How AI Changes Marketing and Business

Session Description: A panel discussion of the intersection of AI, marketing, and business.
The transcript for the session can be found below, with a PDF version available here.

How AI Changes Marketing and Business

(Transcription by RA Fisher Ink)

Hempel: I’m Jessi Hempel from Wired. It’s great to see you all here at Techonomy. It’s about my favorite event every year and I would like to introduce our panelists. We have Heidi Messer from Collective[i] and Rudina Seseri, who is with Glasswing Ventures. Heidi and Rudina.
[APPLAUSE]
Hempel: So in the 15 years that I have covered technology, I don’t think I’ve ever gotten so many headlines about a topic until AI came on the scene. It makes it into every press release and every pitch. And from those press releases I can tell you two things. AI is going to change every single thing and also you should be terrified of it.
[LAUGHTER]
Hempel: So we’re going to break that down a little bit for you and we’re going to try to separate the hype from what is actually happening. And that’s why David has done a great job of inviting two people who really have the hand on the lever of the future of the technology to break it down. So Rudina is investing; she’s seeing a lot of things. Heidi has been working on one project for a good long time and so I want to start with Heidi and I want you to tell these guys a little bit about your startup.
Messer: Sure.
Hempel: We can still call it a startup, right? Because you guys are not small.
Messer: I think it might be beyond that. So the company’s called Collective[i], which is short for collective intelligence. And the area that we’re focused on is the sales function of enterprises and SMBs and bringing them an application that’s based off of a network of data that uses machine learning and deep learning to basically guide salespeople to their most high value activities. So today, most salespeople spend 70 percent of their time on non-revenue-producing tasks and after Collective[i] that equation gets flipped and they spend more of their time selling and generating revenue for their companies.
Hempel: So I want to dig into that a little bit. So those nonrevenue-producing tasks. Say I am a typical saleswoman on a typical sales team. What am I doing with my time, typically?
Messer: So one, you’re doing a lot of guessing. So you’re guessing who I should speak to at a specific prospect, you’re guessing which opportunities to focus on; you’re spending a lot of time in meetings, telling your manager what you did last week, and then you’re also typing it into things like CRM. And basically doing manual research about every single person that you talk to all day. So by the time you actually get to talk to a prospect, you’ve had to talk to a manager, talk to a machine, you’ve had to do Google searches, LinkedIn searches—it’s very, very inefficient.
Hempel: And with your software?
Messer: So with our application, you are actually told who are the decision-makers you should be speaking to, which opportunities have the highest value odds, and what the likely outcome of your activities are and the impacts every day. You’re also able to collaborate in real-time with folks on your time around scientific insights. So the relationship between you and your manager isn’t, did you lie in what you put into CRM? Here are the real odds, now what can we do to improve them?
Hempel: And so that, let’s call it a black box, that provides that information, that comes from data, real data that companies give you that you, in turn, decipher and feed back to them. And so the question that I want to start with is why would companies want to give you that data?
Messer: So, I mean, you talked about AI being sort of overhyped or you hear a lot about it and I actually think real AI is almost underhyped. I think you can’t underestimate the impact that artificial intelligence will have on enterprises. The reality is—and this is getting to the answer to your question, “why would companies share data?” The only companies that I believe that can do real AI, as opposed to fake AI, which is the term you like, are the ones that have real networks of data that cross multiple industries and multiple companies. So one of the questions I would ask if I were looking at applications that could help my company be smarter and my employees be smarter are what kind of a graph have you built? In our world, Collective[i] has built the largest sales graph that’s out there by pooling data across different enterprises. You cannot do that with individualized data sets. So no company on their own, I don’t care how big they are, can generate meaningful AI, meaningful insights off of an isolated data set for functions that involve external parties.
Hempel: I love the term fake AI. It’s better than fake news. We’re going to come back to it in a few. But Rudina, I want to bring you in here. You are out in the world, looking at opportunities, AI-based businesses. Are you seeing a lot of fake AI?
Seseri: So the good news, when you do AI investing—my firm is an early-stage firm and we’re the people who ride that first $100,000 to $3 million check—more along the $3 million than the $100K, but we’ll do both. When you do this for a living, it’s pretty quick to sniff out the noise from the real substance. And I would look at it in the following manner, which is AI has been around for a long time. I mean, there used to be a term called GOFAI, right? Good old-fashioned AI. Fundamentally, the reason why it’s sexy again only to surpassed by blockchain, thank God—
[LAUGHTER]
Hempel: I get a lot of pitches for blockchain too.
Seseri: It’s fundamentally because two things have happened. In academia and then later on in industry, we saw the emergence of deep learning, which is only one approach around machine learning, but that layer of the ability to combine algorithms of different kinds and different neural nets, from the technique point of view, created a new wave and new, improving outcomes. And then fundamentally, the data piece. We have always had data, we have never had data that is richer, more diverse, touches all senses, and in larger amounts. And I’ll dare to venture and to say that we’ve only scratched the surface. So with those two pieces in place, whether you’re directing, in my case, investments in startups and the enterprise, cybersecurity platforms, robotics or otherwise, there’s something that you can really have an impact and you can sniff that out fairly quickly if you do it for a living.
Hempel: I would imagine that you can. So tell us what you’re seeing. I mean, where are you seeing AI have an impact on business today? Not opportunities you want to see develop but the stuff that’s already happening.
Seseri: Again, I vote with my dollars so I’d better believe that it’s happening today. So I’ll give an example. Tala is a portfolio company of ours and they’re basically a platform that’s focused on creating an intelligence layer around enterprise so when departments are communicating with each other, how the heck does that information get captured and then turned into action? So that’s a platform play and then the hooks are around IT support and HR. So if you have an issue and you call IT in a large enterprise, what happens? You get a ticket and you wait for hours on end. Well, how about getting a bot or some sort of level of intelligent agent to—and you’ll hear a term that Heidi used—automate portions of it so that only the more difficult tasks can be done by humans.
And that’s an important theme. The pop culture paranoia around AI, this is not applicable in this moment in time, not that we don’t need to be thoughtful with narrow AI but what AI can do today is quite limited, let’s just be honest about that. And so it has to be narrow in purpose and highly effective. So we are in a paradigm, in my view, of augmentation rather than complete automation. We augment humans or ourselves in the lower productivity, mundane tasks to make more room for the creative or what makes us as a species I think more interesting.
Messer: And I think just to build on that, I think the fear that you mentioned around AI is very misplaced. Often the fear is around how the data is collected when the fear I think at least from the enterprise should be on what you’re missing out on. We were talking about this before—unlike the internet, where you could see when your competitor launched a website and you could kind of track their performance and say where do I stand relative to them, AI is silent. It’s happening on the back end so, you know, why Amazon is so frightening to a lot of companies as a competitor is that they’ve used data strategically for everything that they do and when they launch something, they’re typically so far ahead of the person behind them who hasn’t been doing that level of analysis and doesn’t have access to that level of a graph of data, that’s what makes it more frightening, I think.
Seseri: And I’ll pick up on that thought, to the point of how powerful data with two cases, one being, you know, you think Google thinks it knows who we are and what we want to? Amazon has so much data and is so good at it that they know exactly what we are doing oftentimes before we’ve actually been able to formulate it our own selves.
Hempel: So Rudina and Heidi, you’re both talking about big companies here. And we call them the Frightful Five—you know, Google, Facebook, Amazon, Microsoft, and Apple. I almost missed Apple there—telling. Is there room beyond those five, given that data is so much of the raw material that powers the real advances in AI, for startups to change business?
Seseri: I fundamentally believe yes and this is why. For those of you who have been in tech in some time, remember big data? Who’s talking about big data five, seven years later? It was de minimis data. I think what has driven this current wave and powered it is it manifests in data as the barrier to entry and differentiator but it’s been the connectivity piece and intelligent piece in the diversity of devices we’re using. The fundamental points made over lunch notwithstanding, we’ve gone from smartphones to everything being connected, from your toothbrush to your TV to your Fitbit to your Apple Watch to devices that we have in experiences that we have yet to imagine. Healthcare, think about all that’s coming up, even when you’re not taking an action; data is being generated. Privacy’s a very, very important point so just because I’m not explicitly speaking about it; it’s not a foregone conclusion. But to answer your question, there is a huge amount of data that the Five control, there is a lot more that’s about to be generated. So the landscape will change.
Messer: And I think the five companies that you mention have done very little in the way of business. So how people work is dramatically different than how they live. So if you think about it, you know, I don’t have my phone with me right now but all of my life is captured in the applications that I use on my phone. I go to work and I open up an Excel spreadsheet from 1980; I do an in-person meeting—you know, things are very, very antiquated and that’s because a lot of the graphs of data that would be built around that—LinkedIn being a notable exception and Collective[i] on the sales side—haven’t been built. And so I think there’s a lot of room for companies to start up and create the graphs of data that are relevant specifically to how people work.
Hempel: So Heidi, you’ve been working at Collective[i] for a long time now. I remember our first conversation about it. When you began, did you envision that you would be here? Did you know where you were going and that it would look like this?
Messer: We did and the thing about building networks as opposed to building a pure SaaS company is that networks are just a lot harder to build and not only that, when you build them for the enterprise, it’s a lot harder to build than for consumers. The consumers have a very easy decision-making process, like, “Do I want to download this or not?” Yes or no, that’s the decision-making process. There’s not a procurement group, there’s not a security group, there’s not seven to ten purchasers on the other side evaluating it. So we knew that this was going to be a brick-by-brick build and we knew the area that we were focused on and, to Rudina’s point, focusing on a very specific function within the enterprise was important to us because we felt like that was how we could have impact.
Hempel: I’m curious, where are we realistically on the road to real AI or a general intelligence, insofar as it helps business? Because I hear you say, Rudina, we’re not that far—narrow, generalized applications. I also watched Google yesterday introduce Duplex. Suddenly I was watching voices have conversations with each other that were computerized voices that sounded fairly definitively like humans and I thought, gosh, how do I really understand where we are? Because that actually feels almost like science fiction.
Seseri: Yes and by the way, you’ve seen that in pop media as well and more general media around images and what can be done, you know, who’s a real person, who’s a computer-generated image so you see that in voice, you see that in image—it’s still narrow AI. Let’s be clear about it. Artificial intelligence is not anywhere at the level of generalization abilities that our brains are so I still think that you could combine them together. The analogy I would give you is like we go into our phones, one app in, one app out. Well, if you didn’t have to do that and the time was shortened, you’d feel like you’re interacting almost in a human-like manner but you’re still in a very, very narrow purpose for generalization.
The point, if I may deviate slightly from your question, what we’re seeing that’s different about AI and particularly the impact it will have on tech companies is the ramp up time is a bit longer and the founder and business dynamics are a bit different. What I mean by that is you’re not just coding software, trying to solve a problem, you QA it and there you go in beta and now go get those first enterprise customers. Here you’ve got to train your algorithms. You have to have the right data. It’s not just about massive amounts of data because the neural nets are still quite hungry. It’s about the right kind of data. Otherwise, junk in, junk out. That takes time. Overlay on top of that, the people factor. You know, AI be damned, people make or break you still.
Hempel: And by that, you’re talking about the people you hire in your company, not what your technology’s going to do to people.
Seseri: Yes, the people, the humans adopting on the other side. But just before you get to the adoption, just to get the tech out the door, you no longer just have the tech founder and the business founder. You have this third person who’s usually a researcher and historically researchers teach and research; they don’t do and they tend to be perfectionists. So how do you manage the dynamics between running fast, where it’s good enough, and that academic-type mindset? It creates an interesting challenge in this three-legged stool in how you balance growth.
Messer: So where I thought you were going was in terms of the application’s capabilities and then the adoption of it in business and I think the application capabilities are far enough along now that machines are really learning and able to add tremendous value to companies. I think the human adoption side of it, you know, a lot of things change about how people operate and there are two paradigms that exist right now, I think, in most of the companies I talk to. There’s the up-and-coming paradigm, which is shouldn’t there be an app for that which makes my life easier? And then there’s the established paradigm, which is how do I avoid risk at all costs? And the reason I say that AI is tricky in the latter paradigm is because not adopting AI is actually a choice to your detriment. So it’s sort of like saying okay, I’m going to educate a kid and I’m going to give them five books to read their whole life and they’re going to compete with somebody who went all the way through to university and graduate school. You know, the amount of knowledge a machine takes to learn, every minute you’re not doing it, your company is getting less smart. And that paradigm goes against the “Why should I not do something?”
Hempel: It makes me think of what you were talking about earlier as AI as a silent killer.
Messer: It’s the silent killer. I think you will see a lot of very large companies be displaced by companies that are actually smarter because of the applications they’ve given their employees.
Seseri: And that goes back to your earlier question, do startups in the face of the Big Five have a chance of succeeding? It’s very hard to retrofit AI. You can try to do it through acquisition, in certain passes, they’re quite good. You know, some are really good, like Amazon. But in retrofitting AI or retrofitting for data, the initial data sets are incredibly important to the outcome. So if in the past patents were part of your barrier to entry, it’s definitely data.
Hempel: Well, Rudina, you gave a sort of passing nod to privacy. What do we need to consider and keep in mind as we’re designing these systems and moving quickly into this world on the privacy front, in the business sector in particular?
Seseri: So I think one, that data has at the consumer and enterprise level a lot of other parties are squatting on their data and when was the last time we invited someone to squat in our homes? So I think that—
Hempel: What do you mean by that, exactly?
Seseri: So even though it’s data I generate and information about me or of my behavior, Amazon owns it or Facebook owns it or what have you. I think that as a societal and regulatory environment, that’s turning the corner there so privacy is not dead. And if you don’t believe me, look at what happens with kids, Snap took off in part because no one could be mischievous and the information went away so kids are aware of privacy. So to come back to your question, I think the general environment around privacy is changing so there will be abuses I think but more carefully done so that’s one part.
Secondly, AI’s actually part of the solution because if you’re looking at certain behaviors, you can apply AI in cybersecurity and privacy-type environments where there is a dark web to recover your data, whether it is, you know, predicting what data is being stolen, whether it is creating a needle-in-a-haystack problem around your data so that attackers can’t as easily get access to it. So I think it’s a good story. And lastly, regulatory, as I said, with GPDR or if Congress can actually get to ask fundamental questions beyond IT support type questions of Zuckerberg, it would be great.
Messer: I’m just going to add to that. I think we have to be careful that we don’t let fear dictate AI’s future. There’s a lot of benefits that come from data sharing and transparency and I’ll give one example on the cybersecurity front. There are regulations that are proposed now that companies will have to share when they have been hacked. And there are good reasons for that because the faster you can share data amongst companies that even are competitive, the faster you can stop the threat. And it’s actually the protecting of that data and hoarding of that data that stops that benefit from being realized. So when I think about the privacy question and I think about it from a business point of view, it’s more about tradeoffs and less about some blanket right to not have anyone know about you and be forgotten. I think a lot of countries will make large mistakes on the regulatory front if they look at like this right that we’ll protect at all costs.
Seseri: And we’re already willing to give the data, I think, both consumers and enterprises are just looking for more value creation in return, not abstaining from it.
Hempel: Well put. Okay, so we’ve heard about the importance of looking beyond fake AI, taking away narrow applications as the real opportunity, and that we really shouldn’t walk forward in fear or we’re going to miss a lot of opportunity. Thank you all. All right.
Messer: Thank you, Jessi.