Episode 35: AI for B2B Sales with Collective[i] Co-Founder Stephen Messer

Host Rob May interviewed Stephen Messer, Co-Founder at Collective[i], the creator of the largest global network mapping enterprise buying behavior using data, artificial intelligence and predictive technologies to guide sales professionals through the activities that lead directly to revenue. Tune in to learn more about Stephen's background, the founding idea behind Collective[i], challenges in deploying AI today, and much more. 

Subscribe to AI at Work on iTunesGoogle Play, Stitcher, SoundCloud or Spotify

Rob Circle Headshot

Rob May, CEO and Co-Founder,

stephen messer

Stephen Messer, Co-Founder at Collective[i]

Episode Transcription   

Rob May: Hello, everybody. And welcome to the latest edition of AI at Work. I'm Rob May, the co-founder and CEO at Talla. And today I'm here with Stephen Messer, one of the founders of Collective[i]. Stephen, welcome to the show. And why don't you tell us a little bit about your background and what Collective[i] does.

Stephen Messter: Well, thanks for having me. A little bit of my background, I'm an entrepreneur my whole life. I started a company with my sister, who is my co-founder over at a company called LinkShare. LinkShare, in 1996, looked at the world of advertising and asked, why isn't there paper reform into advertising? We launched that business and over the span of 10 years grew wildly successful. We ended up selling the company, for about a half a billion dollars, to Rakuten.

For those who don't know what affiliate marketing is, think of it as making a commission-based model for the internet. So, if I'm Dell and I sell computers, I can offer commission anyone with a website. If they promote my products and someone buys, LinkShare would track that user, calculate the commission, collect the money from Dell, and pay out the website. We did that for about 10 million websites globally when we left in 2006. It continues to this day under Rakuten and has done very well.

We left there and we loved LinkShare. It was an amazing business. We could tell you, at LinkShare amongst the 10 million websites, by day three of any given month we could predict within pennies what that website would do for revenue. It was unbelievable. So we loved that business.

On the flip side, we had 50 sales professionals. Their job was to work for us full-time. They were on commission, just like our websites, except we had a distant relationship with the websites and a very hands-on relationship with our salespeople. And they were unable to actually predict any dollar amount accurately that they were going to close. It was super frustrating.

That became the foundation for a Collective[i], which is just short for Collective Intelligence. But our goal was to figure out, why is it that we were not able to figure out how to get our sales team to be predictable, to know what they're doing, and help them grow. And I have to be honest, we thought it was just us. We then invested in 80 different companies. And saw that it was all 80 had the same problem. That was frustrating because we thought then it was all entrepreneurs. And then we sat on some growth companies boards and even some public company boards, and found that they were even worse. Then I just realized, that's just it. Everyone is equally bad and so everyone pretends to be equally good. That was the foundation for Collective[i].

RM: For doing something like what you're doing a Collective[i], which you're using some artificial intelligence, amongst other things. And is it similar to-- I have to think if you go back to the late 90s, that trying to figure out how to do affiliate links and track that was actually difficult technology in the early days of the web. Are there similar parallels to where AI is today and the stuff that you're doing? And how much of it is technical and how much of it is adoption and cultural related?

SM: There are unbelievable similarities. Think of it this way, when we started playing in this performance advertising place, where you had all these brand marketers. Now, no one remembers who a brand marketer is. It's kind of like trying to find a T-Rex, they're all gone. And that's because over 20 years performance-based marketing has just crushed that world.

There was a time when people would look at the color of a logo and think that was more important than a click-through, for example. They just didn't care. They thought the tagline mattered. These people would stress out over how the business card should look. And they would have guidelines for their brand. And they would go nuts over these things. And they couldn't understand when Google built an ad without spending-- built this number one brand without spending a single dollar on ads. They thought it was a crazy thing that could never happen again, until Facebook, until one by one they started realizing it wasn't working.

We would go and sell companies. I remember going to one company in particular, where I was thrown out of their office because I got into a whole discussion about, well with brand marketing aren't you wasting a lot of your dollars? Because in performance marketing your only paying for exactly what you get. Why wouldn't you try this out? And the person came back to me and said, you're out of your mind. That's not how we operate. We're buying very specific brands, very specific places. We're doing all this work to make it work. We know that our customers appreciate that.

I turned to the person and said, “I'm a man, you sell women's lingerie, you send me a catalog simply because I moved into an apartment where there was a woman before me because it's got her name on it. So you're wasting a lot of brand dollars on me. There's a lot of waste.” And he threw me out of his office.

That's what you see a lot in today, I think, around AI, which is people are used to bad data, in general. It's acceptable to have data that's unusable because the human is able to interpolate, usually wrong. They think that's the skill. Take for example the industry we play in, CRM is the predominant technology. It's been around for 30 years. The only other technology I know it's been around that long is email, except email works. Because when you look at the stats around CRM, they're so unbelievably abysmal. 70% of a seller's time is spent logging, not selling. It's become the job. In fact people start changing comp plan to pay you in part for logging and in part for selling. That's insane. That's sort of become these weird models. When you go into them, people believe their gut and their skill is the expertise. If you come in and say, I can fix it, it's almost an affront to what they've done for 30 years. I think that's generally also what's happening in the AI world, at large. Which is people think some of these early simple skills, that don't really work, that's their job, and it's not.

RM: We see this all the time. One of the big objections people will say about buying Talla, but a lot of AI tools in general, is well, let us get our data cleaned up first. And my first question is, how long you have been working on that data cleanup project? Because it's been forever and it's not something-- you know.

SM: What's interesting to me about it is, it's not something that humans should do. It's something that machines can do and it's something that, when you integrate machines now into your workflows, the machines will keep the data wherever you're putting it in a good format. And they will know what's in there and they will know what they put in. And that's the way the world should work. And the other thing that's funny, is CRM is one of the most hated pieces of software. Any company you talk, like everybody that uses it hates it.

Everyone hates it, but if you ask them their top five products, they'll start with CRM. But if you ask them to rank them, and then you ask the follow-up question, you'll say, wait, so you're tell me CRM is more valuable than video conferencing. They'll go, oh, no, no, no. By the time you're done with that whole comparison, it's not even in the top 10. But they believe it's the technology that's there. And it sort of feels like everyone-- I believe everyone's bought Salesforce, not because it works, but because they're able to say we're all using the best technology because it doesn't work. It's sort of everyone's fooled themselves into believing because without it what else do you have? And so it's challenging.

To go to your point about data, you're right. Humans shouldn't-- we tell people we take data from CRM, but we actually don't because we just see that as bias. So, we actually tell them, because they've invested so much money in it. So we say, yes, give us your API. But in the way we operate we will not let Collective[i] at all operate with human entered data because we know it will actually make everything worse. But we have to make them believe we took the data.

You had a great article about sort of this concept that you have to make people believe the AI is there behind the scenes. In ours, these people have sunk millions of dollars into an investment that's completely useless. But we actually have to go out there and tell people we're going to pull this data. And, actually after the fact, tell everyone look, we took the data but we excluded from what we could use because I wanted to give you good outcomes. And you weren't getting good outcomes with what you had today. Why did you want me to take it?

RM: Right, so what kind of outcomes does AI provide. Like let's say I use Collective[i], I'm a junior sales person at a big company, sales team of 500 people. What's the experience like for me?

SM: Our model is a flip AI. We believe that the benefit of what we do is made greater by being a part of a network. In 2017 the Wall Street Journal put out an article where they called us "The ways for sales." Now the reason why is, when any company joins us they're contributing their data so that our algorithms can learn across everybody how the best deals are won. So it allows us to study the seller.

The bigger thing, and the big epiphany beyond our business, was that we need to study the buyer more than the seller. If I can observe a same buying group at a company, across multiple sellers, I can reverse engineer how they make decisions. And that became sort of the real big thing about it.

So, Waze, when you type in your opportunity you want to go after, the first thing we do is we tell you here are the people you're going to be talking to. And it feels uncanny, but it's actually really easy to do because no one's given up any confidential data. We just learned from everyone else in a blind way. You get things like who should I talk to, their most up to date contact information. If anything changes, like they quit their job, you get immediately updated, you get new titles. Then we also tell you everybody that you know that's done a deal with them. Or if you've done a deal with those people, but maybe when they were at another company. So you get to do all these things. And I can keep going on and on and on. I guide you through the deal. I tell you how long it should take. In other words, why does it have to be a mystery? And that's what we love about the biz.

RM: Yeah, exactly. If you're going and you're selling to Walmart, which is a huge company, many, many companies have sold to Walmart. And when the salesperson comes back and says, oh yeah, we're going to close this in 45 days. You know nobody closes a deal in 45 days. I don't know if that's true, I'm thinking of Walmart, I don't really know.

SM: It's a great example. They think they're going to get a deal done, they have all this hope. They may have heard where they wanted to hear. And you, as a company, are banking on that money to pay your employees or to go to your investors. And low and behold the deal does come through and they look at you as being a problem. But how could your rep have known that? If they've never sold Walmart before, how do they know the process? I'd rather learn from 50 other people who came before me. And that seems like a crazy thing is CRM is all about logging what I just learned, not about what I need to win the deal.

In fact, people expend an inordinate amount of time, money, and effort going from LinkedIn to Discover Data, or you can name 50 other things, to try to figure out the basics and most of the time they're wrong.

RM: Now do you guys have problems like many of the companies-- you and I are both investors in a lot of companies. And we've both seen companies that like, they have great products, companies should want to buy, they go deploy them, and they struggle with some of the workflow behavior change. Is that ever a reason that you guys struggle with deals, where the customer, it's good for them and you can show it's working but there's some cultural barrier. And do you think it's an AI thing? Do you think it's like--? What's the--?

SM: There's two problems. One is there is this sort of what I call the Tarzan problem. I'm used to doing it in this old transactional way. It hasn't worked, but I've created so many rules of thumb to make it work the way I can make it work. But, they're trying to figure out, how do I fit AI in that old model.

A lot of times people say, where do my people work? I have to sort of walk them through this logic of, well the machine is doing the work for you to extract all the data, to keep everything up to date, and tell you what's going on. In fact, we send you an email in the morning to focus on the deals that you need to look at and then you're done. Please just spend the day selling.

They go, I don't understand. Where should it be? What they're missing is, they're thinking the work is logging what you do and I've eliminated that. My goal is to improve or optimize your outcomes. They're looking for incremental improvement and I'm trying to optimize. It's two different ways of thinking, that I don't know have anything to do with AI, per se. What it is, is such a change that it's more you're leaving yesterday and legacy behind. I think people have a hard time with that.

The second thing is everything we do is looking forward and their people have a hard time getting their arms around the idea that it could change. Number one question I get is, how accurate is this stuff you do? And I have to always say, we're predicting the future it shouldn't be accurate. I should tell you, based on what changed yesterday, what the most likely outcome is. And I want you to change it if it's a bad outcome. These are concepts that they get when they use Waze every day. But when they go into their day to day job, they want predictability in the future. I think I said this to a client the other day, I think there's like only one Marvel character who could see the future and it's the most boring of all the characters. You don't want that one.

RM: Well yeah, it's like Waze, right? You get in you get to get in your car to go to work, Waze says 23 minutes. Waze doesn't know that in 14 minutes there's going to be a wreck on this route. It's like you can't predict that stuff.

SM: And you don't care. All you need to know is it happened, what's the impact, and is there an alternative I can take. Because that's the real value, not that it would have kept you off that road. The value is, as soon as something happens, you're aware of it and you can adjust. By the way, the funny part is when people go live, after about a week of just using it they get it because it's actually much more natural in this sort of hack on hack to hack something that hacks. They've sort of gone down this road of sort of managing horribly, but they don't realize that life is so much easier but they just can't let go. It's been the job for 30 years. You have three to four generations of people who know only one way of doing things. You come in and you're like, just drive forward and it's weird for them.

RM: Yeah. That's really, really interesting. And so where do you think AI powered sales is going in general? What other ideas and stuff do you see that's interesting or that's coming down the pipeline or things that are really going to revolutionize how people do this?

SM: I start off with the premise that sales today is so bad that people aren't looking at it correctly. So the typical win rates for a new business is 10%. So 1 in 10 people who came to you, that you qualified, that usually raise their hand by asking for a form, got through the review process, that have a project, have money, are the right people, got to the place where it was an opportunity, 9 of them chose not to go with you. That is a horrible client experience. They actively chose to either not do it, which is bad for you, or choose to go with somebody else, which is bad for you. That, to me, is a sign of just bad-- if we were to do that for customer service, you'd fire everybody who was involved in that pursuit.

It's just become the thing. It's cars before they had seatbelts. Everyone just thought I'm a good driver so I don't worry about an accident. It doesn't one matter the word accident means that it had nothing to do with you, but it just means like that. So I think what happens is people wake up and they realize 9 out of 10 people who wanted to do business left and chose actively do not work with you, is an unacceptable level.

The way we look at it is, if you know what's coming in advance. Think about things like multithreading. Today in sales you have this idea of a champion, like when you guys go in and sell your product, Talla's an amazing product, the first thing you're always looking for is a champion. All right? I actually think that will go away. Champions are the worst thing you can possibly do.

Having been a buyer, forget me being a seller, but being a buyer, if I want to do something I have to bring a group of people together who all have different goals of what they're trying to do. And the champion comes in and says, please pick my pursuit over yours and then the politics begin. All the sudden everyone thinks, what's the horse trading we're going to do. You need my time, you need my budget, you need me to make this a success. I'm not going to get credit for it because it's your project, so I have to figure out what can you do for me because if I'm going to do this for you, you better not be able to do something.

In other words, all these people are thinking about, why am I doing this for you? And yeah, maybe there's a problem that it touches everyone, but for the most part, everybody is getting involved. But if I know that five people will be involved in the deal, why not multi-thread it? Find a reason that works for all five of them, then find a connection to all five of them, and have those people all give them something they all want so they all come to the conclusion right around the same time where they're all familiar with your product. You're actually helping them solve a bigger problem and you're actually getting into the meat of what they care about not looking for the champion to try to win it.

Because how often you have somebody that says, I want your stuff, but look now's not the time? That meant they don't have the political clout, somebody else won over them.

How many times do you work a deal and it gets to the end and you realize, wow, we haven't been dealing with the real decision maker and it just gets kabooshed right off the bat.

Thrown out the door, right? You've invested tons of money and time, and you've exhausted that organization. Because when the big person gets in there and says no, everyone's unwilling to fight it at that point. There's already been so many political battles, it's just gone. I don't think that's a good way. I think we have to learn how to navigate the buyer, instead of having the navigator learn to buy through our playbook or things like that. It's very personalized.

The good part about that is, I also believe it means you spend more time trying to actually solve their problems by learning their problems, as opposed to stating facts about what you do, but actually digging in and saying how do I actually solve your problems and let's walk you through it and that leads to post-sale and things like that.

Look, we are user of Talla, so we love you guys. Your customer success team has done an amazing job with us. They've made us think about our business in the ways we haven't and that's actually what we value. I think that's actually sales, in general. That's what it should be.

RM: Yeah, without a doubt. It feels good. I don't know if I've ever told this story on a podcast, but so I was an engineer by training. And I wanted to go, when I was young, I wanted to go work for this AI company that was in Melbourne, Florida, which is where I lived at the time. Very small company, you know a couple dozen people. And I cold called the CEO and I left a message and I said, George my name's Rob May. You've got to eat lunch, I'm really interested in what you do. I'd love to just buy you lunch someday.

He called me back, we met for lunch. He said, well if you want to start your own company someday, you want to work in a new space, you really need to learn to sell. And I said, I don't think I could sell. And he said, why not? And I said well, I couldn't cold call anybody. He said, well, you cold called me. And he pointed out that it's not really selling when it's something I want to buy and solves my problem and you feel good about it. People have this connotation of sales of you're using Jedi mind tricks to get me to buy this thing that I don't need and don't want. That's not what it really should be.

SM: Well think about the technologies now that are people pushing, all right. They're pushing things like Outreach, and Salesloft, and these are basically spam bots. The goal is I do an email with you or call, you say don't call me, call me back in three weeks. What I do is I put you on a campaign that will blast out things to you. So talk about how impersonal that is. I have one good conversation with you, I tell you when I want to talk to you, and in the meantime, what you do is fill up my inbox with a lot of junk that I'm not going to read.

I do this test with everybody when they try to understand what I mean by personalizing the sale. I ask them to go through their inbox and show me spam. Most the time they can do it in like 30 seconds, they go through their entire inbox they're like delete, delete, delete, delete, delete. They've seen nothing. They haven't read it. You could just spot it. And by doing that so quickly what it tells you is, you've become low value in the eyes of the buyer, you're not actually trying to figure out their problems, you're just trying to bully them into buying it.

I saw that happen in the marketing world. Spam was actually an advertising model for a while. People did say, yeah, it may be horrible but I'm getting a great conversion rate. If someone said that to you today, you'd be like, oh my god, what's wrong with you people? Because people want such highly targeted information that influencers have become the single biggest thing because that person knows me. And they want someone who's just like them to tell them what to do. And imagine now spamming people. It's literally like we're learning the lessons of 20 years ago in sales and everyone wants highly personal. So that's where we think the world is going.

RM: So, let's move more broadly towards AI. You've done a lot of investing. You've seen AI from the work that you guys do. What are people missing? Or, what excites you about what's coming? You saw the advent of the internet and you saw what was going to come. Now it's 20 years and all of that's played out and it's as we were talking about before the podcast, very exciting. Where's AI going to be? And what's exciting for you when you think about a decade from now, 15 years from now?

SM: I'll first answer that from, if I were an executive and then an entrepreneur and then in the engineering world and in that process. If I'm an executive, what excites me is there are very few things I can do that will give me such an advantage in the market that I can still share. Because my products have a life cycle, my customers can only adopt them so quickly. But if I can do something that gives my customers a completely different experience, or such efficiency that I can actually lower my price or do more together. These are things that can give me an advantage to get share, where it seems to my competitor like we've gone crazy. Oh, they're buying market share.

That's what you want as an executive. You want your competitor thinking, they're buying market share, but what it really is you've figured out an innovation that they just haven't clued into. Because those people hang out there for two years arguing that you're going to go under at any moment. This was Barnes & Noble, Amazon's crazy. They're going to go out of business. Well, here they are later and I think Barnes & Noble's probably thinking themselves, yeah, OK, I made a mistake. And so I love those types of opportunities because the early adopters get such benefits that the middle and late adopters can never get caught up. So I love those markets because those are markets where I don't need to get a lot of customers to really dominate a market and then everyone else has to follow. So that, as an executive, that would make me thrilled. So I think that's it.

Now the flipside is, I don't want to be the middle or late adopter. Because if I wait, and I don't learn how to use this technology, I am trying to figure out how to catch up on the heels of it but I'm not getting ahead, I'm trying to get back to par. And that means I'd never get my market share back. So that's Nokia trying to come up with their touchscreen device, after Apple had completely eviscerated their business. And you might argue a flat screen that's touchscreen isn't that big of a deal, except there's no Nokia left, that matters. So these are things that you want to pay attention.

As a founder, I love businesses where it's so disruptive that the entire industry's value chain that came before you can be gone in five years and the client is grateful. And I've seen nothing in any of the things I've looked at. And I invest in a lot of different things, but I've never seen a technology where the value of what's created is so dramatic that almost nothing can come close to it.

It's funny, all the five-year goals were beat at year three. So, people ask all the time is there an AI hype cycle when I'm on these panels. And I always say for a hype cycle to exist, you have to have failed. We've beaten every goal that we set for ourselves, and even the bigger stretch goals keep getting beaten. So it just feels like this is an area where you've got that.

The last is as an engineer. I think it's a very different skill set to work in AI and I think it's almost counter to the old engineering, so this is the time to learn early.

RM: A lot of people don't know this but you have a legal background. And so based on your understanding of AI and your understand on the legal system, are we setting ourself up for some problems? Is the legal system ready to handle this or not? What do you think about that?

SM: So, that depends on where you live. In the US, historically, we've been very good at not regulating business out of existence. Usually the US takes an approach of, let them go, if something really bad happens address it. And that's why we tend not to have a lot of monopolies control anything, compared to Europe. Europe's historically taken a very forward looking approach, which I would not call forward. It just more means before they know the new trade offs, they're trying to decide what trade offs they want. So if Facebook had existed in Europe, it would not exist. Because the idea when Facebook folks first sprang out, that we would be giving everything away and losing our privacy, they would have gone nuts. The US people made a new trade. So, I worry a lot about that.

I also worry that data is becoming the new farmland subsidy. In other words, in the 70s and 80s, Europe would build these barriers to protect their local farmers. And they created all the things like, the seeds had to have this or you couldn't do this or that, which basically made it impossible for America or any other country to sell their produce into Europe. I feel like they're doing that today with data because there are no major European companies that are on the forefront of this.

What ends up happening is they're starting to say data has to be local. All right, fine. Then you can only do data with this certain things, we have to get approval. People forget, in the direct marketing world in the early days of Germany, if you wanted to use a credit card you had to go to the post office and sign an authorization per catalog to allow them to take your credit card. These were things that were crazy, but I think you're going to see a little bit of that over the next five years, mostly as tax mechanisms. GDPR being a great example of a tax. It's not really a law, they just want to find a way to get money. That's my two cents.

RM: Interesting. Are there any technology problems that you guys aren't working on at Collective[i] that you think, I wish somebody would solve this problem and it would even help our business tangentially?

SM: I mean, look, obviously there's other simple things like labeling, we'd obviously like to find a more effective solution. That's painful. It doesn't add a ton of value to the end user, except for the benefits of what they get, but they don't see you doing that work. In fact, the number one question we get is can you take this data, put it back in our CRM? You sort of look at them you're like OK, it's a very fair question. I know why they want it. But you're trying to explain, OK, I built a jet engine and you're asking me to put it on a wooden bicycle. I don't know what to do.

I think when I look at the tech, I think that the challenges are you're so focused on the logistics of how to get the machine to operate well, that sometimes it's disconnected from the use cases. Because you're so trying to get the data into a usable format, that oftentimes the use case gets lost in getting your stuff into a place where it works well. So I think that sometimes can create more problems. But I think as the logistics of getting the data in usable format go away, that goes away.

RM: So, last question. We talked a little bit about this debate that happened sort of at the end of last year, that I call the: Gary Marcus versus everybody else debate, where Gary Marcus sort of said deep learning is not going to get us there. There's more things that have to happen to get to reasoning and thinking and true intelligent machines. And then there's a lot of people that think, no,no,no, deep learning's got a long way to go. You'd be surprised if you give it time, et cetera. I say Gary Marcus vs. everyone else, there's actually a lot of people on Marcus' side. But what do you think about that?

SM: It depends on the question you're asking. So, can society handle that change that fast? I worry more that, forget the tech itself. Think about what we have to explain to people today, something that's actually not that hard to figure out. Think of it as simple as the machine is learning from trial and error. If I could explain as simply as that, people would be like, oh, OK. But the fear around what does this mean and what if it's wrong? Well you hire people all the day all day long who make bad decisions and yet you're comfortable. Half my team can't explain to me why they did what they did and I'm OK with it. Because you get comfortable with it, you've used technology helps you this way.

But, there's this period of time I think where we're solving a lot of big problems with what we've got. And even as it gets better, it's just going to ease people into that moment where you do get to a place where, you're right, probably deep learning-- I actually agree. I don't think deep learning gets us all the way there. But I think, for where we're at now, I actually feared what getting all the way there could mean. And I think there's a good middle ground right now where we're getting a lot done and a lot of value where people are getting used to it.

RM: Good answer. All right, that's Stephen Messer. Thank you for being on the program. If you want to learn more, you can check out Collective[i].com. If you have guests you'd like to see on the podcast or questions you'd like us to ask, please in those to podcast@talla.com. Thanks for listening, and we'll see you next week.

Subscribe to AI at Work on iTunes, Google Play or Spotify and share with your network! If you have feedback or questions, we'd love to hear from you at podcast@talla.com or tweet at us @talllainc.