[ad_1]
Subjects
Synthetic Intelligence and Enterprise Technique
The Synthetic Intelligence and Enterprise Technique initiative explores the rising use of synthetic intelligence within the enterprise panorama. The exploration appears to be like particularly at how AI is affecting the event and execution of technique in organizations.
More in this series
Khatereh (KK) Khodavirdi is targeted on utilizing AI to create higher buyer experiences — a course of she compares to creating an “AI Legoland,” wherein numerous know-how parts match collectively to construct cohesive options for PayPal’s prospects. That is an method she is making use of in her function as senior director of knowledge science within the on-line cost techniques firm’s client merchandise division, the place she oversees knowledge science groups for PayPal, its peer-to-peer cost app Venmo, and e-commerce coupon-finder Honey.
On this episode of the Me, Myself, and AI podcast, KK joins Sam Ransbotham and Shervin Khodabandeh to explain how PayPal’s numerous client merchandise work collectively to assist customers have a seamless expertise throughout its merchandise. She additionally talks about AI’s function in additional personalizing the shopper expertise throughout the corporate’s model portfolio, knowledge governance challenges following company acquisitions, and her method to creating efficient groups.
Learn extra about our present and comply with together with the collection at https://sloanreview.mit.edu/aipodcast.
Subscribe to Me, Myself, and AI on Apple Podcasts, Spotify, or Google Podcasts.
Give your feedback on this two-question survey.
Transcript
Sam Ransbotham: What do Lego should do with how PayPal thinks about AI? Discover out on immediately’s episode.
Khatereh Khodavirdi: I’m Khatereh Khodavirdi from PayPal, and also you’re listening to Me, Myself, and AI.
Sam Ransbotham: Welcome to Me, Myself, and AI, a podcast on synthetic intelligence in enterprise. Every episode, we introduce you to somebody innovating with AI. I’m Sam Ransbotham, professor of analytics at Boston School. I’m additionally the AI and enterprise technique visitor editor at MIT Sloan Administration Assessment.
Shervin Khodabandeh: And I’m Shervin Khodabandeh, senior associate with BCG, and I colead BCG’s AI follow in North America. Collectively, MIT SMR and BCG have been researching and publishing on AI for six years, interviewing lots of of practitioners and surveying 1000’s of firms on what it takes to construct and to deploy and scale AI capabilities and actually rework the way in which organizations function.
Sam Ransbotham: At this time, Shervin and I are speaking with Khatereh Khodavirdi. She’s the senior director of knowledge science at PayPal. Khatereh, thanks for becoming a member of us. Welcome.
Khatereh Khodavirdi: Thanks for having me. Nice to be right here.
Sam Ransbotham: You construct and oversee large knowledge science groups for the various totally different entities inside PayPal. Lots of people know PayPal, however in all probability some folks don’t notice the extent of all of the PayPal actions, so perhaps let’s begin there. Inform us a bit about what PayPal does and what all these totally different subentities do as nicely. How are they linked?
Khatereh Khodavirdi: Sure. Clearly, a majority of the folks know PayPal by way of PayPal Checkout, which is a core product that now we have, and the corporate began from that product. However now we have a wealth of various merchandise, particularly on the buyer and service provider aspect. We’ve entire suites of merchandise to supply folks the aptitude on the service provider aspect to run their enterprise, from invoicing [to] having the offline functionality for monetary providers after which the net transactions, and on the buyer aspect, ranging from peer-to-peer cost to monetary providers to various kinds of credit score and “purchase now, pay later” functionality, and financial savings accounts. I can go on and on, however on the buyer aspect, now we have a wealth of various merchandise.
Over time, we additionally acquired many different firms to assist us speed up and likewise be incremental when it comes to the worth chain that we’re creating for the buyer and service provider, equivalent to on the buyer aspect, a few years in the past, we acquired an organization — its title is Honey — which truly lets you discover the perfect deal on the web whenever you’re purchasing on-line.
I began at PayPal on the buyer aspect, so I used to be serving to the small-business group, serving to speed up and resolve the issues for our prospects by way of knowledge and knowledge science functionality. After which, over time, I supported all of the service provider aspect of our equation — all of the enterprise retailers, channel companions, our relationship with Shopify, WooCommerce, Magenta — programmatically bringing a service provider for us.
After which, a few months in the past, I truly switched fully to the buyer aspect of the home. So I’m new on the buyer aspect however tremendous excited, as a result of you’ll be able to leverage a number of AI and knowledge science functionality to unravel a number of fascinating issues on this area.
Shervin Khodabandeh: That’s nice. So clearly, PayPal is a giant, multifaceted firm with totally different enterprise and subdivisions, and also you talked about these. However because you’re within the client group, let’s speak about that. Share with us a bit how AI’s getting used within the client enterprise to drive the themes you need together with your shoppers and what use instances is AI getting used for.
Khatereh Khodavirdi: We will truly carry all of the totally different accounts collectively, so now we have a viewpoint across the consumer interactions with us. For instance, if Shervin has a relationship to Honey with us, however on the similar time, he’s utilizing peer-to-peer or the Checkout product or our credit score functionality, how can we make it possible for we are able to truly take a look at Shervin’s relationship with PayPal by way of one lens?
So that’s, I’d say, one of many basic issues we are attempting to unravel as an organization. However the piece I’m personally tremendous enthusiastic about is across the truth of understanding the shopper journey with us and the way we are able to leverage personalization and AI to truly resolve buyer use instances — that “What are the roles they’re coming to PayPal to get completed?” and “How can we present them the customized and related messages to assist them?”
So give it some thought: Internally, we at all times mentioned, “Hey, there are a bunch of joyful paths with PayPal, and there are a bunch of unhappy paths with PayPal. How can we migrate folks from one joyful path to a happier path with us, and the way can we keep away from the unhappy path with the shopper?” And I really imagine that that is principally an AI functionality that we have to develop for purchasers. For those who construct that, it’s going to unlock a humongous quantity of worth for our shoppers and for us as an organization.
Shervin Khodabandeh: Let’s speak about a few of these paths. You mentioned, “What sort of jobs are folks attempting to get completed?” so let’s speak about that. Inform us how personalization might be useful there.
Khatereh Khodavirdi: Give it some thought this manner: Think about Shervin is among the folks that’s truly utilizing PayPal Checkout to buy by way of totally different retailers all through the web as nicely.
As you’ll be able to think about, as soon as we acquired Honey, we additionally had a complete wealth of the coupons which are [available] on the market, and over time, in the event you get a greater sense round what sort of classes or what sort of service provider Shervin is excited by, we are able to truly present Shervin the precise offers on the proper time.
For instance, it’s back-to-school time, and we all know that traditionally, at the moment, you store in any such class. At present, , Goal or Walmart or a few of the high retailers for again to highschool are working these offers, after which we remind you and present you related and customized offers to truly drive exercise with us.
Otherwise you may use Venmo on a weekly or biweekly foundation to ship any individual who cleans your home some cash. We will remind you that, “Hey, Shervin, it appears to be like such as you’re doing this,” and thru simply one-click contact, you go and do this. So principally, we change into a part of Shervin’s life and perceive what sort of exercise Shervin is attempting to do and make it a lot simpler for him.
The opposite a part of it that we’re all tremendous enthusiastic about is monitoring. While you store on-line to a special retailer, one a part of it’s, you wish to observe that order and see the place your order is and when you’re getting it. We additionally wish to make it a lot simpler — you’ll be able to truly get the notification and see the place your order is, and you may take a look at your entire commerce exercise and monetary exercise in a single place.
I’d say for me, additionally, the opposite half is that, sure, I’m in an information operate and we construct fashions, and I take a look at the result of our mannequin, but additionally the extra vital half for me is — I at all times name it … qualitative and quantitative. … I’d say, “Hey, I need a pattern of a few of the prospects on this group. I wish to perceive their journey, the exercise they’d with us,” or, , truly attend the consumer analysis research that we’re doing with prospects, so firsthand, I truly hear the challenges and the issues we are attempting to unravel.
As a result of for me, an important half is that I simply can not go and in isolation construct the mannequin and resolve all the issues. It’s actually doing the qualitative and quantitative points and studying from one another and bettering it over time.
Shervin Khodabandeh: That’s very useful. I’ve to think about that being extra of a tech firm than, let’s say, a monetary providers firm, that a few of these challenges are literally loads simpler so that you can take care of than it could be, let’s say, for a huge international financial institution that’s attempting to personalize throughout merchandise and enterprise traces and all that. Share with us a few of the challenges. What’s tough about what it’s essential to do?
Khatereh Khodavirdi: I believe one of many largest challenges now we have is that a few of the different firms that we acquired over time, every of them is utilizing totally different knowledge stacks, so principally migrating all of them to 1 knowledge lake and having one form of know-how to make use of throughout the board and with the ability to [build] that widespread knowledge layer platform … finish to finish, and understanding all of the contact factors and all the info factors now we have with the shopper. So having that widespread knowledge know-how platform is among the widespread challenges now we have internally.
Sam Ransbotham: That’s large, I believe, for everyone. I bear in mind my first expertise with this. Again in my previous life, I used to work on the United Nations, and I used to be working with databases, and I seemed down and we had simply actually dozens of databases we had been supporting, and I requested, “Properly, which is the usual?” They usually mentioned, “Properly, that is the usual.” And, I mentioned, “Properly, what are all the remainder of them?” “Properly, these had been the requirements then.”
And so that you’re in that very same state of affairs whenever you’re buying firms, the place you’re pulling collectively … numerous know-how stacks [but] you don’t wish to simply form of rip them out and begin over. How do you handle that course of? How do you get these in a cohesive knowledge science course of?
Khatereh Khodavirdi: Yeah, so, I’d say, Sam, you introduced an excellent query as nicely — that it’s not solely the know-how drawback side of it. I name it the info governance side of it as nicely: … what the definitions are, and the way totally different folks outline and take a look at various things otherwise, and how one can outline that widespread language throughout the board internally throughout the firm as nicely.
So that’s the reason we are attempting internally to develop the perfect practices that each new firm that now we have … listed here are the steps, one to finish, that we’re going by way of to include them as a part of the remainder of the info belongings now we have for the corporate. However you’ll be able to think about it’s not a straightforward train. It’s a humongous activity.
However that knowledge governance side of is it additionally crucial, as a result of when you’re bringing totally different parts collectively, it’s essential to take a step again and take a look at the definition from a special lens as nicely and see if these definitions nonetheless are related within the new assemble or not.
Shervin Khodabandeh: Let’s speak about groups a bit bit. You’re speaking a few collection of challenges with personalization, with different use instances, after which the capabilities which are required to get that completed, from knowledge coming collectively and id decision and plenty of, many different issues. Inform us concerning the staff. Clearly, they should have very robust technical capabilities, however what else do they should should work in that form of an atmosphere?
Khatereh Khodavirdi: I’d say this space is among the areas that’s so multidisciplinary, and you may think about the various kinds of issues you wish to resolve want various kinds of talent units and a lot of these expertise.
So I at all times mentioned that in knowledge science, AI, or the general knowledge area, one of many issues that’s actually vital is variety of expertise. And by variety of expertise, I don’t solely imply variety of gender or background, which is essential, however variety of thought management, variety of problem-solving, variety of technical talent set, as a result of on the totally different levels of the issue, principally, it’s essential to follow a special sort of muscle as a way to get the specified consequence.
So, for instance, one of many areas that I actually really feel [is] underappreciated within the knowledge world is enterprise acumen — individuals who can truly sort out the issue by way of a really structured framework and have the ability to synthesize the advice and the “so what” to the enterprise.
As a result of the worst factor that may occur is that you just take a look at your knowledge staff and you’re feeling like they’re constructing a bunch of black containers for the remainder of the group, and also you’re not investing within the final mile — that individuals are truly translating the “what” and the “why” and the “so what” to the enterprise group, to the product group, and to the remainder of the group. So you wouldn’t get the adoption that you’re hoping from the capabilities that you’re constructing.
So the way in which I’m it’s that we are literally additionally constructing a product group that helps the personalization and AI, as a result of like some other product growth cycle, you principally have a product technique behind this — that we’re not virtually constructing the AI mannequin to unravel particular use instances. We truly take a step again, understanding our client personas — what are the roles they’re coming to us to get completed? — and construct a product highway map and product imaginative and prescient round this and sort out this drawback in a cross-functional vogue as an alternative of simply internally throughout the knowledge group.
Sam Ransbotham: Is it tougher to get that final mile with AI-type tasks? Is that one thing the folks have a tougher time understanding? Is it one thing that’s tougher for folks to narrate to?
Khatereh Khodavirdi: It may be a bit bit tougher due to simply the dimensions of the issues that you’re attempting to unravel right here, as a result of folks can not relate to it whenever you’re speaking about tens of millions of shoppers. So for me, it’s how I can truly break down the issue and resolve smaller use instances by way of AI to create that adoption and championship within the group [so] that it’s going to assist me to unravel the most important drawback. It’s a humongous activity, as a result of I’m not solely speaking concerning the product contact factors that now we have. I’m speaking about each contact level that the shoppers have with us, both by way of customer support, by way of threat — by way of all of the totally different features throughout the firm.
Rallying all of these cross-functional features round fixing this drawback could be a lot tougher, versus my method, which is that I’ll first begin with fixing it throughout the product group, understanding all of the contact factors with all of the merchandise, understanding how we are able to perceive the product activity, and personalizing that element. After which you’ll be able to add a further layer — like carry the chance element — with every of the product parts, then you definately add the customer support.
I’d give it some thought as extra of a Legoland — that on the finish of the day, we could have an AI Legoland for PayPal. However proper now, the way in which I’m attacking to unravel this drawback is to construct every of the person Lego items with the hope that I can orchestrate and construct a Legoland and it received’t change into a bunch of separate Lego that aren’t orchestrated to unravel the widespread equation.
Shervin Khodabandeh: I really like the Lego analogy, as a result of my children are completely into Lego and now we have in all probability like 900,000 totally different items of Lego occurring at any given time. And in the event you simply take a look at it in isolation, you suppose, “OK, that is all we’re doing.” However, in fact, you’ve acquired to begin there, after which the items come collectively.
So then my query to you is, as with Lego, once I see my 12-year-old or 8-year-old constructing stuff, and I’m it in isolation, I may not have a full sense of the imaginative and prescient [of what] the entire thing goes to be.
So, I would say, “Oh, that is nothing,” or “What are you constructing? It’s like a small piece. Didn’t you do one thing like this earlier than?” After which they convey out the field with, like, 8,000 items that’s going to appear like this. After which, I am going, “Uh-huh.” How are you doing the massive “uh-huh” right here at PayPal so folks don’t lose sight of the massive imaginative and prescient and don’t get myopic concerning the little issues that it takes to get there?
Khatereh Khodavirdi: You may think about there isn’t a scarcity of particular person use instances — that there are various particular person AI, or no matter you title it, capabilities throughout the group. However whenever you take a step again, you don’t have that tenet to see how they may help you to truly construct that Legoland.
And truly I wish to sort out this drawback in a reverse order; first, I wish to take a step again and say, “Hey, what’s going to the blue sky appear like when it comes to the AI functionality and personalization for us?” So constructing that product technique and imaginative and prescient round it, after which attempting to unravel it backward, after which breaking it down into smaller items of the element Lego and being very prescriptive round what are the important thing issues every of those Lego will attempt to resolve, and why we’re constructing each bit of the Legoland — what would be the “so what” to the group.
Then you’ll be able to truly construct one thing … as a result of in the event you present that entire imaginative and prescient to everyone, it may be a bit bit an excessive amount of for a few of the folks to soak up it, so it’d actually decelerate your progress within the group. [In comparison,] whenever you present the larger imaginative and prescient, you’ve one thing to rally the entire group round. However on the similar time, you’ll be able to break it down into extra tangible parts so you can begin making progress whilst you’re truly holding that vitality and enthusiasm across the group to the north star that you’ve.
Shervin Khodabandeh: The breaking down truly is kind of vital. To construct on the Lego analogy, normally you get these 5,000, 6,000 items of Lego, and so they are available like 20, 50, 30 containers or little luggage, and so that you first do that, and then you definately do this, however then you’ve the entire thing.
Properly, I acquired one two days in the past that had 3,000 items in 20 totally different luggage, however all the baggage are unlabeled, so that you don’t know what goes with what. And so then, what now we have is, like, 3,000 items, and we’re attempting to construct a bit, and the analogy I’m attempting to attract right here is, as you’re constructing these little capabilities that then come collectively and get stitched collectively to assist your greater imaginative and prescient, how do you make it possible for these items are literally connecting fairly than the group is seeking to discover what piece goes the place, or how does this join, or do I’ve eight of those as an alternative of 5 of these?
How do you keep away from that? And that’s, by the way in which, one thing that goes on in a number of different organizations, the place there are silos of oldsters constructing issues, and so they don’t all come collectively. How do you method that?
Khatereh Khodavirdi: You introduced up an excellent query that, Shervin, principally, hey, I can not go to run a step from that imaginative and prescient and product technique for the AI and personalization to be a tactical element of the puzzle that you’ve. However when you construct that psychological mannequin across the widespread themes of the issue you are attempting to unravel, then I’d say the opposite actuality is also that I can not resolve all these issues on my own, [nor can a handful] of individuals resolve these issues within the group. You want to create that tradition, and it’s essential to rally your group round this.
Sam Ransbotham: That is sensible, as a result of I just like the Lego analogy, however the world isn’t so simple as Lego. You don’t have that completely labeled bag that will match collectively ultimately. You’ve acquired to carry these folks alongside to drag that in.
One of many issues that we did early within the pandemic was type our large issues of Lego and, when you’ve these Lego that aren’t within the little luggage, it’s virtually unimaginable to search out the precise piece to drag it again collectively. Shervin, it jogs my memory of after we had been speaking to Arti Zeighami at H&M, and he was speaking about working with particular person items …
Shervin Khodabandeh: On the wheel, proper?
Sam Ransbotham: His analogy there was tightening every lug nut a bit bit as you go round, and there’s some logic right here, too, that what’s totally different, too, about Lego is that you may resolve one bag and transfer to the following bag, however in actuality, you’ve acquired lots of people engaged on numerous totally different luggage who’re transferring at totally different paces.
Khatereh Khodavirdi: No, completely, Sam, and I’d say that’s the reason it is rather vital that, for my part, as a way to make progress on such a fancy subject, it’s form of a top-down and bottom-up method, and also you simply must have that check-in regularly. … The highest down, I’d name it your blue-sky technique round the place you wish to be with AI technique, and the underside up is simply primarily just like the tactical variations of the Lego that by some means exist within the group, or there are totally different parts that totally different groups are constructing, and how one can truly carry the 2 parts collectively.
And also you’re completely proper that the velocity of the event and making progress for some components is harder than the others as nicely. So that you additionally coordinate the totally different parts collectively, so on the similar time, you make progress, you’ll be able to rally the group round that, but additionally be life like that another components are extra advanced and it’ll take extra time.
Shervin Khodabandeh: And the one factor that I believe firms have that Lego items don’t is that they nonetheless have P&Ls and targets and numbers. And so I believe that’s in all probability why it’s such a nuanced method, as you had been saying, KK, that it’s such as you’ve acquired to determine the place essentially the most sensible path is to your group, given all of the gamers and all of the stakeholders and all of the items, and perhaps for Arti, it was a bit “one lug nut at a time” to get the entire thing going, and perhaps right here it’s like, “No, we’ve acquired to get personalization excellent earlier than we transfer on to threat or pricing.” And I believe that’s the nuances of various organizations a bit bit.
Khatereh Khodavirdi: Yeah, and I’d say, Shervin, you introduced up a very good level as nicely, that on the finish of the day, all of the organizations are very value-focused, each for the shoppers and likewise for the corporate … and the shareholder as nicely.
And for me, one of many largest dangers that individuals can do is simply trying on the output and the result of the P&L, [whereas] for such large tasks like this, like personalization, the truth is, it’s going to positively take a while so that you can truly get the true advantage of this from the output standpoint of your P&L. However what are the main indicators and the KPIs you’ll be able to have to truly maintain the staff and the group accountable, to make progress so that you just remember to are transferring in the precise route, whereas it’s going to take you extra time to see the entire profit as an output in your P&L?
As a result of the truth is that the worst factor that may occur is that every one of us know that that is completely the precise factor to do for the corporate, however since you wouldn’t … there isn’t a magic, there’s all this good work persistently over a protracted time period. There isn’t a magic that [lets] you see the output in a single day, so how will you maintain the staff accountable, the working staff, to make progress, however to the main indicator that that finally it’s going to get you to the result that you’re searching for?
Shervin Khodabandeh: For certain, proper? As you’re saying, the worst attainable factor could be to set a giant consequence aim for the imaginative and prescient, however for the improper time.
KK, this has been fairly insightful. I believe, Sam, we must always transfer to the rapid-fire query phase. So this can be a phase we do, KK, the place we ask you a bunch of questions in rapid-fire type, and please inform us what involves your thoughts first.
Khatereh Khodavirdi: Sounds good.
Shervin Khodabandeh: What has been your proudest AI second?
Khatereh Khodavirdi: My proudest AI second goes again to grad faculty, once I was at Carnegie Mellon, earlier than this area was this a lot in demand, and as a part of my graduate analysis research, I used to be constructing capabilities and platforms for vitality administration — good vitality administration for residential buildings. That was essentially the most proud AI second of my life.
Shervin Khodabandeh: Cool. What’s your favourite exercise that entails no know-how?
Khatereh Khodavirdi: Taking part in tennis, as a result of it actually helps me to give attention to the second.
Shervin Khodabandeh: What was the primary profession you wished? What’d you wish to be whenever you grew up?
Khatereh Khodavirdi: In all probability, for my first profession I wished to change into a professor or a pilot. I don’t precisely bear in mind which one got here first, as a result of in our household, schooling is a giant piece, and my mother truly had an academic profession, however in all probability both a professor or a pilot.
Shervin Khodabandeh: What worries you about AI?
Khatereh Khodavirdi: I’d say there have been a number of conversations about accountable AI and bias, and I chatted about this earlier — that it’s qualitative and quantitative as nicely. I believe it could be an enormous mistake to imagine that AI can resolve principally all the issues with out having the precise checks and balances in place.
Shervin Khodabandeh: What’s your best want for AI sooner or later?
Khatereh Khodavirdi: I’d say there are a number of challenges that the humanities are going through, from local weather change or a bunch of different stuff, so I actually hope that increasingly more folks truly play a job in utilizing AI to unravel these issues. A whole lot of my colleagues truly began investing extra of their time and vitality, and I actually hope that finally in my profession, I can also play a job as nicely.
Shervin Khodabandeh: Nice. Thanks very a lot.
Sam Ransbotham: So, KK, I believe that, clearly, the Lego analogy goes to be fascinating for folks, however I additionally like what you had been saying about issues like governance that, I believe, are basically vital for this, and the concept that you’ll come on and point out a few of that significance of issues like governance getting you to that scale that you just want — I believe that’s one thing that perhaps is extra widespread or common. Thanks for taking the time to speak with us immediately. We actually admire it. Thanks for becoming a member of us.
Khatereh Khodavirdi: Thanks guys a lot for having me.
Sam Ransbotham: Thanks for listening. Be part of us subsequent time, after we speak with Fiona Tan, chief know-how officer at Wayfair. Please be part of us.
Allison Ryder: Thanks for listening to Me, Myself, and AI. We imagine, such as you, that the dialog about AI implementation doesn’t begin and cease with this podcast. That’s why we’ve created a bunch on LinkedIn particularly for listeners such as you. It’s known as AI for Leaders, and in the event you be part of us, you’ll be able to chat with present creators and hosts, ask your individual questions, share your insights, and achieve entry to precious sources about AI implementation from MIT SMR and BCG. You may entry it by visiting mitsmr.com/AIforLeaders. We’ll put that hyperlink within the present notes, and we hope to see you there.
[ad_2]
Source link