Summit | Harrison Rose (Goodfit & Paddle): The future of AI in GTM

0:00Harrison Rose:But what do these strengths add up to as a skill set? So they kind of ladder up to decision making for me. I think this is AI's superpower. When we start allowing AI to make decisions and go to market, it's then that I think we're going to see its real benefits. We can walk through an example to prove this out. And using that, AI can do a much better job than us as humans to determine who do we ping, when, with what message, across what channels, in what order, with what budget, with what spend. It can even determine what creative it should use and create that itself. The future of go-to-market for me is giving AI the power to make decisions in go-to-market, not just writing you better outbound emails.

0:38Harrison Rose:AI will evaluate your entire market in real time. It's ultimately going to decide who gets sold to when, with what message, across what channels, with what spend. We are live. We are live. In five. We are live. Four. We are live. Three. We are live. Two. We are live. One.

1:04Harrison Rose:Today, companies have access to the most powerful decision-making tool in history. It's not news to anyone in this room that AI has the potential to radically change how we go to market, among many, many other things, right? But for me, to maximize its benefits, we also need to start using it radically differently, too. Over the past year, I think if you've asked many of the companies that you speak to day in, day out how they're using AI, You'll hear about some lots of different use cases. You know, Andreas has mentioned one there in terms of people using as a tool to write outbound email. Today, I want to propose a different solution, a different future.

1:39Harrison Rose:But before we talk about how we should be using AI and go to market, let's spend a little bit of time talking about this implementation in the past. So I want you to think back to whatever was going on in your life in 2017. It was a nicer time, quite frankly. Go wherever your mind takes you. some prompts you harry styles released his first solo album certified banger i don't know if you've got any fans in the room and i didn't know what a black swan event was to be honest that was a reprieve that i did not know that i needed at the time i also had a lot less gray hair albeit it does show up when it's a little bit longer i had a haircut yesterday i was also the co-founder of paddle i was running go to market uh not hadn't yet started good fit uh we were the fastest growing software company in the uk around 2017 for those of you that don't know paddle we were selling revenue infrastructure to other software companies.

2:26Harrison Rose:So like checkout, recurring billing, payments, taxes, fun stuff like that. But as we were scaling Paddle, finding software companies, those that we wanted to sell to, was really, really hard. And this was confusing to me. At this time, software was eating the world. Software companies were going to be everywhere. I was like, why is it so hard for me to get my hands on the folks I want to direct my team at to go and win and create opportunities with? And it was really weird. Maybe some of you guys have encountered this problem yourself looking for software companies that you might want to invest in.

2:52Harrison Rose:If you go to places like LinkedIn and filter for software, you get tons of false positives like the software development agencies all over the place. Even worse than that, like the false negatives, like Hootsuite is a great example. They list themselves as marketing and advertising. Actually just getting visibility into the set of companies I wanted to target was really, really hard and a big limiting factor. So what did we do about this? So we felt like we invented this role called the lead development representative, the LDR. and it was their job to manually review thousands and thousands and thousands of companies and domains websites to determine whether they were in this case software companies you wanted to sell to or not as well as manually enrich them with a bunch of additional data honestly it was terrible terrible job horrible um it was expensive i think we probably had seven or eight ldrs full time grads on london salaries they were very insular the jump from like ldr to bdr was huge like you were just researching stuff, and then you were this customer-facing rep.

3:50Harrison Rose:It just wasn't working for many, many different reasons. Now, fortunately, one LDR I did hire was a guy called Alexander Berry. He's now my co-founder here at GoodFit. And honestly, such was his distaste for what he was doing and what his colleagues were doing. He didn't half let me know about it. But we quite quickly worked towards just automating what it is they were doing. So what did we do? We built what's called a classification model. This was my first successful, I think, deployment of AI in go-to-market. And again, we're talking back in kind of 2017 now. For those of you that are unfamiliar with classification, I'll do my best job of explaining it super quick.

4:24Harrison Rose:But you basically take hundreds, if not thousands, of accurately labeled software companies in this example's case, as well as hundreds, if not thousands, of companies that aren't software companies, and you can train a model on this data. And eventually, you can throw other companies, or in this case, domains at this model, and it will spit our prediction whether it is indeed a software company by our definition or not. This was long before LLMs, kind of an early and overlooked form of AI, albeit one that's still really valuable today. Classifications are still better at this job than LLMs. You can normally iterate and train a classification model to like 90 % accuracy.

4:57Harrison Rose:LLMs normally top out about 75%. But what did we learn from this experience from this deployment of AI? Well, I learned I could use models to evaluate and kind of map this entire software market in real time. This wasn't something I had access to before. Normally, humans are going through this stuff one by one. and suddenly I could throw hundreds of thousands of companies at this all at once and eventually I got this like real-time view into every software company out there that I could sell to, every net new software company that was emerging. I was also gathering a huge amount of information on these software companies.

5:25Harrison Rose:We called it the software universe. And actually a number of VCs we were pitching at the time were more interested in this than Paddle and a number of them actually tried to buy the data set. Yeah, it was pretty mad. It kind of prompted us maybe to start GoodFit. But what I learned in the context of Paddle was it was cheaper, faster, and more accurate at what it is that it was doing. compared to the humans that I had on the job before. Frankly, I think they were just limited by their own boredom of just going through all these thousands of companies. Now, if we fast forward a number of years, Paddle is still using classification, actually.

5:55Harrison Rose:I've got a lot more gray hair, but hair nonetheless, more than I can say for my co-founder of Paddle. But we see LLMs hit the mainstream in November 2022, then agents and everything else that has come since. But how does this change things to go to market again? Well, what do most people do with the most powerful tool we've ever got our hands on. Unfortunately, most people just started trying to automate the work they were doing before. Unfortunately, a lot of that work was frankly just bad work, like stuff that wasn't working. And because it was automated, they also just decided to do this bad work at really high volumes, right?

6:29Harrison Rose:But what are some examples of that bad work? I mean, we've talked about email. Let's stick with it as an example. We've all been on the end of an automated email that's just crap, quite frankly. Where people used to wait for me to raise like our series A and B and decide that was the opportune moment to contact me about buying their tool for 50 grand, now they can automate that same terribly ineffective message. Like, it just doesn't make any sense, right? People began automating the work they were doing before and at much higher volumes. Now, this has not made people happy. In fact, it's made people feel like the guy on this slide.

6:57Harrison Rose:I haven't seen this show, but it looks like a good one. I mean, in the case of automating mail, for example, there have been cries of, like, outbound is dead and all sorts. I think, honestly, just a much higher volume of people were doing this at much higher volumes. It's hard. It always has been hard, and it will continue to be difficult, but it is not indeed dead. However, what else has happened? There's definitely been lethargy of being on the end of some of these tactics, maybe some frustration too. I think in Sales Martech more broadly, some of the emergent tools around this time also just massively overpromised stuff.

7:27Harrison Rose:Like some of these tools out there were saying, look, we're going to find your ICP. We're going to find the personas in this ICP. We're going to automate messaging to some of them. We're going to deliver that message automatically across all these different channels. We're going to deliver you millions of dollars of pipeline all hand tied in a bow. All you need to do is sign my POC or annual contract or whatever. Funnily enough, this didn't work. And there was huge churn in SalesMartec, right? Not good. It was a bloodbath. And worse yet, I think there was a lot of skepticism as well about AI and go-to-market.

7:55Harrison Rose:Now, we can do better than this. I do think AI will have its moment in the sun with go-to-market. But to get there, rather than using AI to automate the work we were doing before or ramp up volumes of that work at least, I think we need to start using it to do a better job than we were doing as humans. Just like we did with classification back in 2017, I guess. That was a bit of inspiration for me. When AI starts improving the quality of our work, of our decisions, of our strategy, it's then we're going to see it as kind of transformational benefits, right? But we need to lean into its strengths.

8:27Harrison Rose:That sounds good, but what is AI? What are AI strengths? What is it better at than us as humans? I started to think about that a few years ago, And I kind of thought about it in the context of how I would be a member of my team, right? We all have lots of different members of our teams. They're all very unique people. They each have unique skills. And it's all about playing those people in position. It's helping them succeed in whatever they're doing by giving them the chance to succeed at what they're doing by leaning into their strengths. And I was like, what are AI's strengths, right? Like, I'm sure we might all have a different answer to that question.

8:55Harrison Rose:But for me, three really stood out. AI has massive processing power. It can compute a lot more than simple old me. It can hold context across thousands of inputs, much better than my short-term memory. It's my wedding anniversary today, actually, and I remembered that one, but even without the help of AI. And it's also highly objective, right? But what do these strengths add up to as a skill set? So they kind of ladder up to decision-making for me. I think this is AI's superpower. And when we start allowing AI to make decisions and go to market, it's then that I think we're going to see its real benefits.

9:27Harrison Rose:If you give a rep like one week to research and a company they want to sell to and create an opportunity with, they're going to do a pretty good job, right? Like a rep will research who are the individuals in this company, what do they care about, what are their responsibilities, and what channels do they live, and what problems do they have, and how does my product, how is it uniquely positioned to solve some of these problems? They're going to do some decent job. Maybe they'll collaborate with marketing, come up with a quirky direct mail campaign. There's all sorts of things they can do, right?

9:54Harrison Rose:Not bad. Like in enterprise sales, we call it account mapping. It's like the first thing you do often when you get distributed your accounts at the start of the year or quarter. Not bad, but AI can do a much better job than this. It can do it across every single account in your market in real time to start with, which is already a huge benefit. But not only do you get those efficiency gains, it can do a better job. And it's because with AI, I can feed it everything I've ever tracked from that company over the past decade. I can feed it every email, every reply, every win, every loss, every bit of copy I've ever delivered, every successful and failed interaction I've ever had with companies that look similar to the one in question, right?

10:29Harrison Rose:And using that AI can do a much better job than us as humans to determine who do we ping, when, with what message, across what channels, in what order, with what budget, with what spend. It can even determine what creative it could use and create that itself. Now AI can undoubtedly do a better job here and a much more efficient one. I think we focused on how it can improve the outputs for the seller in question here. But when is AI answering some of those questions rather than the humans? I think it will also result in a better experience of the buyer too, right? Maybe we'll be less likely to throw our computers in the bin as you saw in the gif just before.

11:05Harrison Rose:We need to start focusing AI on its strengths, the things that it's better than us as and us humans are good at, I guess. But not only do we need to lead into its strengths, decision making, I also thought about how do we give it the inputs in order to make those decisions as effectively as possible. One of the ones we've been playing with at GoodFit, and there are many, is this concept of expected value. We've been using it in our own go-to-market, a good fit, and also rolled it out to a number of our clients, and it should just show you the possibilities here. So to start with expected value, you need to enrich every account in your market, have a deep level of insight into them.

11:37Harrison Rose:That's a given. You also need a bunch of first-party data, your wins, your losses, and what contract values you won these companies, the stuff that's normally in your CRM. And then you need to run two models. You want to run one model which predicts the probability of winning an account, your win rate, and you want to run another model predicting the size of the contracts if they are indeed one, right? The contract values of those particular companies. Now, this is incredibly valuable. Like, this is only possible as a result of the advancements in AI, ML, and data we've seen in the past few years.

12:05Harrison Rose:And it's a huge step forward, right? Like, being able to predict the win rate in the ACV of an account before you've ever spoken to it is just something that we couldn't do before. Frankly, if you ask a CRO to do that for an account before they've ever spoken to them, they'd look at you like you were mad because they struggle to do it at pretty late stages in the sales process, right? But today we can do this into a high degree of accuracy. And having that level of insight pre-engaging with a company massively informs the decisions we make around how we go to market. Better yet, when you multiply those two things together, you end up with this concept called expected value.

12:35Harrison Rose:You get an expected dollar value against every single account in your market, a numerical value against every single one of them, which we can use as the input to making some of our decisions. Now, how does that work? How is it helpful to us? Knowing an account's expected value in advance of speaking to it, I can start to answer questions like, is this account worth a rep's time, or am I better to send automated programmatic messaging at it? Is it worth me running ads at this account, and if so, what spend? But we can also start to ask and get answers to some more complex questions too, like, what's the optimal combination of channels I should deliver for this account based on what's worked historically?

13:11Harrison Rose:How can I maximize for CAC without shirking on my opportunity creation rate? you can start asking really complex questions that will be very difficult for us to answer as humans before. AI can do all of this stuff dynamically against every single account in your market in real time. And as those accounts evolve, as they change, it can update, it will learn from every attempted interaction with some of those companies too. And it's just one example, if you feed AI the right inputs as well as performance data in this case, it can start suggesting much more effective ways as to how to go to market.

13:42Harrison Rose:It starts making decisions on how to go to market for you. AI making decisions for you in go-to-market might feel unnatural, but it's not without precedence, I guess. It's happened before. Ten years ago, and I'm not sure I'm even right with that figure, if you wanted to run digital ads, you would decide manually in which places to put those ads, on which websites we were going to run them, right? Which is just mad. I actually don't think I can remember that reality, to be honest with you, which is why I'm questioning the ten years. I'm sure that's true of many of you in the room, too. but with the emergence of programmatic advertising, who sees an ad, when, at what price, and what websites, right?

14:16Harrison Rose:That's all completely decided by machines today, and it happens all in real time. I think moving forward, we'll see that extend to all manner of decisions and strategies in go-to-market too, not just programmatic advertising. All of this stuff will be decided by AI and enabled by inputs like expected value. What does the future hold for go-to-market and AI, right? The future of go-to-market for me is giving AI the power to make decisions in go-to-market, not just writing you better outbound emails. AI will evaluate your entire market in real time. It will learn from every single change in that market, every single new account that's emerging, everything that's happening within those businesses.

14:54Harrison Rose:It will learn from every interaction you've attempted with the folks within that market. It's ultimately going to decide who gets sold to when, with what message, across what channels, with what spend. We're not quite there yet, but the foundations are being built. we're pretty close. And I think the companies that figure this out, that AI is going to do a much better job of answering some of those questions than us as humans, those that are going to thrive. And I'd also hope that some of you as buyers with better answers to who should we target when, across what channels, it might result in a better buyer experience too, which we should all probably welcome at this point in time.