Episode 300 / Emily Keller-Logan / Coursera / Director of Global Marketing Strategy and Experience
AI Agents & Data That Answers the Right Questions
Subscribe to Our Newsletter
Emily Keller-Logan is the Director of Global Marketing Strategy and Experience at Coursera, working to reach the world effectively with the largest online learning platform. She credits data with finding the answers to key questions, but not without qualitative nuances that give it context.
Her top data-driven marketing tip centres on curiosity and accessibility: anyone can become data-savvy without formal analytics training. Emily encourages marketers to dive into data tools like Looker and Amplitude, emphasising the importance of asking clear business questions. Her approach involves creating cohorts, pulling reports, and understanding campaign performance.
Critically, she recommends not just analysing numbers, but complementing data with qualitative user insights. She uses tools like Reddit and surveys to understand user motivations and experiences, ensuring data tells a comprehensive story.
Her shiny new object is AI agents - sophisticated tools that chain multiple steps together to solve complex marketing challenges. Unlike simple chatbots, these agents can sequence tasks, such as finding relevant SEO articles, adapting content for specific platforms, and tailoring messaging to different audience segments. Emily's AI agent experiments at Coursera focus on improving creative processes. These agents help teams communicate more effectively with creative professionals by providing structured feedback language and ensuring brand consistency across different channels and global markets. The key benefits of her AI agents include:
Faster content adaptation
Clearer creative communication
More nuanced audience targeting
Reduced time spent on repetitive tasks
Emily is transparent about the technology being in a learning phase, encouraging a test-and-iterate approach. Her ultimate goal is not just efficiency, but creating marketing experiences that truly understand and meet individual user needs. Her philosophy combines data curiosity, technological innovation, and a deep commitment to understanding human experiences - making her a forward-thinking marketing leader.
Check out Emily’s Substack where she covers her key learnings and more data driven marketing tips here.
Transcript
The following gives you a good idea of what was said, but it’s not 100% accurate.
Emily Keller-Logan 0:00
Get in there and try it and see what you can learn, and make sure it's tied to kind of one of those key business questions you're trying to answer.
Speaker 0:11
Are you a fitness loving foodie? Then Defi Snacks is for you defy offers real high quality chocolate you love, combined with functional ingredients, giving you as much as 26 grammes of protein per bag. So you can be both a little bit bad and a little bit good at the same time as a female owned brand, defy dedicates a portion of its profits to other women owned businesses currently sold in the US. You can find us on defy with an i snacks.com and follow us on social at defy snacks.
Tom Ollerton 0:44
Hello and welcome to the shiny new object podcast. My name is Tom Ollerton. I'm the founder of automated creative, the creative effectiveness ad tech platform, and this is podcast about the future of data driven marketing. And every week I have the pleasure and the privilege of interviewing one of our leaders. And this week is no different. I'm on a call with Emily Keller Logan, who is Director of Global Marketing Strategy and Experience at Coursera. Emily, I normally ask if I'm saying your name right. Have I got that right?
Emily Keller-Logan 1:12
That is perfectly right.
Tom Ollerton 1:14
I'm sorry. I should have asked. But anyway, Emily, thank you so much for coming on the show. Could you give the audience a bit of background on who you are and what you do?
Emily Keller-Logan 1:24
Sure, yeah, I work at Coursera, and so if you haven't heard of us, we're the world's largest online learning platform, and we have 350 top companies and universities. And the whole idea is that wherever you are in the world, you should be able to learn whatever it is you need to do whatever it is you want to do. And so leading global marketing strategy and experience, my job is to figure out how we help reach the world effectively with kind of how they can discover Coursera, use Coursera and benefit from it. Been in startups for 11 plus years, so I've been learning a lot about how to do effective marketing and really thinking about bringing brands to life.
Tom Ollerton 1:58
So in your career, what has been the best new belief or behaviour that has improved your work life?
Emily Keller-Logan 2:05
Yeah, I kind of feel like I am learning every day. And so I guess probably the shift I've made over the last few years is, you know, when you're working in startups, typically the idea is, like, no jobs too big or too small. If something needs to be done, you roll up your sleeves, and you get it done. And I think kind of starting out that's like a really good move, right? Be the one who shows up, takes on the task, figures it out. I think the biggest shift that's helped me as I've grown is kind of what I call it, like, knowing when to let it break, like you can't go do everything at a certain point, especially as you get into more positions of leadership. Advice I've gotten over time is, you know, by doing too much, you diffuse your impact, and then you don't have kind of as great of results as you would like to have. And if you're always filling in and catching gaps, especially as the organisation grows too from like early stage startup to growth stage company to now I'm at a public company. If you're in there filling all the gaps, the company doesn't know what's broken. It's a fast path to burnout. It doesn't let you be as effective as you can. And so it's really kind of figuring out, what can I uniquely do that I need to make sure I'm preserving my focus, investing in and getting done for the company. And then, like, you know, when you let it break, it's not just you're silent on the sidelines. You're trying to figure out, Hey, wait. You know, as other teams are kind of defining their charters, does this sit with you? Does this sit with someone else? Or making the case that, like at this stage of this company, I think we need to invest in a resource to take on this thing that I used to help get done for us. And so I think that has improved the quality of my life and also the quality of my work.
Tom Ollerton 3:37
That just sounds like a dream-like state that I've so far been unable to manage. I'm like, that, yeah, cool. That sounds great. Yeah, let's do that. Oh, what about this mad idea? Let's do that as well. I'm like, say yes to everything. I'm equally excited. Like, no job's too big or small. What's a, like, ridiculous challenge over there? Let's go and, like, Yeah, jump on that. But, at the same time you have to kind of manage and do all those other things. So I love the idea of putting your hand up and, you know, passing it on, or processing it, automate it, whatever it is. But how do you know when you you're broken, or your process is broken, or you've taken on too much without sort of winding up, you know, making bad life decisions?
Emily Keller-Logan 4:17
Totally, I It's like a kind of learn as you go. And so that's when I say, like, I'm learning every day. I think you're kind of kind of figuring out what's mine to do and what's not mine to do. And so typically, the way I look at it is like, is there no one else in this organisation with this skill set? And I think too, as I've as I, you know, aged up it, there's often other team members who are earlier on in their career who might just totally love the opportunity to step in and take on a piece of sometimes it's me getting out of the way, creating opportunities for other people to shine and show the skills that they have. And then other times, it's really recognising that, like, there are 20 ideas we could have right now. I remember my old CMO, she would advise me. She would say, like, you know what? That's a great idea. Why don't you put that on the list? And like, there might be a right time for it, but it might not be right now. So it's also just having the discipline to be like, I kind of love that idea, but like, this probably isn't the right time for it, because right now I need to dedicate my focus here. And typically I use our business goals as the guide of like, right now we need to improve retention, or we want to improve conversion rates here. And so if I go take on that top of funnel idea, like, that's not the focus of the company right now. So that can kind of too.
Tom Ollerton 5:21
I like that CMO play. That's a great idea. Yeah, put it on the list. It's a really that is a really clever soft No, isn't it like, yeah, brilliant.
Tom Ollerton 5:33
Let's get some specifics here. What's your best bit of advice for someone listening to this podcast to become a better data driven marketer?
Emily Keller-Logan 5:45
Yeah, so I'm coming from the strategy and the brand side, and often, maybe we have a little bit of a reputation of like, you know you're not metrics driven, which I think is unfair. So for anyone out there who maybe doesn't have formal training in analytics or using data. I guess my encouragement is, you don't need formal training to start using data to make better decisions. And kind of, my approach has just been being very curious. And I almost wonder sometimes if, like, not having formal training helps me, just like, go after it and not know, kind of, maybe what I don't know. And so we have tools like Looker and Amplitude. And so the nice thing about this is, like, data is really democratised in the company, but it's also my job to figure out, like, what what questions can I answer with this? And so we had a marketing ops director once who really encouraged all of us on the marketing side to be like, What business questions are you trying to answer with this data? So often I'm figuring out, like, what is it I need to know about how is the audience experiencing that campaign I just sent out? So then I'll go into Amplitude, I'll create a cohort and really look at, like, what, what did they start doing on the site after I sent that series? Did it work the way I expected? And so my advice is just to get in there and try it. Get curious with the data, pull reports, and really think about the question you're trying to answer, and then do if you're someone like me who maybe didn't have formal analytics training, like, find your performance colleagues or data science and analytics partners and go bring them your discovery and help them validate it. So sometimes I'll we have, you know, a lot of companies on Slack now, I'll go into our Data Science Channel and say, like, Hey, I pulled this this look, and I was trying to answer this question, here's what I found. Here's how I set it up. Does this look right to you? And more often than not, they're like, Yeah, that's right. Or sometimes it'll they'll reference a table that you want to pull and use differently. And so my biggest advice is just to get in there and try it and see what you can learn, and make sure it's tied to kind of one of those key business questions you're trying to answer.
Emily Keller-Logan 5:46
I feel like a massive fraud saying this, but I'm just about to publish a book called using creativity and data in marketing. And I interviewed a guy, a friend of mine, and actually advisor to the business, called Simon Kemp, and he said, The great answer to the wrong question is useless. And that quote really reminded me when you when you said that, right?
Emily Keller-Logan 5:46
Yeah, yeah. And it's, I think, knowing what you're trying to solve for, yeah. There's like, the Einstein quote too, right? Like, if you have an hour to solve a problem, spend, you know, most of it figuring out the right question to ask, yeah.
Tom Ollerton 8:03
And I think another actually quote from the book. Now, you got me thinking, I think, I think it was Perla Bloom and she, she said that unless your data dashboards and decks inspire you. You're not asking the right question. That's not a direct quote, but it's along those lines, you know, like, and that's what we always say at automated creative. It's like, we can give you those things, give you the three Ds, data flows and decks, but all of them should and do inspire what you should do next. And I always think, if you're pulling data and you're like, cool, got some data, but you're not going, oh, right, I'll go and do this now. Then it's not doing the job.
Emily Keller-Logan 8:44
Yeah. Can I add one more piece, I think, in addition to your data on your dashboard, I think, and I would be remiss if I didn't say that, like, hear from your users, right? So I think often I can pull reports and get the percentage trend or the graph see kind of, you know, how people are experiencing the platform in large numbers, but then I we, I really need to understand, like, what they're thinking and saying. And so we have a lot of qualitative surveys I will use Reddit. So there's this tool called Giga brain, where you can actually search Reddit at large for key questions about, like, what people are saying about their experience with your brand. And so I think kind of pairing data with the insight of like, what's motivating someone to do this, what's getting in their way as they do this, I think, often helps varying levels of the team make sense of the data and act on it, versus, like, sometimes you can kind of get caught up in testing madness and not really understand what the user really needs. And so I think really understanding what people are saying to you complements some of that data you're pulling on various dashboards too.
Tom Ollerton 8:44
This episode of the shiny new object podcast is brought to you in partnership with MADfest, whether it's live in London or streamed online to the global marketing community, you can always expect a distinct. Of and daring blend of fast paced content, startup innovation pitches and unconventional entertainment from Madfest events. You'll find me causing trouble on stage, recording live versions of this podcast and sharing a beer with the nicest and most influential people in marketing. Check it out www.madfestlondon.com.
Tom Ollerton 10:19
So we're going to now move on to your shiny new object, which is AI agents and creative effectiveness. So for anyone who doesn't know what that is, can you explain?
Emily Keller-Logan 10:36
Yeah, yeah. So I have been building AI agents internally with kind of the LLM large language model tools that are accessible inside my company. And the idea with an agent is like it is taking actions for you based on instructions you have given it. And so on the brand side, we have all kinds of documentation. You've got brand guidelines, writing guidelines, audience research, user research, and then for creative effectiveness too. Every platform has specs and requirements of how to show up well in that platform. And so what I've been trying to do is thinking about like, all of this information that like is hard for any one person in my company to know it all right, so we're building creative for YouTube, or now we're on meta, or now we're getting into Tiktok, or we're writing emails, but emails in kind of the US versus emails in India, we're going to try different approaches. And so it's hard for everyone to stay on top of everything. And so what I've been testing out and piloting internally is really trying to build out an AI agent that takes all of my information, and then I've set up instructions so based on what the user will query. So this could be a copywriter on our team. It could be a paid media manager. It could be me, and you're at a different stage in the creative development journey. It can kind of prompt the agent with different questions, and then the agent has guides on like how to score writing for clarity or visuals for visual effectiveness. It references at certain points, it will reference our writing guidelines, or it'll reference a platform guideline. And the idea here is to have our teams have better language to use as they're working with creative teams. What I found too, is it can be challenging to know how to give the right creative guidance. Like you don't want to solve for a creative you want to tell them the problem that you're experiencing. And so kind of these agents that I've been trying to create and build and test out internally are really trying to help teams get to better ultimate creative that really meets our audience where they are versus kind of where the company wants to be.
Tom Ollerton 12:40
So I've got a bit of a problem with the word agents. I think this is marketing trying to make itself sound cleverer than it is. So we do, we do the same thing. I wouldn't call them an agent. I call them like a mini GPT or a tool or a thing or a bot. So maybe I'm missing something. Maybe my own agents can't work hard enough for me, but what's, what's the difference between a bot and an agent like, what? Why is this agentic approach different to just creating a resource? I understand it's conversational, but can you help me? Yeah, help anyone listening to this podcast understand that difference and the opportunity?
Emily Keller-Logan 13:15
Yeah, I think it's like when it's chaining a set of steps together. So typically, as I've created, like, you know, in Chatgpt or Claude, you can kind of create a project, and then the project, they're referencing a set of documents. So that, to me, is more like chatbot LLM conversation. Agents to me, it's like, when you're chaining a set of steps together, where it's like, we're going to start here, then it's going to do this, and then we're going to go here and we're going to add that. I think that's where an agent becomes a little more useful. I think something I have not dabbled in, that I'm excited to try, is like, when you like, hand things over to the agent and have them do things for you. So I have someone actually on my team who is working on we have this kind of library of SEO articles. And so these are some of just really helpful, like, 1000s of writing pieces on the internet of basically, like, answering people's questions, so the first agent can go query the article library and find you the right topic for what you're doing, and then the second agent will pick it up and figure out, you know, if I wanted to take that long form article and turn it into like a feature in an email, how do we meet the specs of that email? So the subject line should only be this long, the you know, headline should only be this long. The CTA should always be that long. Then we compare it with and now adapt that for this audience. And so maybe we've built an agent that's all trained to be like data professionals and the kind of questions that they have. And so when you're chaining all these steps together, I think that's when you get agentic, versus kind of when you're just querying a chat bot with, like your brand guidelines, I think that's more chatbot LLM style. I don't know if that helps.
Tom Ollerton 14:43
And so then this is fascinating. Thank you. So, so how help me understand? So, in a an environment like chatgpt you have projects, and then you whack in however many documents. Doesn't take many for it to fall over in my experience, but, and then you go look, well, you know I need to do, I need to make, write this article based on this thing. And. And I need advice on bla bla... how is that handoff working? What's, what's the difference there? Whereas I'm just, yeah, just trying to help people understand how a project is different to a sequence of of mini projects, or mini bots, if you like.
Emily Keller-Logan 15:12
Totally. And I'm not sure how that that stuff works in chatgpt. So our company has kind of built, like an internal version of LLMs where, like, you can pull on like the LLM Foundation, from Anthropic, from Claude, from Gemini. But then our development team has basically built this like agent system, where then you can kind of create one agent, chain it to the next agent. And so there's basically just a set of steps in our kind of internal LLM. We call it playground, where you can kind of build one piece, connect it with the next piece, and there's documentation that our development team has written that then, you know, we're in marketing over here and trying to figure out how to connect these things together and trying it out. And so I will just be really transparent too, that, like we're in the learning phase. And I think everyone is of like, how do I use this as something that can work repeatedly? And kind of the way we've approached this is, you know, we look for volunteers in the team. We're like, Hey, this is the problem we see all of us having. We've come up with this solution, and we think this can really add to your life by, like, taking away the thing that's hard for you to do, or the thing that's really time intensive. Give it a try, let us know how it works. And then you kind of keep going back and you edit the way it connects to this piece, to the next piece, or the set of instructions you have for one agent. Maybe needs a little bit of refinement when someone comes at it with a question that, like we didn't think of on the strategy side, but maybe on the paid ads side, they have a need that we didn't predict. And so just a lot of like learning and doing it is a little bit of the like the answer to the data driven question, just like, try it and see what works.
Tom Ollerton 16:44
What kind of wins have you had so far? I mean, the win is trying, obviously, hard, but in terms of, like, you know, running upstairs and going, Look, I've created an agent that does blablabla?
Emily Keller-Logan 16:54
Yeah, some of them went so I've had people tell me I didn't know how to show up in a conversation with creative to like, give that and I see this so often, of like, how do you help a designer or a copywriter see the like, give them feedback that isn't solving it for them? It gives you the language of how to like, describe what you're looking for. And so people have told me, like, because it's kind of got all this right language for me, I can have clearer conversations with creative to get better output. I've heard from our email marketing team who really would love to leverage that, you know, great article library, but it's really time intensive for them. They're like this took would used to take me hours to go, you know, sift through the content, adapt it for the channel, and now it can happen in a matter of minutes. And so I think there's just those are some of the smallest. And then I think somewhat like the non wins are like, wait, I tried this, and then it couldn't figure it out. So then you just go and you adapt, and you see how to change the instructions so that it can meet that use case. And and then maybe we realise that actually, you know, that's too much to happen in one of these, so let's build out a separate flow and experience to solve that use case?
Emily Keller-Logan 16:54
Yeah, I always think that the test that I have is, if it's doing something faster and quicker, then you can just assume that your competitors doing it faster and quicker as well, right? But like, however, it's like, how is this allowing us to do something that we couldn't have done before? So that's always my measure. Like, what does this unlock for us that isn't time and effort?
Emily Keller-Logan 16:54
Yeah. And I think for us, like a, what I hope to be Coursera's differentiator is the fact that, like, where we're going to approach this is, like, the learner is always the hero of of the story, right? And so I think it's really easy for companies to be like, here's my brand, here's what my brand does, here's who we are. But instead, I think what could, what can and does make us different is really meeting people where they are and what it is that they need. And so the more I can help teams and but we're a global platform, so it's hard for one person to know the nuance of every last person on the planet. So how do I take all of that audience research? How do I take user research? How do I take what it is we can know about the challenges someone has in advancing their career, and help them, especially on the creative side, get into the heart of the problem. Someone has to then build a story or an email or an ad that when you see it, it meets you where you are and like, that's the journey you're on. Like, are we perfect in that? Absolutely not, but I think that is what can so if we can save time, great. So is everyone else. But if we can show up in a way that really gets you and knows you and matches you to the right, high quality experience that only we have, then that's, I think, where we win.
Tom Ollerton 19:24
Emily, this too. It sounds too good to be true. It sounds magical, but we've got to leave it there. Unfortunately, we're out of time, so I would have to keep on talking to you. I'm sure other people would like to get in touch as well, but where's the best place to get in touch and what makes a message that you respond to?
Emily Keller-Logan 19:39
Yeah, that's great. I'm at writing on substack now called it's at marketing in the messy middle. And so I love to have conversations about, what are you learning and what's effective in marketing, and what's not? How are you using AI, and where is it working, and where is it not? So those are the things I write about. And I've gotten people who've reached out to be like, You know what? This helped me? You know? Have the confidence to start talking and sharing my own learnings. Or I've had people who have talked about like, this is how I'm going into my next marketing meeting to talk about the use of AI, like, what do you think? And so I think it's just engaging on some of the harder topics that we have when you're at this level in marketing and and you're figuring out as you go, so that's kind of where to find me and how to reach out.
Tom Ollerton 20:20
Right, I found it already. Let's share that as well in the in the notes. So Emily, thank you so much for your time.
Emily Keller-Logan 20:28
Yeah, thank you for the opportunity. It's been great.
Subscribe to the ‘Shiny New Object’ Podcast on Apple Podcasts, Spotify, YouTube and Soundcloud.