Automate Your Job With Generative AI — Fact vs. Fiction? (Guest: Eric Fraser)

Andreas Welsch: 0:21

Today, we'll talk about automating your job with generative AI. And who better to talk to about it than someone who's doing just that. Eric Fraser. Hey, Eric. Thank you so much for joining.

Eric Fraser: 0:32

Thanks Andreas. It's great to be here.

Andreas Welsch: 0:35

Awesome. Hey Eric. Why don't you tell our audience a little bit about yourself, who you are and what you do?

Eric Fraser: 0:41

Yes, thanks. So I'm currently the Executive Vice President of Revenue Operations and Technology at a consulting company with about a hundred people in it. And I'm also currently playing the role of sales manager. That's the role that I went to my boss and said, Hey, I bet I could get an AI to do this whole thing. But, I was wrong by the way, but that's what I started off with as a premise and a hypothesis, and my boss, because he's really supportive, he said I bet you can't, but I love the curiosity and I wanna see what you can do. So go for it.

Andreas Welsch: 1:18

That's awesome. I'm, gonna say I'm super excited to hear more about you and your journey. I know we've been connected for quite some time, actually over LinkedIn. So I was really intrigued when I saw your post a couple weeks ago that's what you're doing. And so I wanted to invite you to the show and also for you to share with the rest of the audience what you're seeing. What is actually realistic and where maybe some roadblocks that you're hitting?

Eric Fraser: 1:43

Yeah, absolutely. Overall, the reason that I was wrong and that I can't just automate my entire job with AI is that the role of sales manager has too much human to human nuance and connection that's needed to be really good at it. Both conversations inside the company and conversations with prospects and customers. And I haven't found a generative AI like the Type that you see when you use ChatGPT. I haven't found one of those that's just good enough at being a sales manager. So this is a challenge with the current large language models like ChatGPT, is that they are broad. They are somewhat shallow. So you can ask ChatGPT about a huge number of things. But when you start getting really focused on what should I do about this deal though? Should I increase the price? Should I reduce the price? Should I offer somewhat sort of incentive? That's where it starts to become not so smart and not so helpful.

Andreas Welsch: 2:41

I see.

Eric Fraser: 2:41

So we don't have any large language models that are specifically for sales management, although I do think that some of them are being built by some of the large CRM vendors. But they're just not there yet and they're not stable enough. They're not deep enough for my boss and, my CFO to be happy talking to them instead of me. So if I said to my CFO, Hey take these questions and ask them to this large language model here, he would be incredibly frustrated. He'd say, look, I tried that and it just gave me some really dumb answers. I want to talk to you. You give me these answers.

Andreas Welsch: 3:17

I know we have a few questions lined up and maybe a quick question to the audience. We've already jumped into the topic. So I'm glad you you're providing some context here. But for those of you in, the audience, I'm really curious where you're joining us from today. I already see in, the chat folks joining from India, from Austria, from Nigeria, from Greece, Cameroon. I'm always blown away seeing in which parts of the world people are watching. What's the BUZZ? So please let us know where you're joining from. Super excited to have all of you with us today.

Eric Fraser: 3:52

I love the global nature of your audiences. I've been in the audience several times and I really like the range that we can reach here.

Andreas Welsch: 4:00

Yeah, so that's fantastic. Hey, Eric, to stay with the theme. Should we play a little game to kick things off? What do you think?

Eric Fraser: 4:08

Let's do it. Absolutely.

Andreas Welsch: 4:09

Alright, so let's do this. So this game is called In Your Own Words. When I hit the buzzer, the wheels will start spinning. And when they stop, you'll see a sentence. And I'd like for you to complete that sentence with the first thing that comes to mind and why in your own words. To make it a little more interesting, you'll only have 60 seconds for your answer. Are you ready for What's the BUZZ?

Eric Fraser: 4:34

I'm ready.

Andreas Welsch: 4:35

Okay. If AI were a bird, what would it be?

Eric Fraser: 4:42

Okay. I think if AI were a bird, it would be a condor, an Indian condor. And the reason that bird comes to mind is because they fly at extremely high altitudes. And so they can see a broader set of the world than birds who fly at a lower altitude. And I think that's the potentiality of AI is that it can it can see a perspective that the human brain or a single human brain would find very difficult to accumulate just because of the biological constraints of a human brain.

Andreas Welsch: 5:21

That's awesome. I haven't heard condor yet as an answer. So that's fantastic. We've had people mention Phoenix rising from the ashes. The majestic nature of a of eagles. Now condor with the high altitude view. That's amazing. So maybe from high altitude to our topic. You already mentioned a few things where you're seeing some of the limitations. But I'm wondering you said, Hey, I want to consciously and purposefully see how far I can push the boundaries of automating my sales manager role. And I know people are usually afraid of losing their job, especially to a machine. So why would you actively try to automate your job?

Eric Fraser: 6:09

Sure. First of all, I have to acknowledge that I work in a culture where I don't have to be necessarily afraid. If I find a way to automate something that I do, no one's gonna come and say, oh you're fired. So I have that privilege and I'll acknowledge that. But also, even more broadly. Frankly, there are just some parts of my role that are not full of joy. Some of the traditional sales manager role is pulling numbers out of software, putting them in a spreadsheet, doing a bunch of spreadsheeting, and then sending that spreadsheet to several other people. I personally don't get a lot of joy out of doing that, and that is absolutely AI-able. Not with ChatGPT, not with Bard, not with Claude, but with more old fashioned AI. And that's what I was able to do. I was able to take a lot of the number crunching and number pulling and pushing aspects of my job and just AI it and just either shrink it or zero it.

Andreas Welsch: 7:07

That's, an interesting perspective, right? I think there's a lot of promise of what you can do or what you're supposed to be able to do. But really putting it to the test and seeing what is real is great. So for those of the in the audience, if you have a question that you'd like to ask Eric, feel free to put it in the chat and we'll take a look in a minute or two and take some of the questions here as well. Now you said some of the number crunching and analysis is what you've been focusing on and seeing some success for sales management. I'm wondering what other tasks are you seeing that you're able to automate or that you're not able to automate? And maybe you've already alluded to it in the beginning.

Eric Fraser: 7:48

So, there's a part of it that I'll just generally call pattern recognition. This is about 75% AI-able. I'm sticking a very broad number on there, but pattern recognition in sales management is looking at what's happening to your deals and understanding broad patterns: oh, if we don't get a CFO from the buyer's side involved by at least stage three, we're always gonna get problems with pricing and discounting in stage four. So that's a pattern that if you are paying attention to your deals, you will probably see it as a human. An AI will see those patterns faster than you and maybe better than you. Maybe it'll see some patterns that you don't even see, regardless of how much you're paying attention, because AI is a way better at reading numbers than humans are. If you've got good data or you're able to capture good data, then you can trust an AI to show you patterns that you would have taken a lot longer to find yourself, or that maybe you would never have found. Now, the reason it can't be a hundred percent delegated to the AI is because there's still some human interpretation of the pattern. So let's say the AI tells you, oh, when you sell to this industry, you always single thread yourself, meaning you don't talk to enough people in the account, you only talk to, I don't know, like the chief nursing officer or something of a hospital instead of the CFO and the president. So it might tell you that pattern exists, but what do you do about it? So usually there's a smart human that has that answer. AIs right now don't yet have a great answer for what to do about single threading on large accounts.

Andreas Welsch: 9:21

So you mentioned something about data. If you have data, if you have good data, if you have clean data, I think if you're in the space of AI or data itself, you know that's hardly ever the case, right? That's what we all wish for and dream about. But how do you go about getting data and getting data that is useful for the purpose that you needed for?

Eric Fraser: 9:44

Yeah, this was one of the early points of challenge for me, like when I realized how incomplete our CRM data was, and I don't think we're particularly bad. I think we're average as far as CRM users go. I looked at it and I compared it to reality and I thought, this is not a high fidelity picture of reality. This is probably not gonna work. But what I was really impressed by is if you take certain AI tools and ask them to collect data, you can fill in the gaps of the picture in very surprising ways. So this is where, at first I got discouraged early on, and then I got super encouraged when I just pushed the boundaries a little bit and asked the AI to do a little bit more. See if you can interpolate the picture for me. And it just really was a lot better than I thought. So even people on this thread who might be thinking, oh, our CRM data's terrible. This will never work for us. There are technologies available now that will make your data a little better than you think. I'd say I was also discouraged, and then I got encouraged when I just pushed a little further.

Andreas Welsch: 10:44

One of the questions that I see here in, the chat from Justin is: do you find the efficiency gain to be worth the learning curve of experimenting with and learning how to use the tools?

Eric Fraser: 10:55

So my personal situation is that I like learning about AI. Now, if you don't like learning about AI, like if you find math really horrible, it may not be worth the squeeze. I actually enjoy it. Sometimes in my spare time I just like listening to podcasts about AI math. So maybe I'm a little unusual in that respect. But it's not just the efficiencies of what I don't have to do anymore. It's all of the ways it has this kind of unforeseen benefits effect. So it's like the space program where they, discover something and at first they think that's a nice discovery. We have no idea how this is gonna be useful. But then a year later they realize, oh my goodness, we just totally transformed laser technology and suddenly it's used everywhere and all over the world. So there's a little bit of that to it as well, where at first you think, oh, I'm gonna save myself one hour of spreadsheeting, but then it turns into something way bigger than that.

Andreas Welsch: 11:56

I think that's especially interesting. Like you said if, you're curious and you like to tinker around with it anyways and see what's available, how can you make it part of your role and also see for yourself? Maybe what are some things that you haven't considered yet, but that are new tools that are available and so on? I think there's another question here. Savant was asking, with OpenAI now recently releasing ChatGPT for enterprise. How do you see something like that either influence your role as a sales manager or the roles of sales managers in general?

Eric Fraser: 12:28

Directionally it's gonna have a huge influence. So first of all the, three things that it solved when it, the big things security, privacy and encryption. Privacy meaning you don't necessarily have to give your data to OpenAI now for it to train its models. That will just result in a much broader adoption of these types of tools. So now that they've done it, everyone's gonna have to meet that standard. And so you'll have this opt-in where, look, I choose to give you my data or maybe I don't. And companies want that choice. So that'll result in a much broader adoption within enterprises. But also where it's going is the business use, not just the consumer use, but the business use of generative AI. And so you also have in the same thread of direction, you also have where McKinsey recently paid a company to make a McKinsey specific large language model. And that sort of effort is going to result in roles like mine being transformed. Because if you start to have large language models that are very specific in a particular area, let's say you have a large language model that is really good at diagnosing the types of ailments that people go to their family GP about, then you can start to also make a large language model that's good at sales management. So there will be this development curve where steps like releasing enterprise versions of ChatGPT and what McKinsey did, they'll push us towards a world where the large language models get better and better at the things we ask them to get better at.

Andreas Welsch: 14:18

I really like where you're going with this and that there is a journey, there's a trajectory where we're starting at one point right now. And there's a lot of manual work still involved in either doing your prompt engineering, doing your fine tuning, adjusting it to what you need, connecting it to data sources and so on. And I believe, too, over the next couple quarters, I think we were talking about this earlier, maybe the next year or two we'll see this mature even further and become more and more user friendly. Yeah, I think that's where it gets really exciting.

Eric Fraser: 14:49

It is. And I think there's also this element where, right now there's a lot of human effort needed to train a large language model, but people have already started to think what if we could get AI to train AI? So people who are interested in that should check out a company called Anthropic. They have a design where you can refine and improve the training data set, which is the data set that trains a large language model, by using AI. So I would check that out. And that's also I think a direction that AI is going is that we'll try to reduce the the human, the low level human energy that's needed to train data sets. In other words you don't have like warehouses full of people in a third world country furiously tagging images saying, that's a dog, that's a cat, that's a car.

Andreas Welsch: 15:40

Yes. I think that was in the news earlier this year by TIME Magazine and others. So definitely also the toll that it takes on mental health. It's something where if there's more automation and quality that they can be addressed differently. We've talked about journey and we're at the beginning still, obviously, at an inflection point of using generative AI and AI as a whole and where this might go. But talk about journey. What's been your journey like so far? In addition to keeping an open mind and seeing what else is available, what can you use, how have you gone about it?

Eric Fraser: 16:17

Yeah other than the fact that I found out I was wrong and that I couldn't AI my entire job, but I could AI parts of it, the other big lesson that was humbling for me was to realize that I'd completely underestimated the cultural or the human experience effect of introducing even a little bit of AI into the company. So when I started introducing little bits of AI, I thought I'm only really doing this for my team, the revenue team, so other teams in the company won't care. But in fact what happened was they did care a lot and some of them were alarmed, right? They thought, what are you doing over there? Are you trying to get rid of our jobs? Luckily, it was fairly quick for me to adjust to that.'cause we only have a small number of people in our company. But I was thinking if I was doing this in a 20,000 person company, the ripple effect of that concern might get quite serious and hard to undo. So the lesson that I took is if I ever was doing this in a larger company, I'd start with some reassuring communication, clarifying, this is exactly what I'm doing and this is why I'm doing it, and here's what it will probably do, and here's what it certainly won't do. And course being communicative I think is super important. When you do introductions of AI, people are concerned about things like will AI take my job? So if you don't say anything, they tend to assume the worst. So we have people in finance concerned about something that I thought, oh, why are they concerned? This has nothing to do with them. They were concerned because I missed the step where I was supposed to communicate to everyone. Here's what I'm doing and why.

Andreas Welsch: 17:49

I think that's a great point. And if I think back to previous roles in my career, especially in IT where we're going through some outsourcing, there were similar concerns, right? Even, if management assured you, nothing is going to happen. We're, outsourcing parts of the, roles to a different provider, but that gives you more time to work on other things in a similar way. If we're now having this conversation about AI, I can also clearly see that there is concern. And does management just tell us that's what they're doing? Are they really doing it? So I think that the part of around trust and obviously having a, trusted relationship and culture to begin with. And then to your point, to communicate openly and transparently and also show that you're for real through the actions that you think that's very important.

Eric Fraser: 18:37

And if you're trying to do it for efficiency, for example, if you are introducing AI because you think, oh, I can save a million dollars by removing some labor cost. Also consider that if you create the wrong experience for a bunch of people, those experiences might lead to negative beliefs about the company and or their roles in it. And that will definitely affect their behaviors. And those behaviors might start to produce results that completely wipe out the savings that you were trying to get to true.

Andreas Welsch: 19:03

I see a question here from Fred in the chat. Have you actually created that communication now to continue your exploration and journey? Or maybe to phrase it a little differently, how have you adapted your approach and your communication since you've learned about that?

Eric Fraser: 19:17

Yeah, so I've spoken to the teams in particular. So in our organization it's important to speak to the teams. I can also speak to them at a company wide level, but it's more effective in our organization. If I go to one team and say, okay, so here's what I'm doing. It's just for the revenue team, but here's the effect it's gonna have on the revenue team. Even if I'm describing things that don't even touch their lives at work, it's reassuring to them to hear, oh, that's what they're doing over there. Got it. Oh, and that's why they're doing it. Okay. It's a better experience for them than just to know that something's going on over there with AI and I have no idea what it is.

Andreas Welsch: 19:54

So I'm, curious building on that. I hear a lot of leaders in the AI space also talking about creating FOMO. Fear of missing out. So if you are doing this with AI in your revenue operations role, What about your peers? How are they viewing it? Are you indirectly creating some FOMO and they saying, oh, Eric is becoming so much more efficient and effective because of using AI, I want to do the same thing. And are they coming to you and asking you how are you doing this? And what can I do in my role?

Eric Fraser: 20:26

Yes, they are coming to me and to be honest, I hadn't adequately thought about the FOMO effect. It's just a general challenge that I have as a professional and sometimes I'm just not aware enough of the experiences that some of the decisions I take create for people who I think, oh, they don't need to worry about that. They're not even on my team. But I have to keep reminding myself no, when they hear about this sort of stuff, they will have an experience and I need to be conscious of that experience. So maybe there is more FOMO than I realize, but what I'm experiencing myself is that people do come to me and say, show me that tool again. What does that do? Could we use that? And so I remain open to everyone in my company to answer questions about it. A lot of the time, the answer is, No, that's probably not that relevant for you. But here's a tool that might be. So we have a legal and contracts team, for example, that have not yet come to me and asked me about can I use ChatGPT to do legal drafting. I'm gonna have a lot of warnings for them if they ask me that. But I'd be happy to show them like yes and no. You could probably get it to do some simple things, but be super careful because especially in the area of law, it's actually done some screwy stuff.

Andreas Welsch: 21:36

We've seen that just a couple months ago in the news, right? When when users attribute more capabilities to the AI than it can actually deliver. Or if they're not aware of all the boundary conditions and limitations. But I'm curious in general, what's your recommendation? How can leaders in different parts of the business actually embrace the AI driven change and do that responsibly?

Eric Fraser: 22:00

Yeah. I would start with the discipline of writing down what you get asked to do, and just humanly observe the patterns of what people ask you to do all the time. So when you notice that several times a day you get asked a question, let's say in a sales management role okay, so what are the 10 largest deals that we absolutely can guarantee are gonna come in? Of course no deal is guaranteed, but you get asked this question a lot. What can we absolutely rely on that will come in this quarter? So if you get asked that question all the time there's a good candidate for applying some sort of AI or some other sort of automation and you could save yourself some time and improve the answer. So I would start with just becoming aware of what is it that you're being asked to do. Get fairly disciplined about being aware of how much time do you spend on certain things and where are the patterns and where would you like to spend less time? I would like to spend less time spreadsheeting, for example. What do you get asked all the time? I get asked all the time about what can we rely on to come in a quarter? And how can you maybe improve an answer that you think is a bit soft. Like sometimes I tell people, I think we can reliably rely on these deals and that myself, I'm thinking one of those deals, I'm not even sure of myself. I think there's a candidate where if I apply some AI to look at the signals that is happening in that deal, maybe the AI can say, oh, you're dreaming, man. Like that is not gonna come in.

Andreas Welsch: 23:30

Thank you for sharing that, Eric. We're getting close to the end of the show. So I'm wondering if you can summarize the three key takeaways for our audience today, especially, if you can focus on automating your job. Is it fact? Is it fiction?

Eric Fraser: 23:42

Yeah, I would say in my job it's greatly fact. But I was over optimistic about how much of it could be automated. I would say be careful of the hype around generative AI and look for ways to use older and less fashionable types of AI. There are some elements of machine learning, for example, that are not generative AI, but are super useful in lots and lots of professional roles. So I would look in the vault of older AI. And pick through that and see if that can help you. Just because generative AI can't do it yet doesn't mean that there isn't some other AI that could do it really well. And the third one and, maybe the one that I got the most wrong is just to be more aware of the human impact and the experiences that you are creating when you decide to use AI to do anything. Even if you think, oh, no one else will care about this is just for me. No, I guarantee you someone's caring about it and you need to reach that person and create the right experiences for them.

Andreas Welsch: 24:47

Thank you for sharing. Especially that last part around becoming more conscious of the communication and the impact on others as you're embarking on this definitely sounds like it's an exploration worth spending time on and seeing how far you can actually push the boundaries.

Eric Fraser: 25:04

Absolutely.

Andreas Welsch: 25:05

Yeah. And one comment that I do want to pick up before we wrap up. Martin says basically, Hey, maybe you don't even need AI. Maybe something like robotic process automation can do the trick for some of the things as well.

Eric Fraser: 25:20

I think so. And that actually that's one of the older types of AI that I think is worth looking at. So there's lots of different types of AI that is, that are not generative AI. I would absolutely check them out. Yeah.

Andreas Welsch: 25:36

Fantastic. Alright, thank you so much for joining us and for sharing your expertise with us. Eric, really appreciate having you on the show and hearing your experience.

Eric Fraser: 25:45

Happy to be here and it was a pleasure. Cheers everyone.