product that sure we do automate the BDR function for a lot of companies, but for a lot of our other customers and for a lot of our enterprise customers, we don't fully automate the function. They still have BDRs doing cold calling, they still have BDRs working strategic accounts. so the messaging didn't really fit there. so it was it was probably a good idea for us to do it, but our messaging has since shifted to be more aligned with what we actually do as a company. Haris: Can imagine as well. And probably also tailoring the message to the audience as well, I can imagine. Yeah, yeah, yeah. No also to the previous point. Yeah. Jaspar Carmichael-Jack: Yeah, because I say yeah, because for example, an SMB, when they sign up to Artisan, they usually are using Artisan and hiring Ava as an alternative to hiring a BDR because they can't afford to hire a BDR yet. Whereas when an enterprise company comes in, they're usually augmenting and maximizing the productivity of their existing BDR and AE functions. Haris: makes a lot of sense. And it's it's funny like you mentioned earlier that with the AI slop and the product, I think what is really becoming a mode as well is the domain experience. And then like you basically are becoming the expert in BDR workflows and how the process and you package that in your product basically, right? In a in theory, your product could also be just an MCP server that runs to that if it gets to the same results, right? So the SAS is becoming the UI in a way, but the key aspect is the domain experience and that you can deliver that. Do you agree with that? Jaspar Carmichael-Jack: I completely agree, yeah. Like we if I tried to build artisan again today from scratch, without the knowledge I've gotten over the past three years, even with AI, it would be extraordinarily difficult to understand the problems and workflows and the interactions that our customers have with software and and with their teams. Like it's impossible to build a great product unless you deeply understand the role and the workflows of what you're trying to build. Haris: no definitely. Okay. Jaspar Carmichael-Jack: AI can fill the gaps but it can't tell you everything from a blank slate. And it and it will try Haris: No Jaspar Carmichael-Jack: and and it will tell you confidently, but it will usually be wrong and if you don't know you can't spot what's wrong. Haris: Yeah, exactly. If the level of hallucination it's just like incredible. So recently I started on top of documents and sometimes write, please write your own thoughts first and then let it AI contribute to it. To Jaspar Carmichael-Jack: Yeah. Haris: then it's like what AI adds, it's a little bit filling the gap that you maybe forget, but I think we all need to upskill ourselves in how much can we trust AI in that regard. Yeah. Jaspar Carmichael-Jack: Agreed. Yeah. You need to build your own knowledge so you can fact check the AI and not just blindly accept everything that it says. Haris: Yeah, yeah, yeah. And yeah, we talked a bit about your own operation, we talked about what the product does now. Where do you see going in two years? Where would you like to be? Jaspar Carmichael-Jack: Yeah. I'm gonna give you the roadmap. So this year, over the next few months, we're launching an inbound SDR product. So Ava's gonna be able to do inbound web chat. Haris: Okay. Jaspar Carmichael-Jack: and we're launching the AIA under a different name, which we're announcing later on this month. so those are the two big launches we have coming up this year. We're also doubling down on Ava in the entire outbound ecosystem. So we have a bunch of different campaign types that Ava can run now from event campaigns to MQR reactivation to closed loss to champion tracking to web visity anonymization, really every type of outbound play you could want to run, Ava is able to do autonomously. so phase one is us becoming the default go-to-market ecosystem where we own the entirety of top of funnel, outbound, inbound and sales cycle, either through the AIAE for transactional low value sales cycles or acting as a copilot for human AEs when the sales like it really needs a human, like enterprise strategic mid-market. so that that's phase one for us. Phase two, Artisan isn't specifically a sales company. Like we automate work and consolidate workflows into easy to use software. Not tied to any one vertical, but just in general, that is the mission. So we don't really have a fixed We don't really have a fixed path of what we could do next. We could do customer success. We could do recruitment. We could do any product vertical. after we've nailed sales, we'll launch another vertical separately from sales and and build a artisan to to work with teams to make that function more automated. Haris: No, okay. So acquiring also even more domain expertise than in other verticals to accomplish that. Okay, Jaspar Carmichael-Jack: Yeah, exactly. Haris: interesting. Do you see it on a path to acquire that knowledge? Would you start again from scratch or do you would even consider acquiring through some other means that domain expertise? Jaspar Carmichael-Jack: We would consider both, like potentially acquiring a a small player who's building an AI native version of a a vertical software in in one specific category, or building it from scratch and hiring someone with deep domain expertise to to lead that initiative internally. Haris: Exciting, ex very exciting times. And one thing then okay, let's say I am like a company, I wanna deploy and the product and then but my CRM is in a mess or data quality Jaspar Carmichael-Jack: Yeah. Haris: is in a mess because that's often like I'm having next also remote consulting Jaspar Carmichael-Jack: Mm-mm. Haris: companies. Many companies want to do a transformation, but the moment you lock into their CRM it's like wow, what is happening here? So Which brings to the topic of data quality and the current tech stake, maybe tech debt. How do you see that as when rolling out your product? Is that becoming a blocking point? Jaspar Carmichael-Jack: Yeah. So we have a unique advantage here. Unlike if you use Salesforce or HubSpot's native agents, where the quality of your CRM data is really the ceiling of the quality of your outreach, we have a database of every professional and every local business natively built into Artisan that gets updated on a daily and weekly basis with a bunch of different data points. So we can ingest data from your CRM, run it through our enrichment pipeline. fix any of the data that's wrong and overwrite it with correct values. You you have settings to control what is and isn't overwritten. And then we can use our updated data to actually create the outreach and run sequences. So it's not usually a problem like it is with a lot of other products, if you have low quality data in your CRM and we can layer our own data on top of it to to make it work. Haris: Okay. Yeah, I think it's often a big topic for many companies. I Jaspar Carmichael-Jack: Yeah. Haris: think for you I think the benefit as well because it's top of the funeral. So we're compared maybe to a machine learning algorithm that gives you a churn risk prediction because there you need more company data and they're probably the higher risk of having poor data. So that's probably an advantage. Interesting. And Jaspar Carmichael-Jack: So we do actually we ingest company data as well. So we ingest product usage data and things like that to do cross sell and upsell Haris: Mm. Jaspar Carmichael-Jack: campaigns. In that case, having high quality internal data is much more important than 'cause you don't want to quote the wrong customer metrics to a customer. Haris: Yeah, yeah, yeah, exactly. That is relevant. And and I think many people companies don't have like a data lake or storage of data. Maybe actually do you where do you still store your own company data? Are you already leveraging like a data lake or similar? Jaspar Carmichael-Jack: We see we have billions of records. So we we have a database of basically every person and company. so we have like a whole data team dedicated to to main I'm not super involved in that, but we have a a lot of infrastructure around data. Yeah. Haris: that's good. I think many people who are listening on this podcast are on Ada side that an incubants that want to transform. I think you guys are good fit, helping them. On the Jaspar Carmichael-Jack: Yeah. Haris: other side, there may be AI native comp founders who would like to build something or create something. Do you maybe have three recommendations that you would give such people as an advice? Jaspar Carmichael-Jack: For AI native founders, I would say one, automate everything that you can reliably automate better and more cost-effectively than hiring a human to do the same role. but not things that aren't a good idea to automate. There's a lot of things that are uniquely human and make sure not to give those to AI. Two, when hiring people, hold a very, very high bar. because having the wrong people on board will slow you down much more than not having someone on board for a for a role. and three, iterate on your products as fast as possible, get feedback. If people aren't picking up if if people aren't interested in what you're building, pivot very quickly and iterate and try new things. Because beating a a dead horse is gonna waste years of your life and not have any Strong outcome. Haris: No. It's interesting. It's like the recommendation are like probably similar to the recommendation you would give five years ago, except point one with Jaspar Carmichael-Jack: Yeah. Haris: the AI automation, but indeed iteration and hiring like breaks or helps a company succeed all the time. Okay. Jaspar Carmichael-Jack: Yeah. Yeah, agreed. Haris: Amazing. Jasper, amazing. Thank you for joining. And Jaspar Carmichael-Jack: Thank you. Haris: I'm looking forward, maybe from what's gonna be the next big campaign from your side and Jaspar Carmichael-Jack: Yeah, it's coming this week, so yeah, very soon, keep an eye out. Haris: yeah, okay. I'm gonna be watching. Amazing. Have a good day. Jaspar Carmichael-Jack: Yeah. Cool. Thank you very much.