Sheikh Shuvo:
Hi, everyone. Welcome to Humans of AI. I'm Sheikh. On this podcast, we learn about the wide diversity of incredible people building and working on the technology that's slowly and quickly changing our world. Today's guest is Natalie Lambert, Founder of GenEdge Consulting, where she works with marketing teams to empower them with generative AI.

Natalie, thank you so much for joining us.

Natalie Lambert:
Thank you so much for having me.

Sheikh Shuvo:
You've had such an interesting career journey, working in very different organizations and on a lot of different topics. Could you tell us your career story and maybe highlight what some of the inflection points were that led to where you are now with GenEdge?

Natalie Lambert:
Sure. I always like to tell the one piece that you'll never see on my LinkedIn, which is that I actually started my career as a ballet dancer.

Sheikh Shuvo:
Ooh, interesting. You're the first ballerina I've talked to.

Natalie Lambert:
There you go. I love it. We are out there. We are sprinkled around the world. But yeah, I started my career as a dancer and, as any professional athlete, that means that you are training all the time.

And the challenge with that type of life is that you are never available when your friends are available. So imagine, starting at eight years old, 12 years old, 15 years old, when I am, when everybody's available, I'm in the studio; when I'm available, people are not. So I needed to come up with some fun hobbies to keep myself not bored in those moments. And my fun fact is I created and I taught myself to code and created a Ben Affleck website. He was my hero as a 12 or 13-year-old. I absolutely loved him. I'm from Boston. And I taught myself to code websites and build a fan site. And when a stress fracture prematurely ended my ballet career, I was trying to figure out what to do with my life, because my entire life I had been training as a dancer. But, in the background, I had this computer thing. And I translated that, building and developing websites, into a computer science degree. So I went to college as a computer science major, and then ended up as an analyst at Forrester Research on the computing infrastructure team, helping the analysts at the company with their research and understanding some of the biggest pain points that were happening in the largest organizations in the world.

And spent time, worked my way up through Forrester, was an analyst in the whole virtualization space for a long time, and decided it was time for me to try something new. And that's when I went to work for one of my favorite clients, which was Citrix. I absolutely loved the team at Citrix. They were just such a great group of people.

And I made that shift from analyst to marketing because when you're an analyst, you spend so much time walking in a customer's shoes, understanding their pain points, and it was just a really great shift to then be able to help build the technologies that my customers at Forrester were leveraging.

So did a whole bunch of different various stints in marketing at Citrix. But the one thing that was something my husband and I always talked about is we moved to the valley for me to start that Citrix job, but I wasn't taking advantage of what the valley had to offer, which was the whole startup community. And so when I had the opportunity to go run marketing at a startup, I felt it was something I just had to do. I had to check that box. And so I went to run marketing, which was such a great experience because I'd always been a product marketer. And so when you're running marketing, you now have to do all of those various other activities. You've got to help build websites, orchestrate a website, do lead gen, do email marketing, social media operations. There are so many different pieces of that. And so it was just such a great opportunity to learn all areas of marketing and did that and we had a really great run and then we were acquired by Citrix. So I didn't want to go back, not saying anything about the company, but just I've done that.

So then I went to go work at a larger startup. And rinse and repeated the same thing. And that one was sold and then had the opportunity to go to Google and go back to doing the work I knew well. The storytelling content and leave the broad look at being at a startup behind.

And it was great to go back to my core because as an analyst you're writing, being able to go back to that was an incredible shift for me. And then this past year I had the opportunity to try something new. They were looking for somebody to come in and figure out how to make Google best in class at using AI to improve our marketing. Like, how do we make Google marketers just the best at using these technologies? And so I got to be based on that breadth of skills that I had learned at the startups and all of the different areas of marketing, start to partner with all of these different teams and areas to figure out how can AI help here?

And had the opportunity to really chase down 20 to 30 use cases that AI could be applicable and figure out where is this technology good for, where is it not as ideal for, where is it saving you money, where is it saving you time, where is it improving performance, where is it improving creativity.

And really chase that down. And it was just such a great, it was such a pinnacle of my career. It was such a great opportunity that I decided that I wanted to help more companies do this. So that's really where GenEdge came from. It was that ability to go back to those analyst days and understand customer pain and research all the technologies, but take what I've been doing at Google for so long and apply that to as many companies as possible.

So it was each of those pieces. I can see how it led me to come here to be able to help companies really improve their effectiveness in the marketing teams.

Sheikh Shuvo:
It seems like a very organic path when you look at it in hindsight. It's very well structured.

Natalie Lambert:
It does. And I will tell you, it didn't feel like that at all. It definitely felt almost squirrel in pieces. But it's funny when you do talk about it, you can see how each piece led to what was to come.

Sheikh Shuvo:
Oh, that's awesome. Now you obviously have a technical background in computer science. But when you started really diving deep into the world of AI, how did you train yourself on the fundamentals of that?

Natalie Lambert:
Just start. I know that sounds ridiculous. But just go in and play. When I started at Google, I hadn't actually used ChatGPT much. I had played with ChatGPT once when I was trying to settle an argument with my husband. So I asked ChatGPT a question, and it came back with some response, and we were laughing at the response, thinking, 'Oh, you know, that's not bad advice.' I'd only ever used it for more personal, joking things. But when it came to applying it to work, I had no clue how to start. I went into ChatGPT and Bard, and I just started asking questions. It started very personal, like, 'Should I do X? How should I learn about Y?' Every time you ask a question, you get a new idea. It was incredible to me how, three or four questions in, I realized, 'Oh, I bet I could apply this to work.'

The job I had at Google, prior to switching to AI, part of that was running our blog. And one of the things we were always questioning was what the right titles are. Whether it be the title of a piece of content, subtitles within, I also had the social team and was creating all the social content for that. It became very clear to me that AI could help us here. Once you start diving in and playing, you will pull that thread, and it will keep taking you to new places, and you'll start to see your work reflected in the answers. And that's how I got started. I just started asking a lot of questions.

Sheikh Shuvo:
Yeah. When you were at Google, you had a very broad and both empowering and potentially overwhelming job to see how to improve the marketing department with AI. When trying to identify pilot projects, what were some of the decision-making criteria you used to prioritize?

Natalie Lambert:
For me, there were a couple of things. It was overwhelming in that there was zero guidance. So, I had to start somewhere. I started with what I knew best. A lot of the projects I worked on initially were with my previous team because I knew every single one of their workflows. So, we quickly started with content, whether it be helping to build blogs or create social posts or create headlines for X, Y, and Z. It was just very easy to get them to help me with anything because they were there. I just started with what I knew and expanded. The next set of projects I worked on, I remember there was an announcement that went out about my role, and the next day one of the product marketers came to my office and said, 'I'm not going to lie. I'm scared to death of AI. It's going to take my job.' So I was like, 'Okay, that's how you feel. You're my next project.' It's like finding these willing participants, whether they think they want to or not want to explore something, and kind of pull that cord. Initially, it was just things that I knew or people that came to me about something, but then I started to really understand what the technologies could do, and I would seek out very specific use cases.

Sheikh Shuvo:
That's awesome. Of use cases, one of the things I recently learned about you is that you're a connoisseur of nachos and recently created what I think is the world's first GPT dedicated to nacho recipes. I actually used it and found some amazing recipes for Nordic-inspired nachos with rye chips, and then Burmese-inspired nachos using chickpea flour and some tamarind sauce, which I'll experiment with this weekend. So firstly, thank you for creating that. But the question I have for you is, obviously, GPTs are super new and super exciting. Could you just tell us about your experiences with building your own GPT and launching it?

Natalie Lambert:
I want to tell you it took me weeks and months to plan this. I want to tell you it was so hard and it took massive skill and it took my love and enthusiasm for nachos to be able to build that. And no one else could do it. I'd be lying through my teeth. The whole thing took about 10 minutes, and it was incredible. So, I have a website called The Inspired Chip, which is something my husband and I played around with during the pandemic. We don't have kids. We were looking for things to do. We made creative nachos and we would just make things up. I think ultimately we came up with 30, 40 different nacho recipes over the pandemic. And I started to write them out. I actually haven't written them all out, but I wrote about a bunch. So, I had a base. One of the reasons The Inspired Chip slowed down in its content creation is because I just couldn't beat the ideas that I had the previous week. When I set that up, I set it up with the knowledge base of The Inspired Chip to give it a sense of, I'm not looking for just different creations of standard nachos, because that's boring. I want Thai-inspired, Chinese-inspired, Mediterranean, I have a bunch. So I gave it that base and told it like I always want the chips to be seasoned a certain way, and all of that. And it was done. It really was the easiest thing. It's been incredible because I've had a bunch of people talking about using the different recipes. I did that for personal reasons, but also in my new role, I wanted to take something that has such a business person perspective and give it to somebody so they can apply it and think about it in their personal life. And in showing that GPT to a few people, I had people come back to me saying that they had immediate ideas. They didn't even know these things were possible and they could now create a GPT to help them with their emails and communication style at work or to help them do something else. And I do think, going back to how I got started, which is asking questions, pulling on a thread, and then finding a work purpose. It is amazing just playing with something and realizing, even if it's a completely different space, you'll just start to think about how you can apply it to your work. And I've seen it over and over again with that nacho app.

Sheikh Shuvo:
It seems all this AI for marketing stuff will be cute, but your real legacy will be nachos.

Natalie Lambert:
I'm ready. If it can write a couple of great recipes that I can add to a cookbook, I will give AI credit and we'll see how it goes.

Sheikh Shuvo:
Awesome. Outside of GPT, you also write extensively about using various AI tools to build your business. Looking at some of the tools that you've recently started using, what are some of the ones that you'd say have had the biggest impact on your workflows?

Natalie Lambert:
I will say, before I talk about constant workflows, that when I decided to go out on my own and build a consulting company, I had the full intention to hire a designer and a web developer to build out a web presence, create logos, and all of that stuff. But then I decided, 'No, you're in AI. There's got to be tools that do them.' And so, using AI to help build a logo, understand my color palette, and build a website, I got that up and running in four days. Meanwhile, I'm talking to companies spending $10,000, $15,000 and taking four to six weeks to get these things up and running. I had an MVP in two to four days. So, if you talk about just how AI can have that immediate impact, that was just incredible. It's a once set and forget thing, and I'm definitely adding more to what I do. But just having that MVP to build on was hugely impactful to the work I do.

Taking it to the day-to-day work, I do a lot of blogging, and I'm starting to get into video, showing in two minutes how to use a tool to do X. In both of those examples, I use AI so much as my collaborator. If I'm writing a blog, I always write my own first drafts. I'm a writer at heart. But I give that blog post to the AI and say, 'If you're a marketing leader, when you read this, what questions do you have for me that I haven't answered here?' Or even the more basic, 'Can you do a copy edit for this and make sure it's all grammatically right, in a professional tone, maybe with a little whimsy?' So, I use AI to do that gut check. I am a team of one, I don't have a team that can help me with all the standard edits and peer feedback that I would get in the past. So, having it as that sounding board has been really helpful. And then taking that content and helping me with social posts. I find that AI uses way too many emojis for my liking, but it can do 80%. So, I feel like when I've written really great content, I can turn it over to those different tools to help me build derivative content and get it out into different channels.

Sheikh Shuvo:
After setting up those initial workflows, it seems like you're leading with trying things out and lots of curiosity. But what are some of the pitfalls or mistakes you've made along the way that you'd recommend a marketer just starting with AI tools to be conscious of?

Natalie Lambert:
I'll say two things, in two very different categories. One is, I think a lot of people hear about AI being able to do a lot of writing. So, I see a lot of people who want to write a blog post on X, and they'll hand it to an LLM and get this blog post in return. They're generally not great if you haven't really thought through what you want to write. It's garbage in, garbage out. Good in, good out. Great input, great output. And quite frankly, the opposite is true. So, people who give up on AI because they've asked it for something and just gotten a really bad result, that's one thing. The second thing is, AI is not perfect. Someone said to me, and I have no idea who said it, but it's like, 'AI is like your intern. They know everything and understand nothing.' It's that idea that they can do all the research in the world and have that answer for you, but they don't understand it might not be right.

Sheikh Shuvo:
That's a great analogy.

Natalie Lambert:
Yeah, it's a great one. One of the things I really emphasize is, whenever you get something back from AI, do not just think, 'Oh, I created it, let's ship it.' You saw what happened with the lawyers in New York; the AI can hallucinate. I've seen a lot of times when it's writing longer content, it'll go off in the most factually incorrect way possible, with complete falsehoods. And if you don't know to look for that, if you're not monitoring it, you won't pay attention. So, that's the second point: always be the human in the loop and pay attention.

The third point I want to call out, because it's very personal to me, is the bias in some of these tools. I've seen bias in language and in pictures. One thing I did was create a bunch of headshot illustrations for a consistent look and feel at an event. We wanted to use AI to bring a single look and feel to the event. I assigned the same prompt to every headshot given to us, and because there were eight male and two female speakers, I did all the men first, and they all came out great. When I started doing the females, there was a glaring problem: the AI had turned them all into men.

Sheikh Shuvo:
Whoa.

Natalie Lambert:
Yeah, it was very, not good fascinating, but fascinating. I was looking at the prompt, and it was because I used the word 'executive'. These LLMs trained on existing information in the world have probably tagged more males with 'exec' than females. So, I redid the prompt and said 'female executive' to make it clear. What popped out next was a bunch of images of women, but now in a full suit and tie. Ultimately, I found using the word 'speaker', 'female speaker', resulted in a woman in a nice dress or blouse. It just goes to show that you should be really paying attention to the biases in these tools. Not only should you be paying attention to details that you might not notice, but have someone else review it, because you might overlook something. Every single use of these tools should have the human in the loop.

Sheikh Shuvo:
That's a powerful story. It seems like you really need to be scientific about how you're constructing your prompts and almost have version control there. But outside of adjusting the prompt, do you recommend any other ways marketers can be conscious of different biases that might come up? Is there an aspect of different biases in different products or fine-tuning the models in a particular way for narrower use cases?

Natalie Lambert:
Especially in the world of writing, I haven't figured out all the fine-tuning in images other than a prompt. With writing, one of the things I do a lot is give it a sample of previous work that I've written, to try to emulate my style. So, if you're shy and reserved, you don't want the output to be overly bold or off-character. It's about fine-tuning, giving it things you've done in the past that you're comfortable with, to show it what you're aiming for.

Sheikh Shuvo:
Be your next post.

Natalie Lambert:
Yeah, totally. That's how I'll pitch this webinar or this podcast, but it's that fine-tuning of giving it examples you're comfortable with to guide it.

Natalie Lambert:
So, I use a lot of examples when I'm creating something in AI to get to a place sooner where it sounds like me, rather than having to go around and around with it. AI is really good at if you get something you like in concept but maybe it's not in your voice. It can rewrite it over and you can make it more professional. I do a lot of taking out the marketing fluff because it writes a lot of that. Using examples is a great way to start getting it to a place you want, but you do have the ability to cut a lot of that out.

Sheikh Shuvo:
Interesting. Now, as a marketer diving into this world, there's obviously an explosion of tools that are competing for my attention. And if I'm in a company, I'm typically working with a broader team. How would you recommend I investigate which tools are worth using? And how do I build the internal business case to use one tool versus another to get team-wide adoption?

Natalie Lambert:
That's a great question. Every company has so many different requirements. For me, there are a million amazing newsletters out there. You can sign up for anything, see them online all the time, and you can follow those daily newsletters that say, 'Here are the top five tools today.' I'd just go create a free account and start playing to see if there was something interesting. There's an AI-for-that.com that keeps track of all the tools. You can say, 'I want something for copywriting,' and it will give you a list of vendors. I would just say go in and play. Some of the interesting ones are those that aggregate the hard work into a single tool. So there are marketing tools, like your writers, Jaspers, INwords, Copy AIs, that take the prompt work out of the equation. For marketers, finding those tools that aggregate so that you're using fewer of them is something I would focus on. If you're going to bring in a tool in a big company, you're not going to bring in 30, you're going to try to consolidate as much as possible.

Sheikh Shuvo:
Along those lines, outside of the cost and time savings implied by those tools, are there any other things you'd recommend emphasizing?

Natalie Lambert:
Yeah, there are four things I look at when it comes to these tools: three criteria with hard numbers behind them and one that's much softer. Cost: how much did it cost you to do this task before and how much does it cost with AI? What's the delta? Time: how long did it take you to do something before and how long does it take now? Performance: what was the performance of the activity before and what is it after? And the fourth one is creativity. It's hard to put a number around, but you can feel it. You might get people to stay on a website longer or more engaged in an event. There's a brand boost when you do something creative, but it's less numerical. So, it's really evaluating each of these and recognizing that it doesn't have to be even. For example, there are activities where I was able to drop cost and time dramatically, performance stayed the same, creativity was the same, that's a no-brainer. Then there are things where the cost might go up, the time might go up, but the creativity took you to a realm that was truly going to change brand perception. So, it's not that each of these has to be even, but it is a matter of looking at what you're looking to build and looking across those four dimensions.

Because even for me, I am a fast writer. I know exactly what I want to say, but now that I started to ask AI to poke holes in this, do X, do this, or have a copy edit, I actually take longer now to do some of that content than I did in the past. But it's a better result. So, don't solely look at it in the true numbers, but as an aggregate of what were the benefits of using AI to accomplish this task versus the old way.

Sheikh Shuvo: Looking back on the types of customers that you have at GenEdge or in your work before, are there any cool customer stories that you can share about the impact things have had?

Natalie Lambert: Yeah, one example, moving away from the standard written content, I was working with a startup who had been at a couple of events and they had done surveys of the people who came to the booth. We took those two Google sheets of the data, asked AI to consolidate them, and then to do a summary of all the data. We asked it to visualize the data and run some simple modeling. At one point, it couldn't do one of the models because we didn't have some demographic questions to build out personas. So, I asked it, 'What questions would we need to ask in the survey in the future to do this modeling?' And it provided us all of the questions.

Then, I sent it back to the client, and they asked, 'Can it write out all the answers for us?' I thought it was an ignorant question, but instead of saying no, I tested it. I asked the AI, 'You've provided us these seven questions, can you give us proposed answers?' And it did, with five multiple-choice answers for each question. It was amazing that by giving it a dataset, having it do that high-level analysis, asking questions we couldn't get answers to, it could then help us revise the survey and give us the questions. It was one of those aha moments where you realize the power of AI.

Sheikh Shuvo:
Sounds like the intern's going to get invited for a full-time gig.

Natalie Lambert:
Yeah, exactly.

Sheikh Shuvo:
Natalie, to close things out, I wanted to bring it back to the very first question I asked you. And while you were chatting, I took your LinkedIn profile, plugged it into ChatGPT, and asked it the same question.

Natalie Lambert:
Okay, thank you.

Sheikh Shuvo:
So your background, according to ChatGPT, is: Natalie teaches storytellers how to use the magic computer program to make their stories about products more exciting and fun, like a captain guiding a ship to amazing adventures.

Natalie Lambert:
That's awesome. I love it. And if AI is that tool that makes it happen, there you go.

Sheikh Shuvo:
And you were right about emojis in this one. There is a sunshine emoji and a rocket ship.

Natalie Lambert:
Yeah, if there are only two, that is less than most. I've gotten back some with, no joke, something like 20 or 30. It's absolutely crazy. I had a friend, I'm sure everyone saw, when ChatGPT first came out, that New York Times article with Kevin Roose having that entire conversation with ChatGPT, and you remember all that. I had a friend who literally copied and pasted every emoji in that entire thing. It was this grid of, I don't know, 150, 200 some odd. It might be more emojis just lined up, and you could see very clearly the personality that ChatGPT has.

Sheikh Shuvo:
Fun chat. Thank you so much for sharing more about your world and best of luck as you keep growing your firm.

Natalie Lambert:
Thank you so much.