Podcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | Android | TuneIn | RSS
Generative AI creates new opportunities to create and manage content at scale. And scale is definitely required when crafting content experiences for one of the world’s largest ecommerce companies.
Anna Potapova is incorporating gen-AI across the span of her work at AliExpress: content creation and management, localization, personalization, and other areas where her strategic-content mind guides her.
We talked about:
- her recent promotion to a new leadership role at AliExpress
- which types of content are most amenable to being generated by AI
- the standards they use to guide the creation and ensure the quality of AI content
- the crucial role of content designers and localization experts in the ongoing iterative improvement of AI content at a large scale
- how AI enables the democratization of content creation
- the large percentage of user-generated content on the AliExpress platform
- how AI helps her team with personalization
- how gen-AI content helps them scale their marketing personalization efforts
- the importance of inviting yourself to machine learning and data science meetings to show the value you bring
- the value of case studies when communicating with internal stakeholders to show the value you can bring
- the importance of staying grounded in business objectives when developing relationships with your collaborators
- how a strategic approach to your work can help your org use AI most productively
- how the shift from hand-crafted content to AI content at scale manifests in content operations
- her plans to explore how AI can help evaluate content quality and conduct content audits
- the concept of hyper-localization, which addresses very specific regional and cultural differences
- the importance of proactively engaging with product and tech colleagues to ensure that standards-backed content powers AI products going forward
Anna’s bio
Anna Potapova is Staff Content Strategist at AliExpress (part of Alibaba Global Digital Commerce group). She changed team positioning from pure localization to Content Design, built a style guide and a system to maintain it, established standards for AI generated content in multiple languages and improved business metrics while reducing production costs. Anna has been featured on several podcasts (Content Strategy Insights, Writers of Silicon Valley, Localization Leaders), joined UX Evenings @ Google and helped to build a content community in China.
Connect with Anna online
Video
Here’s the video version of our conversation:
lkjsdf
Podcast intro transcript
This is the Content and AI podcast, episode number 33. When you create, manage, personalize, and localize content at scale for a global ecommerce giant like Alibaba, you need all of the automation help that you can get. In her role as a content strategy and design leader at AliExpress, Anna Potapova is harnessing the power of generative AI tools and techniques to address customers’ individual preferences, to help third-party vendors create better content, and to streamline their internal content design operation.
Interview transcript
Larry:
Hi, everyone. Welcome to episode number 33 of the Content and AI podcast. I’m really delighted today to welcome to the show, Anna Potapova. Anna is a staff content strategist at Alibaba, the big e-commerce merchant in China. She works specifically for AliExpress. Welcome to the show, Anna. Tell the folks a little bit more about what you’re doing these days.
Anna:
Thanks, Larry. Happy to be here again. Should I mention that since the last appearance on your podcast, I was on Content Strategy Insights with Arnaud. Since my last appearance I was promoted, I attributed exclusively to your podcast. Thank you so much for having me again.
Larry:
That’s too awesome. Thank you.
Anna:
Recently, I’ve been talking a lot about AI and my team has been doing a lot of work in this area. Last week I actually spoke in front of front the audience of over 200 people in Chinese about how my team is harnessing the power of AI as we need to generate a lot of content every day.
Larry:
Nice. And you as one of the biggest commerce, maybe the biggest on the planet, I mean, there’s a lot of content. Every one of those products needs something said about it. And every correspondence you have, there’s so much going on there. And a lot of it I’m imagining is either routine or data-driven or in some way amenable to the use of AI. Can you talk just a little bit at a high level about how you’re using AI? And also, I just want to note for folks that all of these conversations, we’re going to dance around anything remotely proprietary. We’re just going to talk in general about how big enterprises can work with big, vast content repositories and how AI can help. Can you talk a little bit about how you use AI to generate content?
Anna:
Well, first of all, it all comes down to what types of content can really be, dare I say, outsourced, can be created with AI because not every piece of content is created equal. Something very important for your product, maybe your core flaw, your gold path, all the UX copy over there. You really want it to be based on empathetic research, based on your users, based on very holistic view of this flow and challenges that people might potentially face. For things like that, I think it’s very clear that you will still need that human touch.
Anna:
And recently, I think what I’m really excited about is that I see my company finding good niches, finding good places where AI can really benefit both business and users. We’re diving into different content types and we’re exploring new opportunities to see how we can create more personalized content, more engaging content where it’s appropriate, where we know that it’s not going to fail us and it’s not going to harm the brand in any way.
Anna:
That ties back to quality centers. Of course, you cannot just have a large language model writing everything for you and having it all shipped without any quality control or any kind of involvement from professionals. Yeah. We’ve been building the system where we learn and try AI-generated content in different areas. And at the same time, we’re building standards to make sure that it meets expectations to our customers. It doesn’t contain any inaccurate or false information that it’s surely aligns with our brand overall.
Larry:
Nice. And you’ve said that there’s so much involved in it. And so it seems like developing standards is the prerequisite. You’re like, “Okay, here’s the threshold we have to meet in order to share this content.” Can you talk a little bit about how… You mentioned some of the criteria that it’s got to abide by the voice and tone of the organization, be accurate, all those things. Are there parts of that that AI can support enforcing your standards, I guess?
Anna:
Yeah, absolutely. But in that case, again, it requires more input from content designers and maybe even multilingual content designers as we work with many different countries and different cultures. It’s important that you included your content professionals in the whole process of prompt design, scripts. And make sure that you really use the expertise of people to make sure that we can constantly build up, we can constantly build up, and we can improve from iteration to iteration.
Larry:
Yeah. And I can only imagine the scale of localization that you must do because you’re a global company. You serve pretty much every country on the planet. Or do you constrain that in any way? Are you down to 50 or 100 languages?
Anna:
We serve over 200 countries and regions currently in 16 languages.
Larry:
Holy cow. And so that, again, that’s one of those things like machine translation is probably one of the oldest forms of, if not AI, at least automation in content. But can AI help with that improving those machine translations and just making the localization person’s job easier?
Anna:
Fantastic question. Fantastic question. I think what we’re talking about here is localization in scale. When you have, for example, multiple merchants on a shopping platform, or maybe you have multiple hosts on your apartment sharing platform, you need to make sure that the content that those customers publish on the platform is actually attractive and actually interesting to people who are going to use the services or shop with those merchants.
Anna:
It’s very important to empower people to create better content with AI, as long as you have this clear standard and you can make sure that your AI generates quality stuff. Patrick Stanford had this very good presentation recently on content design 3.0, talking about the impact of AI on the content design overall. And one of the principles that he brought up that really resonated with me is the democratization of content creation.
That more people will have access to the tools to create content. And if we can guide them, if we can provide them with tools that generate better content that is better for their customers, better for their business, then it’s really a win-win situation. That’s also one of the areas where we’ve been working on in order to improve the information that comes from AliExpress sellers.
Larry:
That makes me wonder, what percentage of the business that you all do is third-party merchants who are doing that kind of thing? Like people who maybe don’t have professional copywriters or just who content design is not their forte versus how much of it’s AliExpress. It must be a huge amount of third-party sellers on your platform.
Anna:
Yes, yes, absolutely. I think on the daily basis, most of the content that people browse on our platform or any kind of platform I believe is coming from those merchants. Yeah. Most of the things that they see are not coming from my team. What my team creates is just a very small chunk, very small chunk, especially when we talk about UI. When we talk about sales maybe a little bit more, but still not quite as much as sellers produce every day. Yeah, you can I guess say that they are the actual content team.
Larry:
Yeah, no, and now that makes me wonder how much of the… And I know we don’t have to go into the details of how models work, but I’m just going to speculate that the models in a system like that where you’re trained on data from content designers so that you can give end users feedback, like, “You might want to say this or say it that way.” Or I’ve seen, I’ve talked to other people I know who are developing systems like that, it seems like one of the best uses of AI.
Larry:
Like you said, I’m all about democratization, so the more of that, the better. There’s people in the UX research field who disagree, but that’s another conversation. As you’re talking about localization and translation, I’m also curious about personalization because that’s, I don’t know, another layer down or another instance of it. We were talking before we went on the air about the bicycle case study that you talked about. Can you talk a little bit about that, how AI helps with personalization?
Anna:
Yes. Oh, absolutely. I love this case study. It comes from my teammate, Maria (Li Xinlu). And what she’s doing is she’s working on our user growth content. And one of the areas that she’s looking at specifically is push notifications. And we’re sending literally millions of those every day, a lot, a lot. Many push notifications. What you noticed is a trend when you post, when you send to people some more generic content, they’re actually not quite willing to be engaged with.
Anna:
For example, you have two bicycle lovers. And in my presentation last week for a bipartisan conference, I called them May and Pablo. May is more of a hardcore mountain bike, hiking in the mountains, sleeping in the woods kind of biker. And the Pablo is more of a casual biker who just takes his kids out in the park and rides bicycles with them. Their needs and their expectations for biking related products would be drastically different.
Anna:
And when you send them the same notification and say just like, “Hey, everything for bikers 90% off,” or something, it’s very generic and it doesn’t give people confidence that they will see what they’re actually interested in. And for the longest time, I feel it was a problem for many organization. Creating content is costly, creating content is long. If you have so many users and so many product categories on your platform, how can you make sure that people really get personalized recommendations without spending unreasonable amount of money on that?
Anna:
It didn’t use to be possible. It didn’t use to be possible because it was just too long and too expensive. It wasn’t financially reasonable to do so. But now, when you have a possibility to match our product pool with some content directions, write different prompts for different product types, different product categories, different product pools, you can get very high level of personalization. So that you engage people with your content in a more effective way, and you also give them more of what they really want and what they care about. And that’s exactly what Maria (Li Xinlu) did.
Anna:
Instead of sending one generic push, she worked with the product team to make sure that we had technical capacities for a higher personalization. And of course, she was working on the copy and the creative directions for those different types of push notifications and help trained model based on your designer outputs. And yeah, we did see that people were more satisfied. We also saw a higher recovery rate on our platform just with the introduction of this new feature.
Larry:
Interesting. I’m curious about that. Again, don’t have to go deep into details, but assume there’s something like A/B testing or variant testing to see which message is working better. I guess it seems like is there a smart system that underlies this that’s, what, another learning model about, “Oh, this kind of message resonated with that kind of person,” or does it get down to individual level personalization? It seems like you can go any number of depths on that.
Anna:
Yeah, yeah. I think it all depends on the technical capacity of your team, how deep you can go. And of course, how well it performs for you and whether more investment in technology, more investment in content is reasonable.
Larry:
Yeah. Hey, one thing we talked a little bit about before we went on the air, and I don’t want to talk in particulars with any enterprise about this because data is the new oil. But just in general, there’s this need in any model to have good data underlying it. Now, I’m assuming that the world’s largest merchant probably has a lot of really good content and really good training data around. But I guess my curiosity about that is to what extent do designers and strategists at Alibaba or AliExpress or friends of yours work with the machine learning folks and the data scientists who do the backend stuff? I think one thing that’s come up in other conversations is these new collaborators that are showing up.
Anna:
Right. Right.
Larry:
Is that coming up in your work?
Anna:
Well, yeah, and I think one of the important aspects of that collaboration that I don’t see brought up very often on LinkedIn or in talks about AI is that you need to invite yourself to that conversation. You need to invite yourself to that conversation. You need to proactively engage with those people. And you need to step in and show them how you can make their work better, how you can make the whole project better with that design expertise.
Anna:
Because I think it’s incredibly important to have the sense of authority and have the sense of value when it comes to this. Because when you understand what you can bring to the table, you feel more comfortable introducing yourself and being like, “Hey, guys, you generating a lot of content, this is great, but I can help you make it better.” Or, “Hey, you generate a lot of content, but it has some legal risks, publicity risks. I can help you solve that.”
Anna:
I think the crucial part is make sure that A, your product team is aware of your existence. People who work on technical stuff know what content design is. And B, with your work and your contribution, show them the value, show them how you can help them reach the new level. I think some of those principles can guide pretty effective collaboration in my opinion.
Larry:
I don’t know if this is the case with you in your interactions like that, but I’ve had a number of folks from Google and Microsoft, other places who say that once they start those conversations, that people are like, “Oh, you know about language. You’re a word person. These are language models. That’s really helpful.” It’s not a stretch to see how we can be of help there. But I guess are there any particular… I know a lot of content people come from less technical backgrounds. Do you have any ideas about how to invite yourself to those meetings, how to begin those conversations?
Anna:
Again, I think one of the most effective ways to invite yourself to this conversation, to any conversation for that matter, is to show people what you can do. That’s why I value case studies so much. And if you ask anybody on my team, I’m talking about case studies constantly, and I’m asking people to do case studies constantly because it is a powerful artifact that can showcase the value that you create in a very clear way. When you can step in and say, “Hey, so I’m content designer on this organization. And look, I did ABC, and by doing so I increased conversion in this pipeline by this much. You want to work on AI together?” See what I mean?
Larry:
Yeah, yeah, yeah. And you’re like, “Hey, I can help you do even better.” Or that’s a great way to approach somebody. I mean I might be projecting my mother’s education of me on this, but there’s a lot of finger wagging in the grammatical and language world sometimes. And if you can come in more collaboratively, it’s probably a better approach. Yeah.
Anna:
Yeah. Nobody wants to be criticized. Everybody wants to be helped. I think that’s the attitude we should take. Yeah.
Larry:
Yeah. No, I love that. And that’s a great mindset to have because I think something that’s come with… I know a lot of people have just quit talking about fighting for a seat at the table. It’s just like, “No, I’m just going to invite myself or ignore the table or turn over the table or make a new table,” whatever it is. Yeah, that sounds like that’s resonating with you.
Anna:
Absolutely.
Larry:
Yeah. Yeah. Making your own table. And then like you said, you come in, you say, “Hey, here’s how we improve conversion rate here or adoption there,” whatever the metrics that matter to the people. That gets into the business side of this. I think one of the main ways that manifests right now is I talk to so many people whose bosses just come to them and just say, “Hey, put AI in this. We need that. I have to show my stockholders using AI everywhere in the organization.”
Larry:
That dynamic of… Because you mentioned earlier how this still has to be a human-focused, human-centered thing, and we’re always working off of UX research and customer interviews and stuff, but there’s still this business prerogative that we have to deal with. Any tips? Or that notion of showing people how they can meet their performance objectives, are there any other things that folks can be doing to stay more grounded in the business objectives of their work?
Anna:
I think staying grounded in business objectives is crucial. It’s absolutely crucial. And this is how you get productive collaboration with other people. This is how you have shared goals with other people, and this is how you get supported in whatever you are trying to accomplish. Now, when it comes to AI adoption, it’s very important, I feel to set the boundaries and ask why. Because AI is just a tool. Not all tools are great for everything. I mean, hammer is great for, I don’t know, fixing your house. It’s not that great for, I don’t know, conducting a Zoom call.
Anna:
You need to make sure that your stakeholders and your management understands that it’s not one-size-fits-all kind of situation. That it requires very strategic approach. Why? Because if you don’t check your quality, if you don’t try to be mindful about the usage of AI, then it can easily create some risks for your company reputation. I think that can be one of the ways to frame this conversation. “Hey, look, AI is not always a great…” Look what cases we can find now even online where inappropriate use of AI caused major PR risks for some companies. That can be one of the ways to frame this conversation. Say, “Hey, I love your idea. Dear manager, I love your idea of using AI, but let’s be smart about it. And this is how.”
Larry:
As you’re saying that you’re reminding me of something I wanted to come to you before to make sure we get to is the notion of operationalizing all of this. The last couple of years, we’ve all been frantically learning and try to catch up and how to do stuff. But in order for it to really fly, it’s got to be institutionalized and operationalized. It has to be all this stuff like standards, that’s a total governance thing and having… Which typically falls into that operations umbrella. Do you have a conscious approach to content operations? And is all this stuff starting to fit into routine content operations?
Anna:
Yeah. I think it’s definitely a very different type of projects with very different deliverables. Instead of creating individual polished pieces of copy, we need to create a system that would generate good copy on scale, on long-term basis. Definitely what we can see here is number one, deliverables are different. Life cycle is probably different, somewhat tied with development. And that definitely makes us think how to work in this new pipeline and also how to measure contribution in this whole process. Yeah. I don’t think we’re any close to have very well-defined AI content operations role yet. But then again, to be honest, I don’t even think that we did very good job defining the job of content designer so far. And it’s been at least 20 years. We have a long way to go. We have a long way to go.
Larry:
We’re notoriously bad at that. We’re the word people, and we should be…
Anna:
We should be ashamed of ourselves.
Larry:
Yeah, that’s a whole other… That’s an ongoing theme that comes up all the time in all these. But just that notion of thinking about the life cycle, and that brings up another thing related to that operationalization and governance it was a part of kind of things. It just audits. I mean, the fact that you have standards and enforcing them, that’s like an onboarding into the life cycle – attention to that. I’m just curious about how AI has shown up everywhere, and there’s so much things in audits that are tedious tasks which AIs can be good at. Have you done any content auditing or other kinds of content evaluation using AI?
Anna:
Oh, fantastic, fantastic thing to do. Absolutely fantastic. I do have it on my to-do list, but I haven’t got to that yet. I haven’t got to that yet. The way I’m thinking about using that is, again, we have so many languages, we have so many languages. And I think it would be nice just to start by defining some basic mistakes. I mean, if you have a lot of backstage content that may be not displayed very often, maybe some edge cases, error messages maybe written by developers years ago. Everybody has legacy stuff like that. I would love to just get AI to run a quick check on all that and quickly just define some issues. And of course, same as always, we can constantly build up on what we did. I would probably start with this basic surface-level things like typos, missing punctuation, wrong capitalization, and then try to grow from there. Yeah. Haven’t done that yet, but I’m thinking about it.
Larry:
Yeah, no, and as you say that, that’s like, it’s classic start small, just find the little stuff that you can… Details. And I’ve never met anybody who doesn’t have old developer-written error messages that they need to identify and tidy up a little bit. And also, just to mention, I should be able to get, I’m talking to Paula Land, the woman who literally wrote the book on content audits. I think we should be talking in the next month or so on the podcast about her take on it. And she’s doing some interesting work in that area now.
Larry:
It was something, you talked a bit earlier, you mentioned it, not in passing, but you mentioned that the localization, but you also talked before we went on the air, about a friend who’s an expert in hyper-localization, and I know that-
Anna:
That’s right.
Larry:
Yeah. I would love to hear a little bit more about that because that’s such a just… It’s hyper. It sounds exciting and big and important. But can you talk a little bit about… And she specifically mentioned, I think a presentation that she had done.
Anna:
Oh, that’s right. Yes. There is this wonderful, wonderful person and the content leader, Stephanie Pereira she works for Google. And she did a presentation during content designers China Meetup that I organized together with Arnaud every month. And what she was sharing is some strategies on how you can hyper-localize, what does it mean to hyper-localize? It means that you fully align your brand’s voice and tone, and also the way you communicate with the expectations of local audience, even within the same language. People who speak English in India, people who speak English in Singapore, people who speak English in Malaysia have their own regional and cultural differences when it comes to use of language and when it comes to daily communication. They have different terms for the same things, they have different expressions, and they play with language in the most great way.
Anna:
Stephanie is leading a team that does precisely that, that can help business create content that is hyper-localized, is really deeply aligned with particular audience in different countries. I find it absolutely fascinating. I feel like it’s really the next step for localization, being very local. Not all the brands can afford it, of course, but I do find the idea extremely exciting. How cool that would be for, I don’t know, a Netherlands brand to go to Malaysia and sound absolutely local. I mean, whoever can solve that mystery on scale is going to make a lot of money and become very prominent content leader.
Anna:
Yeah, I strongly recommend to check out her talk. There is a recording actually on YouTube, we can add it to the notes to this podcast. Really incredible, super grateful for Stephanie for sharing with us.
Larry:
Excellent. Yeah, I’ll definitely get that in the show notes as well. Hey, one other thing I wanted to ask about, just at least in passing is, Do you ever deal with SEO? Because I know that getting products found and SEO has always been notoriously difficult for merchants. Is that a concern of yours? And if so, have you found any benefits of AI for your SEO programs?
Anna:
Right. Actually at the moment, SEO related content is a little bit out of my scope, so I don’t have much insights into that. Sorry.
Larry:
Okay, no worries at all. Yeah. I just wanted to try to squeeze that in before the end. But hey, I can’t believe it. We’re coming up close to time, Anna. But before we wrap up, is there anything you’d like to revisit from the conversation or just want to make sure that you share before we end the conversation?
Anna:
Yeah, I think the key point I’ve been trying to make here is the way we create content is going to change, and we really need to make sure that proactively we’re stepping in and engaging with product teams, engaging with tech guys behind the algorithms, behind the code to make sure that whatever they create is not only scalable and usable, it’s also useful and safe to use. I think creating strong guidelines and creating strong quality standards of content and empowering more people to create content based on those standards is really something that we need to be thinking these upcoming years as content designers.
Larry:
Well, thanks for doing your pioneering work in that area and sharing it here. And hey, one last thing, Anna. If folks want to connect with you online or follow you, what’s the best way to find you online?
Anna:
Well, absolutely. You can find me on LinkedIn. Also, if you are based in Southeast Asia and you find Chinese time zone is not too different from yours, feel free to join online meetups. Every month Arnaud Frattini and I organize and online content designers meetup originally for folks based in China, but now we see people from Singapore and Malaysia and even from Brazil and US… How crazy is that? Joining us. Yeah. You can follow me or Arnaud on LinkedIn and we will post the announcements. We have really cool speakers showing up, and yeah, I’ll be happy to see you there.
Larry:
Fantastic. No, I love that you’re another node in the global knowledge network about content design and content strategy. Thanks for doing that. I’ll definitely link to that in the notes as well. Well, thanks so much, Anna. Great conversation. I really had fun.
Anna:
Thank you so much, Larry. Always a pleasure. Always a pleasure to record the podcast with you.
