How AI 10x'd our Sales Throughput (Trey Miles)

Full transcript

Welcome back to the podcast, guys. Today we are joined by Trey Mills, a chief revenue officer at Upstart 13. Trey, welcome to the podcast. >> Thanks for having me. >> To kick this off, could you share more more about your background and what do you guys do at Upstart 13? >> Yeah, absolutely. We like to say we make AI real. So, our goal is to help organizations implement AI in a way that's meaningful. What I mean by that is is uh we we call it boardroom to build. So we start all the way from the left side on strategy. So where could AI impact an organization all the way into execution? We have a pilot to production program that takes 90 days. So we we take people through all the way from ideation through to production implementations in 90 days of a specific work. We we like to say we like to think big, start small, scale fast. So we like to think with the end in mind.

So when we design a a solution, we're thinking for something that that can scale. When we think about implementation, we want to get as small as possible that's going to add meaningful value. So we we scope this thing down to the smallest meaningful impact and then we scale it quickly and then we move on. So once we've done one, we go to the next one. So what we tend to find is we help an organization achieve massive impact on a single use case and we scale. I don't know if you've seen the MIT study, 95% of AI projects fail. That's how we solve for that is we help our customers be in the 5% by having a really clear game plan of how to get to the goal that they're trying to achieve and really knowing what that goal is. So yeah, it's what we do. >> Great. And you describe your lane as revenue leadership in the age of AI. >> Yeah. >> In real terms, what has changed for revenue teams in the last year in your opinion and what hasn't changed at all?

>> That's a good question. So I think what's so let's start with what's changed. So the thing that I see as as having changed is a lot more personalization. So in the past we always used to segment our accounts into like A B's and C's and the A's you'd be like hyperpersonalized outreach really detailed understanding of the organization. You can do that with every account now. the amount of of time it takes to hyperpersonalize content, deeply understand an account, do the account research. Like I used to spend hours a week as an AE back when I was an AE, researching, listening to 10Ks, understanding my account base. You don't have to do that anymore. There's a zero reason to listen, read a 10K or listening to an earnings call because I could feed it into AI and ask it meaningful questions about my business. I can actually take it a step further and say, "Hey, you know, what do I need to know going into this call based off who I'm talking to, based off what you can see about the organization and what do I need to pitch?" So, being able to tailor content at scale, whether that's for outreach, whether that's for meeting prep, you know, fill in the blank. There's an older book called The Challenger Sale that came out a few years ago. The core foundations were teach, tailor, take control.

Those are the like the the key three talking points of how to sell in the modern age. And if you think about it, teaching and tailoring, it's a lot easier with the help of AI. I can tailor my content to the person I'm talking to. I can educate myself so I can educate my clients better. So yeah, I just think it's it's scale, right? I mean, we can we can do for 20 account what's we used to be only be able to do for one. On the other side, that's actually pretty similar to what I've seen be impactful for us is our ability to just to scale. You can read it if you want. There's an article on our website called how we 10xed our sales throughput. That's how we did it. We did it by removing the barrier to personalization and documentation by leveraging AI. We have agents per stage of our pipeline. We've literally created agentic processes across our entire workflows. Um, we have outreach agents, we have calling agents, we have agents that help people prep for calls. We have agents that build documentation.

We have agents that build presentations. We've fully enabled our organization using agentic technology. And we've 10xed our ability to to execute. I can do with one rep what I used to need 10 to do. And that's that's what that means. So it's it's a massive it comes to scale. Yeah. >> And when it comes to the personalization part that you talked about >> what kind of tools do you see most useful as I tried clay? >> Have you heard about clay? >> Yeah, I know clay. Yeah, we're not using clay. We honestly we've kind of built it ourselves. We're using everything kind of inad workflows and chatgpt you know anthropic. We don't lean on a specific set of tools. We use a lot of tools. We use kind of everything we can. And it just depends on which the what what problem are you trying to solve. I think a lot of people try to approach AI like, hey, I want to I want one model to fit all.

And it's that's really not what it what's happening. What's happening is there are different different models are good for different things. And so if you apply the right model to the right problem and understanding and so I think the the challenge is for people what's the right tool? You do I need a hammer or a screwdriver? Right? There are different tools for different problems and applying the right tool to the problem is is as much the answer as any. Yeah. I mean we even have different tools for channels, right? So some tools empower our our BDRs teams to have a have a tailored pitch. Someone picks up the phone, right? We have a different model for LinkedIn outreach. We have a different model for email outreach because there's different tools that that work differently depending on the channels. our are our our tech stacks highly diverse because we've just found as we've implemented these solutions that different tools work better for different different solutions >> and the open claw has been on a big boom recently. >> Yeah. >> What do you think about it?

>> We are at a hard no >> hard no. >> Hard no. Yeah. Nobody We're at corporate IT security said uh nobody puts it on any corporate machines. >> Yeah. because it released a blog article about a a dev who canceled its it canceled the request, right? Like >> it's too sec issue. >> Yeah. Yeah. We it's too autonomous. So we we don't trust it at a corporate level. Now we do love co-work from from Anthropic because it checks in with you, right? Like it's not going to send the cool thing about co-work is it checks in. So if it doesn't understand the answer, it's going to ask you. I mean it's like a good intern versus a bad intern, you know? like Clawbots's like a bad intern. It's like, "Whoops, I accidentally sent somebody $450,000 when I should have spent them $45,000." Where as Claude' be like, "Hey, uh, did I get this comma right?" You know? So, like, yeah. No, we're a hard no claw. Uh, we love co-work. Co-work's a big part of our workflows.

That's the other thing I'll say like from a design principle perspective, not just with how we build the solutions we use, but also how we build solutions for our clients, we always recommend human in the loop. I don't care if it's sending a quote or sending an email. We always think about a spectrum, right? Fully manual to fully automated. So when we think about designing a solution, you want to go from the fully manual to the fully automated over time. And what that means is you're going to do less and less human in the loop. So what you do is you start highly human involved, meaning every quote is being reviewed, every, you know, every email is being reviewed. And as you go, you just start to increase the level of automation as you build trust in the system. So that's what I don't like about Clawbot is is there's not those granular trolls. So that's just not how we design solutions. >> But on the other hand, I've had pretty good success with cloud code. Do you use that?

>> So we leave it up a lot to our developers. So it depends on kind of who you talk to. So yeah, we have cloud code, sometimes it's cloud, sometimes it's you know Open AI's codecs. We've used cursor in the past. So it just depends on kind of what's the project. Again, it goes back to like not everything's a nail. Not everything's a screw. So, you just got to make sure you're using the right tool for the right solution. So, I' I'd recommend anybody who's like, "Hey, what should I apply to this problem?" Try a couple of different things. Test them out. Like, if you put the same prompt into 10 different models, you're going to get wildly different answers. So, becoming too dependent on a single model limits your ability to execute because you start to shoehorn your problem into the into the model as opposed to picking the right model for your problem.

So that's that's a better design principle to think about >> when it comes to you walking into like a revenue organization and you want to know if it's healthy >> fast. Are there any numbers that you look at first? >> Meeting volume. Meeting volume is the atomic unit of the sales process. And it doesn't necessarily have to be meeting volume. It's whatever your first entry point into pipeline, right? If you're working in a in more of a PLG motion, it might be signups, right? So there's two there's two metrics that I always look for. What starts the process and and I call it the pivot point. Where's the pivot point in the sales process? If you think about a fulcrum, right? A fulcrum pivots. So you're looking for the point. So I'm looking for two points. What's the starting point and what's the pivot point? The pivot point is the point in the sales process where the um conversion rate jumps to over 50%. So at some point in the sales process, you go from more likely to close than less likely to close. That's your pivot.

That's your pivot point. How healthy is our entry point? And then where's our pivot point? And what's our conversion rate from the entry point to the pivot point? Because what I'm looking for in that question is product market fit. Do we understand the right fit buyer for us to go talk to to get into the pivot point? And if we understand that, and it all again, that also helps me understand health, right? Like where, you know, how healthy is our sales pipeline? Most people don't understand it. So, I'm usually coming in and diagnosing it. But those are the two points that I look for in the sales process. What's the entry point? What's the pivot point? And what's the conversion rate from one or the other? And then who. And the other thing to note, there's a great story uh from World War II where they were trying to reinforce bombers, the US was, and they were armoring the part of the plane that had bullet holes in it. So that the bombers would come back and they'd look at the plane and they go, "The plane has bullet holes in these places. We should armor those places." And then someone had the brilliant idea to go, "Actually, no, because these are the planes that made it back.

Those are the places we don't need to armor, cuz that's not the places that crashes the plane. It's the places where they weren't shot that we needed to armor because that's where the other planes got shot. So the question is not know across this sales process. Do we understand our ICP? Do we understand the conversion rates? And everything's about conversion rates. There's volume and there's conversion rates. If you understand those two things, you can understand is this business healthy? Like can I effectively hit my number? And again, the answer is when I walk into an organization is no. When I come into an organization, they're going to go I don't know. Like cool. where in your sales process do we convert to 50%. They're like, "Oh, we think it's here." Or you cut tuck into the data and they're wrong. Right? So like, you know, it's but that's the two things I'm looking for. What's our entry point? What's our pivot point?

So if we can get better quality coming into that first point, that's how we ultimately drive outcomes. >> And is there any vanity metric that you see people looking at when it really isn't important? Yeah, anything that doesn't drive revenue is a vanity metric, right? So like every metric can be a vanity metric if used wrong. Every metric can be a not vanity metric if used right. And what I mean by that is to understand how to hit the number. If you're let's say like in a PLG motion, right? Like monthly active users could be a vanity metric if you don't know how to convert them into revenue or if there is no revenue associated with them. Website visitors can be a vanity metric if you don't know how to convert them into marketing qualified leads. Marketing qualified leads can be a vanity metric if you don't know how to convert them into intro calls. Intro calls can be a vanity metric if you don't know how to convert them to your pivot point, whatever that is, PC, whatever. PC's can be a vanity metric if you don't know how to convert them into revenue.

Right? So, every metric is a vanity metric if you don't know how it ties to revenue directly, right? But every metric can be extremely meaningful if you understand what it means along your revenue journey. So to answer your question, are there any like always vanity metrics? No. Website visitors is a perfectly meaningful metric if you understand how that gets into marketing qualified leads, turns into intro, turns into PC, turns into revenue. >> So it's really understanding the whole picture around each metric. >> The question I always ask people is what what are you trying to understand with this metric? What are you trying to get it to tell you? It's all signal, right? And again, it's it's highly valuable signal. Signal is incredible. Like, we want signal. If I'm trying to answer a question, data is better than no data. So, so what is what do I mean by that? I'll give you a really good example. Took over, you know, I'm working on a marketing project right now and um you know, we started running some LinkedIn ads. Before we started running the LinkedIn ads, we had no idea.

After we started running the LinkedIn ads, the number of people who clicked a specific point of our website was almost 10%. But they weren't finishing the form. They weren't converting into to form fills. So the question becomes, what is wrong after that button click? If you think about it, I have a ton more data. I'm asking I may trying to answer one question at that point as opposed to trying to answer 50 questions. Is the number of people who clicked on a button a vanity metric? Maybe. But it's data. It helps me understand what do I need to go change about my process? Now, if I'm not willing to do the hard work to change, then yeah, it's a freaking vantage. You go to the doctor and they're like, "Your cholesterol is high." Well, you don't do anything about it. It doesn't matter. If it causes you to make change in how you operate, then yeah, it was a very important conversation. If it doesn't drive any change, it didn't matter, right?

All metrics are about understanding what do I need to go change or update in my process to drive the outcome that I'm trying to achieve, which is ultimately revenue. >> And when it comes to bottlenecks that most teams deal with, what are they? >> So, it depends, right? So most of the bottlenecks I see have to do with customization, right? Tailoring. It's un it's it's building the right content to convert the client, right? So this is where I think AI becomes hugely powerful. And this is where I think you you really should think about AI in an organization, which is where are my bottlenecks? I have an article about this on our blog. So if anybody just wants to go to upstart 13.com and read my article about this, it's there. But um what I found is we're creating custom content. Like so if you think about a deal right we're creating custom content for a specific deal and it may or may not convert right so it could be on the front end it could be BD right we're creating tailored custom outreach and emails and everything but if you solve that and you start getting really good meetings then you create a bottleneck on people so it's how much work needs to occur to accomplish the outcome of that stage by the person doing it and do I have enough people to do it now there's two ways to solve bottleneck and throughput problems one is we throw tools at it, right?

This is where AI comes in. Or we can throw people at it, right? We can get more hands to do the work. We can lower the cost of those hands. There's all kinds of levers we can pull to unblock that. What I've typically found is when I solve a problem on the left side of a process, I'm going to create a problem on the right side of a process. So when I unlock a bottleneck here, I'm going to create a I'm going to create a bottleneck here. Right? So what we're typically trying to do is we're just trying to solve the next bottleneck. And again, the metrics tell us which bottleneck to go solve, right? Like the bottleneck may be conversion rate on PC's. I'm typically looking for like where am I off on my headcount to conversion rate. So it's like I have enough people but I'm not seeing enough throughput or I don't have enough people and my throughput starts to degrade. If I think about like metrics going back to what you asked me earlier, they're my like instrumentation. They're my dials.

Like if I'm flying a plane, right? I've got all my dials, understand what's my telemetry, like how high am I, you know, what's my odometer reading, how hot's the engine running. So my metrics tell me those things. So, I'm looking for like where is follow follow-up dropping on MQL2 intro call conversion rate. That may mean I don't have enough BDRs or it may mean may need mean I need to create more automation. Where is my like PC to PC completion offs? Maybe I need more pre-sales. Maybe I need automation. And bottlenecks as is a good example. I see my bottlenecks show up in my metrics. And what I typically do when I see those bottlenecks show up is I go ask the question, is it people? Is it process? Because sometimes it's even just a template, right? Sometimes it's well we're recreating the freaking deck every time and we just need to templatize the deck. Like that solves the bottleneck, right?

So there's always a different answer to problems. AI is a great solution, especially if it's like we're creating custom content and slowing us down and it, you know, it takes time. Man, you asked me a really big question just then. >> What's your guys vision at Upstart 13 in the upcoming year? We got a huge revenue goal, but I you know I really think that for us it's it's all about how do we help organizations? How do we empower organizations? How do we enable organizations to you use AI meaningfully? I feel like there's a lot of waste in the AI space. There's a lot of people who are experimenting with client dollars. And so for us, it's all about like how do we empower the organizations that we serve to truly start to see value from these tools. Right? There there is value. I've 10xed my sales throughput. I can do with one rep what I used to do with 10 reps. like that's proper freaking value, right?

There can be real impact, but um there's just a lot of people doing it wrong. Our goal is to help people do it right. Whether that's they pay us to do it or we do it through content and we we serve them in that way, right? But, you know, our goal is to just help help people make AI real and serve them in that way that helps this become a real force multiplier for these companies. >> That sounds like a exciting goal. >> Thank you. Yeah, we also want to triple revenue. So, I mean there's that but >> that comes along with it. >> Yeah, that's a that's an output of the service and and >> Right. Okay. So, thanks for joining, Trey. I will add links so people can check you out. >> Yeah, great. >> And we'll see you in the next one. >> Awesome. Great.