Luca Micheli: The AI Pivot That Took Customerly From $100K to $1M ARR

Speaker A: Luca. Ciao.

Speaker B: Ciao, man. Long time no speak.

Speaker A: It's been a while.

Speaker B: Um, yeah.

Speaker A: So look, I was really excited to connect with you again. Uh, our first episode, if you remember, it was published. Uh, I was looking at it this morning, January 13, 2020. Uh, so we must have recorded it in December around this time. So it's almost been exactly six years since uh, we spoke on the podcast last. And it feels like the last six years of changes have been the equivalent of a lifetime. Like if you think about it, shortly after your first episode here, the world shut down because of COVID You know, that lasted a few years. That literally changed the, the world and people's behavior in a major way. Um, in late 2022, ChatGPT launched and you know, this AI era began and that was the beginning of arguably, in my opinion, the biggest transformational, one of the biggest transformational, if not the biggest changes, uh, in human history. So again, only six years, but definitely feels like a lifetime ago. Lot to catch up on with uh, yourself, customerly, what you've been up to. But if you don't mind, Luca, just for those who might not know you, just take a minute to introduce yourself. If you don't mind.

Speaker B: Sure thing. Thank you so much for having me, Costa. It's such a pleasure to be here again. Um, I didn't realize it was like already six years and Covid passed and uh, of course GPT and everything. So I'm the CEO of Customerly and uh, it's a bootstrap SaaS, uh, helping B2B SaaS companies to improve their customer service, uh, operations. And now more than ever, um, has been, um, major help for um, mid size and enterprise businesses. I'm going to tell you how we changed uh, our business model, uh, from like 2020 to now because AI of course changed everything from the product, but also our go to market and our positioning. So we can help B2B SaaS for sure to improve and automate customer service tickets and uh, way much more. So happy to dive deep into it.

Speaker A: Yeah, yeah, no, let's definitely get into that. So a couple things before though, in our first uh, podcast again, almost pretty much six years ago you shared that customerly was at 100k arrow. Uh, so you guys are still very early. Do you still share revenue numbers? Like where are you guys at today?

Speaker B: So this year we should, if, if a couple of contract, uh, close. We're very close to it. One million mark.

Speaker A: Nice.

Speaker B: And it's uh, actually is, it's way much less the. Our ARR revenue, but way much more the AI conversation revenue. So our business model switch completely from just a seat subscription to a uh, SIPs subscription, plus the AI conversation. So our north side metric for the revenue from now on is going to be way much more the number of conversation we managed because that's direct revenue from us. So it's way much more. Like right now it's like 60% AI revenue and 40% licenses.

Speaker A: Okay.

Speaker B: And if you think about it, in our business, it's going to be way much more um, per unit revenue in terms of conversation rather than seats because seats are going to decrease while conversation will increase.

Speaker A: Right.

Speaker B: So it's, it's challenging. It has changed business.

Speaker A: So that, that's great. So you've, so you're at uh, but you're including the revenue with, with both the seats and the AI like credits.

Speaker B: Right.

Speaker A: So you're, you know, hopefully gonna, gonna pass a, a million in ARR this year. Is that correct?

Speaker B: Correct. Correct. It's again, it's the recurring revenue because actually it's increasing because if the companies we are, we're uh, supporting are increasing their business, their, their figures, basically that's increasing also the number of conversations we are selling to them.

Speaker A: Yeah. So yeah, amazing. Okay, that's great. Um, so you know I mentioned AI has been the biggest transformational change in human history. Um, you know, we're chatting, uh, offline. You said something interesting which you just alluded to it, alluded to it about how AI has completely disrupted your strategy. Um, and you don't even think like customerly is like a SaaS business anymore in some ways. So can you expand on that? Expand on like what happened? Uh, you know, how did you pivot and all that?

Speaker B: For sure, as you might know, we always hated chatbots. Uh, we have this tagline, chatbots are shit. Because we uh, never invested into that kind of automation because in our um, in our experiences we always had bad experiences with traditional chatbots. You know, they were pushing away people and we didn't want to get into that kind of automation. So thankfully we never invested into the shitty chatbot experience. Okay, so never invested into creating the tools to let our customers to automate early years before GPT, um, any chat conversation. So we decided to, went like plain human to human support. When GPT came out, we basically started to think this through. Now we can improve that experience without cluttering it with a lot of buttons and a lot of, you know, shady experiences. So we created this layer of AI on top of what we already had and we focus strongly, heavily on the quality of the responses. We always wanted to escalate automatically to the human being if the information wasn't there, if the customer was requesting for a human being all the things that we hated as consumers. We wanted to implement it into this model. We're using GPT of course, but there are several prompts for every single message. And we decided to implement the best possible solution out there for our customers. And so number speaks from themselves. We are managing thousands of conversations a day with an error rate that is less than 1%. So we wanted to build out one of the best um, implementation of AI out there. And uh, our customers that have been using several of our competitors plus customately AI, they're saying to us that it's actually easier to uh, implement and is responding way much better. So when we had this realization we said okay, we need to change our icp because the customers that are getting the most the value out of this are the mid size enterprise companies that are managing thousands of conversations a day. Because if you go to like smaller businesses and we were serving mainly small businesses, um, they don't feel the value, they don't feel the benefits of having an AI model replying to 10 messages a day. Okay. When you start automating hundreds if not thousands conversation a day, that is something that you feel as ROI on your processes and operations. And so we started, you know, we always been product led. We always focus on creating a product led experience, uh, word of mouth targeting channels, uh, for small businesses. And now all of a sudden we needed a switch, we needed to go sales led. And to me it's like a completely different ballpark. You know, we never had, I didn't know what was a PO number, you know, all that kind of stuff. Ah, that now is, is there, uh, we needed to switch, we needed to hire a VP of sales to understand how to structure a saint.

Speaker A: Okay, so yeah, there's a couple things then to uh, to, to uh, jump in there. First of all, I thought you'd be wearing your uh, your shirt, uh, chatbots or what does it say?

Speaker B: Yeah, I should have. Yeah, it's kind of gold.

Speaker A: Okay, so you, you mentioned error rate less than 1%. Um, I noticed on some of your marketing it was 2%. So has that improved recently?

Speaker B: Improved, yes, we are actually deployed another several um, updates. Ah, which basically now is more skilled into finding the best resource to gather information from. And so we decreased the uh, the, the percentage of hallucination out there. Um, yeah.

Speaker A: Which model are you using? Is it OpenAI.

Speaker B: OpenAI, yes, different models for different uh, tasks. So whenever a message comes through, we have several like four or five prompts that goes and search for like. One of the key differentiation is not that like, is that we're not just a RUG system based on FAQs or documents, your knowledge base, but we also have this layer of intent that you can prompt as you want. For example, if you want to treat in a different way a uh, refund process or an upgrade process, um, you can prompt the AI to manage it in a very skilled and logic way. We're deploying this grounded AI concept where you can define for which intent you can prompt AI to behave the way you want to behave. Um, and on top of that of course you can connect it with your external, um, CRM systems or database to process the refund. So you can automate not only just the FAQs but also second layer tickets from start to end. Um, very complex scenarios actually.

Speaker A: That's amazing. And is 1% error rate, is that like industry norm? Like is there any data on like what is good? Is that considered like better than industry standard or

Speaker B: um, well there are two metrics that uh, most of the time are shared from the industry and is the closure rate. Um, so how many conversations are closed automatically? Uh, currently all our customer base is 71%. So no matter the training materials, no matter their industry, the average customer is 71%, which I believe it's a really, really good number. And as I said, you can have the best model out there, but if you don't train it in a good way, uh, it's going to be shitty. Like it's going to hallucinate or it's going to escalate a lot. Most of escalation or errors we see is actually human error not training the right way. So we are trying to deploy also tools to let our customers improve their models on autopilot. So saying, okay, this should not be here because it's creating hallucination or something like that. Um, I created also Masterclass because right now there is another issue I found in the market, which is the customer service managers, most of the time they don't know how this technology works. There was a gap of knowledge on how they should be training this model or they should change their operations to train models right moving forward. And they're pressured by the sea level to adopt as fast as they can the AI in support in their industry and automate as much as possible. So that's a challenge I saw in the market and I Said, okay, let's create a master class with all the things that we learned by building this software. So that's another thing that we decided to do, um, to improve the experience. Okay. Of the end user.

Speaker A: Okay. So like Customerly was, was hit head on with AI, right. Like the first department that most companies um, try to kind of use AI and disrupt with is customer support. And Customerly is a company that provides directly customer support related tools. So. And it seems that you guys acted and pivoted pretty fast. Was that due to just client demand or did you see the writing on the wall and were proactive?

Speaker B: We definitely saw the opportunity, we saw the quality of uh, the models and we decided to jump on the train as fast as we can. And thankfully we didn't have any previous in house model or automation layer. Um, because we can build on top of conversation, plain conversation, and release the model as fast as we can to all our customers without being stopped or uh, without having any issue to destroy whatever we created before we could have just mounted this AI layer on top. So we were really fast in, I would say in a couple of months we went from zero to understanding what was the best way to train an AI. We actually released um, on the very first um model the intent classificators and the mission. So basically an agent AI that was already doing some intent classification and you know, structuring a flow in a completely different way. It wasn't just FAQs, it was just really advanced at the time.

Speaker A: Okay, so no, that, that's Great. You mentioned 71%, uh, closure rate average for customerly clients, uh, which look to me if I can automate and close 71% of tickets, um, no matter how big, like how many tickets I get, whether it's 10, 20, 50, 100 per day, to me that's, that's very valuable. Um, for sure. But so there's probably a lot of companies including myself, who are thinking about how to leverage AI to handle customer support and correct me if I'm wrong, and you kind of alluded to this, but I feel like a good AI tool is only going to be as good as like the, the internal data that it's learning from. So a couple questions then. So before considering a tool to use, what are some things internally that like myself or anyone listening that should be consider, that should be considering to tighten up first, like your, your knowledge base, your FAQs or like. So that's one part of my question. And you said the average is 71. So what are the companies doing that are above 71 versus like the lowest percentage. Like, where's that difference coming from?

Speaker B: Really, really good questions, actually. Um, so to, in my experience, the way you can choose a really good tool to understand if you can actually deliver on the promise, um, it's not that easy. But first of all, you should understand how easy is to set up the way the model behave with very tricky questions if we are talking about only customer service inquiries and how it could handle like processes rather than just an faq. So that's one thing. And it's not that easy to understand from the outside. Okay, you want, um, you won't discover that until you get your answered dirty, you know, with some um, trials and understand how the model behave. Um, so that's one thing, um, another thing that we, like we've been struggling like people behave that believe that you can throw everything at a model and expect great results. So we've been having customers asking us, okay, can I upload the transcript of all our webinars and expect a really good results? That's not the answer. Can you imagine the kind of clutter that there is into a webinar transcript talking about whatever. So that's not a good idea. It's better if you polish up, uh, the knowledge base articles or the FAQs and structure them in, in a good way rather than having like hallucination later on. Okay, so I'll give you another example. Another customer of ours was having a banner for every single article with a link to their own page because some of the articles were uh, indexed on Google and that was a sales banner basically that was creating hallucinations, um, to like during some conversation we saw that that was uh, the cause for hallucination or some candidate responses used for training. Uh, we're stating some piece of info like, uh, we're working on this bug already. And so that was an hallucination, a very bad hallucination if you think about it. A conversation with a bug is coming through your AI and AI thinks because it's generative AI, it thinks that based on that faq, you're already working on that, your team is already working on that. So that ticket could be completely closed and you didn't know about wasn't escalating to your human team and that was creating other issues. People have this misalignment of expectations versus the actual quality of the output of the AI, probably because the bias that GPT, like the consumer GPT is creating on us. You know, you expect that this thing is aware and trained on everything and Whatever you throw at it is going to perform to your customer facing um, uh, experiences. But it's not like that. So the way we build it is it's not going to gather information from the general knowledge because if the topic should be tailored to your business, you don't want the AI to gather information from external sources. So that's why we build it in a very strict way. On the other hand, you need to properly train it for very specific reasons.

Speaker A: So is it safe to say the core knowledge is through the knowledge base, the FAQs?

Speaker B: I would say yes. And we built it in a way that if the model has a low confidence score, it's going to escalate no matter what.

Speaker A: Okay.

Speaker B: So if, if she doesn't have the knowledge or a very low confidence of what she found in the knowledge, she's going to escalate. So in that way you can close the loop because you will see the topics that are generating more escalation for either missing information or low confidence issue. These two are clear signal that you need to increase. Yeah, the, the, you should add more knowledge base. But there are some topics that needs to be managed in a different way with an intent. For example, the refund. You wouldn't just share the FAQ with the customer. Maybe you want to recover it. So there are customers that are coming through that can automate 80% of the their support tickets in two weeks because their knowledge base is already really, really good. And there are customers like okay, I want to throw everything at this and expect 80% as well. It's not like that you can get 50, 40, 50. But then to close that gap you need to iterate on the knowledge base.

Speaker A: Uh, I'm curious if you have data on that. Let's say someone comes in with a mediocre FAQ section and then over time, with the help of the AI and knowing which uh, topics or conversations have to be escalated and then you, you improve the FAQ section. Like I'd imagine over like a few months that closure rate would, would significantly increase, would go from like 40, 50% to you know, that 70, 80, if you're constantly refining,

Speaker B: depends a lot on the volume that you're getting and the spread of the topics. Yeah, okay. And of course the quality of the knowledge base. So we got um, like this, this customer that had a really good knowledge base, 80%, uh, after two weeks and then they started to implement intents. Okay, another customers with larger volumes and um, a larger topic, um, volume. They started with 45ish. And after three months of iterating on knowledge base and you know, intense. They've got a lot of intents, a lot of dedicated intents to manage. Um, and so that increase the quality of the responses. And some teams actually needs a mixed approach, AI first and then escalating to the human to create different kind of tasks. So it really depends, it really depends on the situation. If it's just informational, the kind of request that you're getting, probably you're gonna get 60 plus percent in no time, two weeks time. But if you need a lot of back and forth, a lot of um, check in your database the information or like order information. Some of them can be easily um, resolved with AI intense plot mission and gets you to 70, 80% fast. Um, but yeah, it's a matter of understanding the business every single time. We need to analyze what's the business look like and what kind of customer inquiries they've got.

Speaker A: Is it able to learn from like um, an inbox? Like let's say you have a support like inbox where like all your support is you're doing it manually. Can the AI like learn from past, like how you answer?

Speaker B: So there are different approaches from our competitors and uh, we have a strong point on this and I want to ask you a question. Would you like the AI model to be trained on um, very specific quotes or offers that you send to one particular customer and using that very specific piece of info with all the customers or prospects out there?

Speaker A: Um, I don't know. I, I, I could see, I, I could see both sides to that to be honest. Um, I'm looking at it just again from my personal um, situation. I, I would say if I had the option, I would say like yeah, I'd be okay with the AI learning from my uh, specific like support inbox because like you know, the way we, we do support, we try to like structure it in a way where it's, you know, it could be relevant for every customer. And we, we try and we're very deliberate on just making support efficient and turnkey. Like I'm not trying to do like very uh, isolated support, uh, inquiry. Like yeah, there are nuances and like certain customers. But um, generally if it's an answer for one person, it's the same answer for someone else.

Speaker B: Okay, that's good. The way we wanted to build it is to avoid this because when you train it with the old database of incoming queries, it can hallucinate a lot. Like there is a lot of work to understand what the source of truth is, right? Uh, without even considering that within the inboxes you can get a lot of spams which the model will not probably know about it. Like you can prompt it to clean it up. But again, I don't feel, I don't feel okay having a, a model trained on whatever it is on the inboxes. Uh, also because like some cases are very specific and sometimes the bugs are resolved already from, I don't know, six years ago. And I wouldn't, I wouldn't train the model that is responding now to, with an updated content that was created six years ago. I don't know, I've got mixed feelings, honestly. Is more is more a no? Because those reasons rather than a yes yet? I can feel the benefits of, you know, again, throwing everything that I was working on that can be considered good as uh, you know, optimizing the training. But at the same time we still need to consider that this is generative AI and uh, she doesn't know how to still think through whatever happened, I don't know, six months ago rather than now. Product, product changes fast nowadays. So that, that's another thing to take into.

Speaker A: Will it ever get to the point though where it's like you, you had a support inquiry like two years ago about a bug or whatever, but like it, it, it learned from like a newer message or like an updated FAQ that like, hey, that post, that uh, email from two years ago has been updated. Like, is it there yet? I'd imagine it will be like, if it's not already, like soon. No,

Speaker B: it could be, it could be there. There would be certain ways to, to train it with several, like past several issues that you might be having. And uh, yeah, it's interesting. Um, again, I don't feel that confidence to do it because imagine you're not checking your inboxes anymore. You got, in our case, you're gonna get the Aura, uh, inbox dedicated inbox. It's like a teammate of yours. So you're not gonna get escalation if she could handle it. Um, and sometimes like you check of course the CSAs, the conversation quality, the sentiments analysis and everything, but if you want to put it onto autopilot, you will never know.

Speaker A: Yeah, makes sense for a typical company then, like how fast is it to kind of implement um, your AI like into their business?

Speaker B: Um, so we can import everything in no time. Like uh, from certain providers we can import also conversations, ad center, customer data, everything. Um, but generally speaking we run um, demos with actual data of the customer no matter where is their knowledge. Base um no time, like 10 minutes. We can train it all and if we need to like run specific scenarios like intent transmission for refunds or upgrade or whatever. Um the, the general setup window is like two weeks or like very structured processes to go live with all the things. But we are talking about like mid size enterprise deals, you know where you got several different departments working all together into this project. So like a small business can actually go live in one hour.

Speaker A: Yeah. Okay. Um and how, how much like um upkeep is like recommended with like the updating the intents and you know the like all that stuff internally just to continue refining it. Is that like uh, like do you have recommendations for that is or does it depend on the business itself?

Speaker B: Again it depends on the business. First of all you track the, the metrics, you know the closure rates, uh CSOTs and you can spot like any issues coming through the supports uh which like we, we track and we our customers also track escalation rates, type of escalations and topics escalations and of course see sots. Um for aura M Generally speaking seesaws Aurora is higher than human teams. Yeah like 20 higher. And I was like you know what, like what's happening why? And I been talking with teams and they say that most of the time is because the speed of the answer. So they, they were used to have the first answer in like 40, 45 minutes. Right now it's 7 seconds and it's accurate.

Speaker A: Amazing.

Speaker B: You know so you're getting support in less than one minute before you had to wait 45 minutes just for the first reply. Um so CSAT is another really important thing to see and to iterate on. Um, if you see that aura is maybe treating a uh, certain topic in, in a not good way you should iterate either on knowledge based on intents intent don't change that much. Um mission can be fine tuned. So the way you tell the AI how to behave that for sure something you should improve. But if it's working, it's working. Like if that process is working already maybe there are some nuances you have to fix. But rather than that it's like there are some teams managing just mhm. Their goal is actually to improve the automation rate and to do so they need to like understand what's what what's the next band of conversation that you need to automate either with a knowledge base article or an intent plus mission. So you check whatever is already automated the quality and you check and improve and iterate with a feedback look loop on whatever is remaining. Right.

Speaker A: Yeah. Makes sense. What are, what are some of the biggest uh, frustrations or complaints that you hear from your clients related to the AI?

Speaker B: Our clients or our client clients?

Speaker A: Your clients? Any, any like complaints like. Mhm.

Speaker B: Honestly now I'm thinking like very few.

Speaker A: Wow, okay.

Speaker B: Very few. Like there are some times that, that hallucinate. Of course most of the time is because it can be um, like a human error or sometimes like the, the signature can let into like uh, weird stuff. But like lately, two, three weeks ago, one of our best customer called uh, me and say look, we, we're getting um, an outage and we are submerged by tickets. Aura has currently 10, 000 tickets open, which is something unusual for them because like Aura is just closing them. And, and she said can Aura handle it? And I said yeah, why not? Like what's the uh, like why the question? I don't know, like this is going to be a lifesaver for, for us. And after the holdings ended, she said thankfully Aura exists because otherwise like they don't. They should like manage more than 10,000 tickets in a very short period of time. So spikes are not an issue anymore. And um, the team can actually focus on fixing the issue rather than panicking and uh, managing all those conversations. So honestly like I don't get any complaints about how our AI is working.

Speaker A: I really like the um. I think the key is the Aura being able to judge when to escalate it. Um, you know, that having that, that understanding of, you know, when a human should step in rather than trying to, you know, solve it on its own. How do you, how do you judge that? Like you know, do you have any, what's this, any specific criteria that you guys use for that?

Speaker B: So as I said, we didn't want to create the shitty shuffle experiences so we prompted within the system that if the customer is asking for a human is going to escalate. We check sentiment analysis on the whole thread to understand if the customer is pissed off. If that's happening, we escalating. And um, there are other triggers. Um, another one is low confidence. So we have a very high confidence score. Um, our customers can change the threshold. So it's like 90, 95%. So if, if it's not that high, it's going to escalate as a low confidence issue no matter what. And the last one is like missed information. So there is no information about this. I'm going to escalate. So those are the four general escalation types that you're going to get and there is no way you can fix your customers into loops. Like there is no way to not uh, like to not escalate to a human team. So at certain points that you get will escalate.

Speaker A: Makes sense. I like that. Um, let's switch gears a little bit to you know, sales and marketing. So you mentioned that you know, you're, you're, let's say like a few years ago you were catering to smaller companies and now you're more like medium mid size, like enterprise. Is that correct?

Speaker B: Correct.

Speaker A: So has your um, has your arpu, like revenue average revenue per user has increased significantly, I'd imagine A lot, yes. Okay.

Speaker B: So like everything started actually three years ago when we went to uh, Saster in Silicon Valley. I heard this speech, uh, from the cebo. What is the CEO at the CIPO now? And um, it's something that hit me hard because they actually had the very same path that we had. Like they were struggling to sell to small businesses, they were churning fast and they felt uh, their customers couldn't get their value, the value they were sharing. On the other hand, mid size and enterprises were getting value, were paying way much more and churning less. I said yeah, this is something that is the same thing that is happening customerly. And so I had um, the pleasure to have a private mentorship with him. And he said look, it's going to be a long journey but it can be done. And so like I felt it, it was already a long journey. And we started, you know, restructuring all processes to close the sign up rates and get into sales LED processes, which is like a bloodbath honestly. Like you see your free trials going down because you are now like filtering at the door, the signups. You're saying no, no, we don't want you. And like this is scary. You know, you're kind of cutting revenue but at the same time you're focusing on better revenue. Okay, Better qualified prospects. And so also my like the demos we're getting are like lesser. Like we're sending quotes for 200k. Um, before we were talking about 1, 2k a year and we were getting complaints about that, you know, and now it's like, okay, we want to expand, we want to work more with you because this is actually helping us a lot. So it's tough because passing from a um, product led, um, experience and team, like all the team is focused on that and now it's like sales led.

Speaker A: Okay, so that's what I want to get into then. So what's your team size now? Like how Many team members at ah, customer. Ten people, still small. Um, which is great now. So, so where have, where does the. What's the biggest marketing channel now for you guys?

Speaker B: Thinking us for sure. And outbounds.

Speaker A: And outbound. So outbound is something new that you guys started doing recently or.

Speaker B: We started um, one year and a half ago. We decided to hire an agency. Three, uh, people, sdr, BDR and the product, uh, manager basically, uh, overseeing everything. Uh, it was a complete failure. Complete like six months, zero results. Like a bit of replies, but zero results. So I understood that if you want to run like a proper sales process, you need to, you need to develop it internally and then if you want like outsourcing it, that was a big mistake. Um, but we learned that we needed to create the old process inside. So then we hired a fractional VP of sales from LinkedIn and he helped us a lot to structure the outbound sequences like the um, Clay Automations and LEM list all together to get to the actual prospect and then structure the deal. Because I was running demos for very small businesses and when you switch from small businesses to thousand employees, company sides or NASDAQ visit, companies are like completely different ballpark. So I needed that kind of help because I never been inside that kind of industry and company sides. So that was a huge challenge. And uh, next year we want to hire um, two people for like sales and like one SDR and one AE and to run the whole process and start um, scaling up a bit more onto this.

Speaker A: Do you still have the uh, fractional, uh, VP of sales?

Speaker B: Not for now because everything is set up the proper way for that kind of process. But it's like a discussion that whenever we need him for like different kind of challenges is there and uh, we can work with them.

Speaker A: Yeah. On the paid side, um, where is that coming from? Is it Google or like Meta or um, on the page you're doing uh, paid ads?

Speaker B: Oh yeah, yeah. LinkedIn ads is the one that is performing the best and a piece of Google Ads. But honestly, LinkedIn, uh, ads is working because we enrich the data with clay. Um, like their targeting is amazing. We can pinpoint the companies and the icp. But we discovered that by polishing up the data with Clay, we managed to get better results. One of the best um, ADV out there we put out is the Masterclass, uh, which is generating lead at a very low price because, you know, uh, it's a knowledge gap that customer service manager have. So that's a really good lead magnet that is working out for us.

Speaker A: What's a uh, better performing marketing channel at this point? Is it ads or is it the outbound?

Speaker B: I would say definitely, yeah, yeah.

Speaker A: Like are we talking like 70, 30, 80, 20 or.

Speaker B: Honestly man, it's like I feel it's not the right answer because it's a mix of them.

Speaker A: Yeah.

Speaker B: So it's a mix of people seeing the ads, then getting an outbound email and then getting the outreach on LinkedIn is a mix of different steps and also different people we are targeting at the same time with different messages, you know. So uh, it's, it's, it's a mix with this kind of contract and go to market strategy is like, it's impossible to say where like to attribute to one very specific channel. It's a mix of them.

Speaker A: So have you noticed then the amount of like just um, product LED turnkey people signing up without speaking to anyone or doing demos, has the percentage of, of companies signing up like that decreased a lot?

Speaker B: Yeah, yeah, 100% also because our efforts are not there anymore, you know. So last year we've been focusing a lot into content, blog content, um, to target those people, target those uh, small companies. And we stopped doing that because we could see the quality of the leads coming through the kind of process and we say okay, we need to focus on a different icp. So as I say like, like trials going down. But the quality of uh, the conversation that we're getting right now during demos are uh, like 10x. Like I need to get probably 200 conversion from signups to get one single contract I'm discussing right now. Don't get me wrong, like closing one of these contract is like six months plus in the making. Uh, so it's not that easy either. Uh, I feel like it's a completely different ballpark.

Speaker A: Totally, you know, for sure. Are you still doing most of the demos yourself?

Speaker B: Yeah, yeah, yeah, absolutely.

Speaker A: Okay. And so there's no. Are there you have one other salesperson now or you're looking to hire?

Speaker B: Yeah, it's uh, a person that is helping me to do books. Yeah.

Speaker A: And then you're looking for SDRs and uh, AES.

Speaker B: Yeah, yeah, very good.

Speaker A: How has, has sales? Is sales like a new skill for you? Is this something uh, like new territory?

Speaker B: I would say that enterprise sales is a new skill for me. Absolutely. Like understanding the complexity of a deal. It's uh, has been hell of a ride to learn all this stuff in such a fast way. You know, I'm a product person. I love to optimize products as A salesperson, you know. But yeah, no, I love the old sales. Like it hits differently. Closing a 200k deal with just one, uh, companies like completely different ballpark, you know.

Speaker A: Absolutely. I, I say this all the time. I wonder if you agree is, I think personally sales is the most important, uh, skill set for any founder or entrepreneur, point blank.

Speaker B: Yeah.

Speaker A: That's especially a bootstrap.

Speaker B: Yes.

Speaker A: Because.

Speaker B: Oh, yeah.

Speaker A: The only source of funding that we have as bootstrappers is sales from clients. If you can't sell, your business dies.

Speaker B: Yeah, exactly. Our customers are, uh, our investors. So you need to do that. Honestly. Um, and I was an introvert, you know, as a programmer, I was an introvert, you know, I didn't like to be exposed to. So I needed to learn a lot, courses, books, whatever, to understand how to close a deal of, of that size.

Speaker A: What are some of the biggest, uh, takeaways or lessons, uh, on the sales front that you've learned in the past year or so? Especially now, selling to more enterprise. Any, like, little skills tips, tricks. I know it's ever changing and evolving. You're continue. You're always learning. But is there anything that kind of.

Speaker B: So recently I've been exploring the Challenger Sale.

Speaker A: Okay.

Speaker B: I don't know if you've heard about it. It's very cool, actually. It's the way it's. It's a book, um, that I was reading and when I started doing some challenge, um, to our prospects and implying how they should be improving their customer service, that started to change the discussion later on and I saw a lot more. Okay, I'm listening, I'm listening. More than just pitching the software and saying, this is, this is the demo, um, this is the product. So I'm also challenging the prospect to think in a different way because the way they're doing currently is not the right way. And you're here because you know that it's not the right way and I will help you to change that way. Change is not always easy.

Speaker A: Yeah.

Speaker B: And so you need to challenge the person you're speaking to get into that kind of solution. But I was not doing that. Uh, I mean, not, not like most of the time I was not doing that. But now that I know, um, the process is actually really cool. So I need to explore and dig deeper into it. But it's something that I definitely love.

Speaker A: Is this the. Just, uh, looking it up? Is it. It's a book called the Challenger Sale. Yep.

Speaker B: Correct.

Speaker A: Taking control of a Customer conversation.

Speaker B: Yeah.

Speaker A: Okay.

Speaker B: It's one of the most Effective kind of sales, uh, salesman is the challenger sale, the one that actually challenge um, prospects. Um, like it's, it's, it was unusual for me too and uh, it actually works. So that probably is one of the best learnings in the last year.

Speaker A: That's, that's one of the things I love about sales personally is that like to me the best sales people are the ones who understand psychology and human behavior. Um, but, but the beautiful thing about sales is that it's not like black and white. Right. You can have, there's like an infinite number of ways you could do sales with like you know, based on your personality. Someone might be an extrovert, an introvert, challenger sale method versus another method. And you can have like two different people doing two completely different methods but still be as equally as successful.

Speaker B: True.

Speaker A: You know what I mean? I strongly believe that that's, to me that's the beautiful thing about sales, um, is just like the, the intricacies of it. I think a major factor of it is like understanding who you are as a person and playing to your strengths.

Speaker B: That's true. Always believed in it. Yeah. Another thing that I was thinking about, you know, passing the knowledge and uh, expand the team into um, more sales people. It's like how can I pass not only like on like passing the knowledge regarding the industry and what already we did with our customers. That's, that's the value. That's the value I'm providing during the demos that I believe is going to be challenging for and not non tech sales guy on how to challenge the prospect in a way that can actually click. So that's, that's one challenge I need to figure it um, out before getting there. For sure.

Speaker A: Yeah. It's interesting about the agency because, and I had a feeling you were going to say it wasn't successful because that's something I learned very early on. I'm, I strongly in, I would, I would advise against hiring for a position that you haven't done before.

Speaker B: Yeah.

Speaker A: You know, like for example, like if you're, if you haven't done outbound and you haven't made calls and demos and you just straight away hire someone, now you're kind of just arbitrarily relying on this person or people to create the benchmarks for you and you're just like trusting like hey, this is how many calls I can make. There's so many opportunities, there's so many demos. You're like, you're trusting that person. Whereas if you haven't done it yourself and especially As a founder, I, I don't think, I don't think anyone can outsell a, uh, founder in general. It's just like, just, just the, the confidence, like, your knowledge of the product, like, how things work. Like, I think people can, can smell and sniff, like, the, the passion and the confidence. Like, there's so many little things that a founder has that let's face, like an employee or someone you bring on. They don't bring that same level of a passion that you have. Um, and I say too, in my business, I don't care how good of a salesperson you are, like, no one can outsell me.

Speaker B: That's true,

Speaker A: but that's my whole point. Like, if I haven't, like, I need to do it first to validate it, to just understand, like, what's realistic, what's not. And then once I've kind of done it and know, like, you know, benchmarks and criteria, then it's. It's a lot easier to hire someone after the fact.

Speaker B: You know, I was mistakenly believing that because they had a really good success case in our industry with some sort of competitor of ours in another market.

Speaker A: Yeah.

Speaker B: So I was already believing that, you know, probably they already got the structure, the, the way to do it, so they can just replicate it in, into another market, adapting to our messaging and could work. It doesn't work that way, so. No, never does.

Speaker A: Because sales, it's. There's so much testing and tweaking that's involved. Like, there's, there's a lot of nuances from company to company. Even though if you think you're a competing company, there's still like, yeah, ICP differences. Maybe you're catering to a slightly different segment or pocket of customers. And this just requires, like, a, A very deep level of understanding that requires a lot of tweaking, a lot of testing. Um, the way I look at it is like, you know, think of like a, you're on a, like a giant boulder, like a big rock, and you're like this little person on top of this giant boulder with a tiny chisel. And like, every day you're just like, chipping away at this, at this boulder. And the, the chips represent, like, tweaks in the process, the messaging, the communication, like, the structure of your emails, your opening pitch. Like, these are all tweaks. And eventually, with enough tweaks, you eventually get a little crack in the boulder. And that crack represents, like, some traction. And again, in my experience, it's just very difficult, uh, to, to have, like, someone external doing those those tweaks for 100.

Speaker B: Uh, on the other hand, like, you might be experienced this, but as a, as a founder and CEO of a small company, you probably are like, uh, a jack of all trades, right? So I thought that if I can't do this right, and I'm playing my cards, I need to figure it out as fast as I can or try to get knowledge out of this as fast as I can to learn and uh, create the process internally. That was my mistake, like, for sure. We learned a lot of things. We learned how not to do sales, outbound sales. Right. Um, so in a way was a lesson that I could save, uh, by learning myself, but somehow I accelerated.

Speaker A: Makes sense.

Speaker B: I got some piece of content, some piece of knowledge that I needed to internalize that. Right. Still I need it because I didn't have in myself. I didn't have the picture, the whole picture on how to do that. So it was a expensive lesson for sure. I wouldn't do it again. Um, but I was thinking, you know, probably it's easier to. To get this, you know, fancy agency with a lot of case studies in different industry, working worldwide, uh, with something that they could replicate. Um, I was wrong. Totally wrong.

Speaker A: That's business, I guess. That's the game, right? Any, any experience, good or bad, you got to take it as, as a lesson and you gotta like, take something away from it, um, you know, just to propel, uh, you forward. Right?

Speaker B: Yeah.

Speaker A: Um, Amazing. What are you. You mentioned like, clay and Lemless. Are these your, like, internal tools that you're using on the outbound front?

Speaker B: Yeah, yeah, 100% plates. Just amazing. Honestly, in the way we're working, um, with. With it to. To get contacts, companies list. To be honest, I built my old SaaS company list elsewhere because getting like the information, a piece of information of who is doing SAS is very complicated. So yeah, Clay, it's the richer. We also use Apollo, um, and different other sources to get the companies. And then we use like pipedrive, um, lamb list, um, LinkedIn, sales navigator, uh, like. Yeah, that kind of stuff.

Speaker A: Is, Is clay just then like in Enricher for, For the data?

Speaker B: Uh, right now, yes.

Speaker A: Or does it also act as like the this or. I guess you mentioned Pipedrive, like, acts as like the CRM where you're actually communicating and.

Speaker B: No, no, we have. Actually it's a very complex scenario and in between some, some sometimes there is relevance, AI, uh, which we fit with a lot of data coming from Clay and pipedrive. And then like we update the data. There Are some agents doing, uh, updates on um, pipedrive? M. I wish Pipe Drive can have a very improved AI because right now it's so old and uh, you still have to do a lot of AI, a uh, lot of manual stuff which can be automated with AI. Uh, but right now we got Clay reaching a lot of data coming from different channel sources, also visits. We use deal front for that. Um, and it's enriching the data and pushing uh, it towards Pipedrive or back to LinkedIn ads to target that kind of people. Also back to customerly because we use customer whenever you start using it. There are, by the way, there are some enterprises getting into the trial and then hitting the support or sales to speak with us. So whenever there is a trial, we enrich it to understand if it's a good deal to reach out to. And it's getting better in terms of the quality of the deal entering the free trial as well. Um, so we need that because there are some enterprises getting into the free trials.

Speaker A: Okay. And is the outbound something you're doing like every day and just. It's ongoing?

Speaker B: No, I'm not doing it personally every day. It's a setup. So, um.

Speaker A: Yeah, that's what I mean. Like it's set up, it's running.

Speaker B: Yeah, it's running every day and I revise them every couple of weeks. There are several outbound sequences for different intents you got, or different processes. The thing that I'm um, struggling the most is the intent, because there is no, like a proper intent of when you want to switch or get. Adopt an AI solution for service.

Speaker A: Yeah. How many new, um, like contacts are you reaching every day? 200 a day. Yeah. Okay. Do you have any data on like, like how many positive replies you're getting?

Speaker B: LinkedIn is working crazy good in terms of reply emails. Very few. Uh, but I would say 15 more or less on average. On all. On all, uh, channels. Like accounting. LinkedIn because like some straightforward questions are actually the ones that are getting us the best answers. And then of course there is the iteration behind that. So again, it's um, for us it's a mixed approach where we shoot a very straightforward question and we start, we start working on that. Um, so LinkedIn, again, email is not performing very well. Like only email is not performing anywhere.

Speaker A: Yeah, I have the same similar experience with email. It's uh, I feel like it's, it's a lot of structural, um, like email was a lot easier like you know, 12 years ago when I was, you know, growing my, my first company. But a lot has changed. Um. Oh yeah, nowadays that's good though. 15%. That, that seems very high. Amazing. Um, that's great. So when you have a positive reply, obviously you're trying to like get a demo. Is that like the next step?

Speaker B: Yes, to try to understand if they want to have an audit of their current situation and to understand if we are a good fit for each other. And of course, like, that's not getting them onto a demo or like understanding what they're doing, how they're doing it and what's their kind of challenge they have, um, because they're coming from another solution. So we need to understand what they're using, how they're using it and what kind of results they want to get.

Speaker A: Yeah. So what, what's like the day to day CRM? Like if you have a positive reply now you're communicating with someone or like phone calling. Is that pipedrive or is that spike drive?

Speaker B: Yeah, Pipedrive. Actually, actually, no. The first layer of communication until you, you get to a deal point point like when we actually engage onto a meeting, um, it's. Everything is on LEM list because the inboxes, the outbound inboxes are merged, uh, together in there. So we use LEM list for the all. And of course like LinkedIn messages because it's faster and easier. But when that becomes a, uh, deal, it goes into Pipedrive and everything else goes into that.

Speaker A: Yeah, makes sense.

Speaker B: We're also using Fireflies and Fireflies is, um, connected with relevance and uh, and pipe drive as well.

Speaker A: Yeah, very good. Amazing. No, this is, uh, this is great, man. I, um, do want to be mindful of your time, Luca, but so we know round of 2025, end of 2026, where do you see customerly?

Speaker B: Well, I definitely want to eat the 3 million marks. Uh, it's gonna be challenging, um, but I believe we can get it and by like, by the quality of the leads that we're getting, I'm sure we're gonna get there. And of course, uh, increasing the team, enlarging, uh, the team is going to be another great, great, accomplished, uh, 2026.

Speaker A: Amazing. So, yeah, that's, that's a, that's an ambitious goal. I think you'll get there. Yeah, I have a lot of confidence in you and your product and uh, it seems like you guys are on the right track.

Speaker B: Thank you, man. Thank you so much.

Speaker A: Random question, uh, just on your website now. Are you still in Italy or is it Ireland? Yeah, you're, you're in Ireland.

Speaker B: Yeah, Dublin.

Speaker A: So you moved there yourself as well?

Speaker B: Yeah, yeah, I'm living there since eight years now. 2007.

Speaker A: Okay. Okay. Okay.

Speaker B: Yeah.

Speaker A: Nice. Very good.

Speaker B: People say that I'm downgraded my life, but actually, I believe it's, uh, it's an upgrade.

Speaker A: Yeah.

Speaker B: Honestly? Yeah. I cannot go back.

Speaker A: No. Is your team all remote as well?

Speaker B: Yeah, all remotes, yeah.

Speaker A: Do you guys have, uh, like, do you have teammates in Ireland as well that you see face, uh, to face at all, or.

Speaker B: Not yet? We got some, um. Some of the times we meet all together when we can, because, uh, you know, it's getting. It's getting difficult to. To have all the team in one place.

Speaker A: Yeah.

Speaker B: Um, spread across Europe. Yeah.

Speaker A: Okay, so just in Europe is where your team is.

Speaker B: Yeah, only Europe. Very good.

Speaker A: Is. Is that deliberate or is that just.

Speaker B: No. Uh, I mean, we tried, first of all, gdpr, access to data and everything. It's must be in Europe. Second of all time zones, it's way much easier to work on. Um, almost the same time zone our. Our clients, uh, and all the teams is.

Speaker A: So.

Speaker B: Yeah, that's. That's mainly the. The reasons why.

Speaker A: Yeah. Are most of your clients in Europe or in the US And Canada?

Speaker B: Yeah, most like. Yeah, most of the clients are in Europe. Like, 80% in Europe, and.

Speaker A: Oh, really? Okay. Very good. Okay. Luca, um, this is a pleasure, man. Thank you so much for taking the time. Uh, I think whoever listened to this got a whole ton of value. I. I certainly know I did. Um, and, yeah, maybe end, uh, of next year, we'll do this again.

Speaker B: I love it.

Speaker A: The yearly roundup where we land here.

Speaker B: Absolutely. Thank you so much for having me. Pasta. It was a pleasure.

Speaker A: Thanks so much, Luca. Um, we'll talk soon, and best of luck with everything.

Speaker B: Thank you. Thank you.

Speaker A: All right, take care.