Amplify: Generative AI: The New Rules of Engagement

Summary

Highlights

• Generative AI is poised to redefine how brands engage with customers by lowering data barriers and enabling personalized experiences at scale ⚙️.
• The shift from traditional to digital commerce has heightened the need for personalization due to increased competition and customer base expansion online 🌐.
• AI's ability to reduce content creation costs to near zero allows for personalization on a scale never previously attainable 📉.
• The importance of data quality and a unified customer profile is emphasized for AI to accurately personalize experiences 🧠.
• Human-in-the-loop AI applications are the stepping stones towards full AI-driven customer interactions, ensuring efficiency and personalization while maintaining control 🌟.
• Organizational structure and data strategy are integral to maximizing AI's potential in delivering enhanced customer experiences 🏗️.

Key Takeaways

• Generative AI is revolutionizing customer-brand relationships, making one-to-one personalization scalable and efficient 🚀.
• A strong data foundation is crucial for effective AI-driven personalization. Companies need to unify and understand their customer data to maximize AI benefits 📊.
• Human-in-the-loop systems are a current focus, but full automation and AI independence in customer service and personalization are on the horizon 🤖.
• Generative AI can significantly reduce content creation costs, enabling more dynamic and personalized customer interactions 📝.
• Companies must adopt a phased approach, starting small with human-in-the-loop processes before scaling to broader AI applications 🐢.
• Understanding and aligning organizational structure and change management with AI capabilities is essential for successful integration and use of AI technologies 🛠️.

Overview

Chapters

• 00:00 - 01:00: Introduction Joyce Gordon, Head of Generative AI at Parody, introduces the session alongside Logan Patterson from Slalom, who leads Slalom's digital strategy practice.
• 01:00 - 03:00: Pre-AI Era of Personalization and Scale This chapter explores the potential of AI to revolutionize the relationship between brands and customers by enhancing data accessibility and promoting customer-centricity. It provides insights into AI tools that lower barriers to data access, enabling teams to be more data-driven.
• 03:00 - 04:00: Impact of AI on Personalization at Scale The chapter discusses the significant impact of AI on personalization at scale, which is crucial for boosting customer lifetime value and attracting more customers. Traditionally, these two goals were mutually exclusive, especially in the pre-AI era of brick-and-mortar stores. AI enables businesses to simultaneously influence and manage personalization and scale, thereby facilitating revenue growth.
• 04:00 - 06:00: Current AI Use Cases in Marketing The major focus in this chapter is on how businesses, particularly in the marketing sector, are using AI for personalization to increase revenue. This is especially evident in specific regions where personalized experiences are advancing rapidly. Examples include personalized recommendations in wine shops based on individual taste and price preferences.
• 06:00 - 10:00: Challenges of Data Quality for AI The chapter titled 'Challenges of Data Quality for AI' discusses the evolution of business operations from traditional in-person experiences to the digital age. Initially, in-person banking and financial advisement focused on understanding customers' personal goals and creating personalized experiences. With the advent of e-commerce, businesses shifted their focus towards scaling operations, enabling them to expand their customer base from local to global markets. However, despite the global reach, personalization remained a vital element, though not as prioritized as scale in the early stages.
• 10:00 - 15:00: Real-World AI Personalization Strategy The chapter titled 'Real-World AI Personalization Strategy' discusses the evolution of customer interaction in the digital age. It highlights how businesses initially competed with local shops but now compete on a global scale due to online business expansion. The growing customer base means that personalization has become crucial for standing out. Initially, personalization started with simple tactics like using a customer's first name and evolved into more complex techniques, including lifecycle segmentation in emails, reflecting the increasing sophistication of personalization strategies in the digital era.
• 15:00 - 22:00: Future of Generative AI in Personalization The chapter discusses the future of generative AI in achieving dynamic personalization akin to the in-store experience with a fantastic store associate. It highlights the imminent capability of achieving one-to-one personalization on a large scale, which has been challenging and expensive in the past.

Amplify: Generative AI: The New Rules of Engagement Transcription

• 00:00 - 00:30 great to be here I'm Joyce Gordon Head of generative AI am parody and I'm here today with Logan Patterson from slalom Logan leads slom's digital strategy practice earlier today we had a chance to hear from our customers out window
• 00:30 - 01:00 and bobit and also see a demo of mgbt and AI so we really had a chance to see how AI is a tool that lowers the access to the barriers to access data and makes it really easy for your teams to be data driven and customer Centric during this session now we're going to be talking about how AI is going to completely redefine the relationship between Brands and customers so let's dive into it we're going to start our ation today
• 01:00 - 01:30 with two fairly modest looking dots on a graph personalization which is the major tool to drive customer lifetime value and also scale so getting more customers through the door these are really the two ways that a business can grow revenue and before AI they were pretty much mutually exclusive going back in time a bit to the era of brick and Border you couldn't really influence scale too much most businesses operated
• 01:30 - 02:00 in a particular geography they were able to engage or interact with customers in a particular region so the major lever they had to actually increase Revenue was personalization and during this era we saw all types of personalized experiences start to grow in store and I think we've all probably experienced a lot of these whether it's going into a wine shop and having them really understand your taste preferences and price point preferences or really going
• 02:00 - 02:30 into a bank and working with a f financial adviser who understands your goals brick and mor was really about creating these inore personalized experiences with Ecom the name of the game was scale businesses were now able to expand from selling to customers within their geography to reaching customers all over the world and what that meant particularly at the beginning is person ization was important but not as
• 02:30 - 03:00 important when you could only interact with customers in your geography overtime competition increased every business essentially moved online and then you were competing not only with businesses in your GE geography but all over the world so the customer base grew but the competition grew as well so personalization became really important in a digital world and we started to see personalization with first name then segmentation and emails so things like life cycle segmentation and then even got of more sophisticated techniques
• 03:00 - 03:30 like Dynamic personalization but we never really reached the level that you would experience in store if you are interacting with a really fantastic store associate with AI this is all about to change we're finally going to be able to realize onetoone personalization in a way that resonates with e with each and every customer while also experiencing scale at the same time why is this content is incredibly challenging and expensive to
• 03:30 - 04:00 create which is why personalization is so difficult and AI drives the marginal cont cost of content creation to zero so it's going to make personalization at scale digitally a reality Logan I know you spend a lot of time working with Brands shepherding them through this change as we think about how to adopt generative AI when do you think we're going to start to see some of these shifts towards AI in marketing yeah I mean I think we're seeing it uh and first and foremost I wish I had a glass
• 04:00 - 04:30 of wine I wish you would have offered that um so not going to hold out but didn't want to make you all all jealous um but I I do think we're already seeing it right and I think just to go back to your story right that entire journey I think was an attempt to personalize the brick and morar to the Commerce you saw you know very rough attempts of personalization of you know direct onetoone sites of selling everything from watches on watches.com to you know birdh houses.com that was an attempt at personalization at the time for putting
• 04:30 - 05:00 out a site that someone is coming with a purpose right I think now we've been talking about personalization so long the biggest question I hear from clients is you know how do I personalize right do I is it a technology piece the biggest question I ask back is you have that content engine right if you want to drive one toone personalization you have the content to actually get close to that and it's always been I think uh attempted but now with with Gen and different Technologies out there I think it can be a reality and so you know
• 05:00 - 05:30 Gartner is already saying that about 30% uh of marketing content is going to be generated by AI in 25 and frankly I think that's an underestimate um if you really think about you know name the technology how it gets started it start by almost Shadow operations like Skunk Works and if you if you pull even solom our own organization the number of people that are using AI even not at an Enterprise level but on their own personal time right a lot of this content is already being at least
• 05:30 - 06:00 drafted by gen right and so I think that's an underestimate I think we're going to see advancement pretty quick but um we'll talk more about how to get started there because I I don't think that we can all Leap Frog to that future that we've all seen on LinkedIn feeds and everything out there it's it's about pragmatic it's anything other any other technology advancement you have to start small in scale so this change is coming soon both in terms of people's personal usage and also usage within Enterprise as Logan
• 06:00 - 06:30 mentioned so we're going to spend a couple minutes talking about the use cases that we see people implementing today and also how we think this will evolve over time today most brands are getting started with humanin the loop use cases so use cases where the AI is really a collaborator to your team it augments your team but content is not going out into the wild without a human taking a look at it first and there's really two classes of use cases one that focuses primarily on segmentation and
• 06:30 - 07:00 generating more content for segmentation and the second is customer service so the name of the game here is really efficiency enabling your team to do more in terms of segmentation and content creation we see things like marketing copy subject line optimization image creation so imagine if you could create different imagery for example based off where someone lives curated product descriptions uh so we know for example that high value very loyal customers they want to hear more about your bronze heritage where customers that are a
• 07:00 - 07:30 little less loyal to your Brands they want to hear more typically about products and the technical details of those products so these are all use cases where AI is really able to augment your team help them get started and help them do more and then there's also customer service as well and in these use cases we typically see AI used is a tool to help draft first responses for the customer service agents the customer service person though takes a look at the response has a chance to edit it and
• 07:30 - 08:00 it only pushes send once they feel good about their response over time we're going to see increasing automation um and more onetoone personalization so things like AI concierge automated customer service without that human in the loop and onetoone creative generation and one thing we really like to think about is like what do we actually need to see or what changes have to occur to bridge this gap between human in the loop and increasing automation I'm going to talk a little bit about the technology side of things and then I'm going to hand it
• 08:00 - 08:30 over to Logan to think about the change management and organizational structure pieces in terms of Technology side it really comes down to three things the first is constraints the AI is not going to be able to do everything it's not going to give you the right answer with everything um so really coming up with ways to constrain the AI so it does well in certain use cases and maybe there's some use cases where the human gets pulled in I think the second is cost um we need cost to decrease here uh the
• 08:30 - 09:00 third is really benchmarking and tuning so the ability to judge like is the AI doing well or not it's an unsupervised problem it's really hard to measure and then giving your team the ability to tune those responses easily there's obviously techniques to do this today but they're really geared towards technical users and we envision over time many folks within Brands will have the capability to tune the AI and really match it make it match the brown voice and also the answers that business are looking for would love to learn a little
• 09:00 - 09:30 bit more about the organizational structure side of things yeah certainly I mean just to to kind of tap into some of what we're seeing today right I think the the thing that organizations get excited about are some of these early use cases but one of the ones that I get excited about and I think how you start to go from now and into the future is really the combination of a couple of these or a number of these use cases to do something really magical one example of that is is just think of all the organizations that have to spin up these micro sites for product launches or
• 09:30 - 10:00 other reasons right we're seeing many of our clients start to adopt all of these human in the loop use cases combining them to say how do we go from an idea creative brief to actually a representative uh page that we can then review all generated by AI I think those are the things that really get get folks excited but it starts you know with some of these early stage use cases I treat it from a transformation perspective like any other technology right we all know the the case for change is uh
• 10:00 - 10:30 really putting it in a um in the point of view of those who are going to have to be using the tools right you can't just go out there and say we have this new technology go out and use it right you have to paint the picture of what's in it for the organization uh what types of skill sets do we need how do we upskill our our teams and pragmatically approach it to drive adoption uh and ultimately efficiency via the tooling so I think really what these use cases have in common whe whether it's a micro site use case um or an increasing
• 10:30 - 11:00 automation use case like an AI concierge is the data Foundation these use cases are only going to be as good as your understanding of your customers and your relationship with them so we like to say that a gen strategy really starts with a customer data strategy to take an example imagine that you go into a store the store associate has never met you before and they start recommending products probably not going to be exactly what you're looking for versus
• 11:00 - 11:30 imagine that you go into a store and you're a regular the store associate really understands your preferences and what price point you're looking to shop at and knows uh what items you typically like it's going to be a very different interaction so to illustrate this we're going to walk through an example together of an AI concierge and the AI concierge is operating on the website of a sporting goods retailer we have a customer who's going seeing at Veil so he types into this AI
• 11:30 - 12:00 I'm going ski at Veil this weekend what do I need the AI comes back and says happy to help what's your first name and email address the customer is now already a little bit peeved because he's signed into the site he feels like it should add have access to his data but it doesn't so he says girl I'm I'm logged in how do you not know my email and this is a very loyal customer friend David here David enters his work email DAV David clampers atw
• 12:00 - 12:30 workmail.com but he's made multiple purchases using different email addresses um he's made many purchases with this brand over the years so he enters his work email and it says okay it looks like your last purchase was a children Sparkle unicorn one piece ski suit and a children siiz medium David actually did purchase this item but it was for his niece it was two years ago thei doesn't know this it's undeterred it says based on your last purchase here are some
• 12:30 - 13:00 recommendations for your first day skiing it recommends this fantastic unicorn helmet a pair of sparkly goggles these teal ski boots personally I think these look excellent I would happily Rock the unicorn helmet uh do you like unicorn my kids do but not me personally well maybe that taste will evolve over time and then David says hm I'll try something else
• 13:00 - 13:30 so what went wrong here what went wrong is a lack of context this AI actually does have some information on David but it only knows a little bit about him it has the orders associated with his work email it's quite fragmented so it's working off this incomplete data incomplete picture David's super frustrated he's not able to get the products he needs it's a slow experience and this is also going to be frustrating for the marketer as well who put this this in place it's going to have impacts on customer retention
• 13:30 - 14:00 so how do we fix this the issue here is really data quality and there are a couple components of data quality really starting at its core identity the average person has two email addresses they're active on five channels 12 if you're gen Z and they've moved over 11 times customer data is incredibly messy it's complex and this problem is even more acute for your best customers who have inter Ed with you
• 14:00 - 14:30 across more channels than anyone else and they're also the customers who are going to care most about personalization so we see that AI is a fantastic tool for driving a source of Truth around identity in your customer data you can use other techniques like deterministic matching or fuzzy matching but they really tend to break down when the data is messy once you have that notion of identity you craft a customer 360 pulling in data from all of your Brand's most important sources touch points whether it's online and offline
• 14:30 - 15:00 transactions whether it's loyalty data whether it's customer service metric uh customer service records and then we go ahead and we augment that data with attributes uh so whether it's a historical metric like uh what someone's preferred size is or predicted metric like lifetime value or churn or predicted product Affinity we augment it with the information that's going to be really helpful for personalization and we also see AI as an incredible important tool for augmenting your data
• 15:00 - 15:30 with the right metrics and projections so we like to say that you need AI to both craft the data foundation in order to prepare that data to actually drive an AI personalized experience Downstream Logan we'd love to learn a little bit from you where do you typically see your customers in their Journey as they're uh getting to a place of data quality yeah I think going back to your point before you went through the uh the challenging example there is
• 15:30 - 16:00 it it goes back to customer data strategy right and and one of the things I learned early on in my career when I you know was a cashier at a retailer was just walking the floor right and and you know as my manager at the time who was always reminding me when's the last time you walk the floor when's the last time you talk to a customer right that's effectively what he was saying I think you can repeat that here and by analyzing the customer data that you have that is our version of Walking the Floor today to really understand what customer data do we have and how does
• 16:00 - 16:30 that inform what we can do with it in order to create those experiences and so I think it all goes back to that let's start there let's start from the foundation let's make sure that the foundation is right and understand based on our vision and what we want to do what we need to close or what we need to collect more of in order to unlock that so I think that's really where we start every conversation is do you we can talk about the vision we can talk about the future but let's go back and understand what do we have to make that a reality and and if we don't have everything which is usually the case how do we
• 16:30 - 17:00 close that Gap over time to make that a reality I love that example because the challenge is really the same whether you're a single cashier at a brand in that freaking mortar experience or whether you're a brand with millions of customers who's trying to personalize with geni all right let's try this again with better data David again he's going skiing at fail next weekend he wants recommendations already we can see there's an integration between the site and David's data so it says welcome back
• 17:00 - 17:30 David awesome looks like lots of grape powder this weekend you already have a pair of all Mountain skis which should perform quite well in veil's terrain but not as well as powder skis want to see some options so taking a look at what's Happening Here the AI actually has access to D's David's purchase history so it knows that he's purchased this pair of all Mountain skis and it's able to take that and weigh it against the request that D David had which is that he's looking for skis for his fail trip
• 17:30 - 18:00 David decides he's actually not in the market for skis right now he says hm I'll stick with my current skis but I'm in the market for a new jacket show me three options sure here are some suggestions I think you'll love based on the expected conditions in your impeccable style and it recommends these three very brightly colored jackets frankly with this I'm surprised David was not into the unicorn um these bold colors will go great with the pants you bought last winter and they're available in your size medium
• 18:00 - 18:30 have a look and just reply order if you're good to go so once again we see that the AI is pulling in David's purchase history it's actually pulling in metadata as well on the products that David bought so it knows these pants are brightly colored and it's recommending jackets that have a similar style it also has attributes on David's profile so it knows that David's preferred size is medium it's only pulan products that are in stock in David's preferred size David super pleased he says love The North Face jacket yes please order
• 18:30 - 19:00 thanks so what's different the difference here is the data quality like the actual AI sitting on top of the data is the same but we can see that the experience is completely different when you have that unified customer profile Logan This was um obviously a synthetic example I know you have a lot of experience working with customers on real AI driven personalization strategy would love to hear about some of your real world experience I mean was it I mean Thomas I think is still out in the
• 19:00 - 19:30 audience somewhere is that integr are we going to integrate that into the epic app and you know media Network and run all of that yeah no more unicorns like okay I think we're going to do that um but yeah I think I think you're right right the the data quality is the differentiator what you can do is based on what you have in that in that data Foundation um and you know I see Thomas in the audience so thank you for letting me pick you out of the audience there but I think it's it's a unique example of how how do you target people in moment and that's a really tough thing
• 19:30 - 20:00 to do unless you have that data Foundation um you can only be as good as product recommendations based on product relationships which we've been doing for a very long time uh but what you can do here if you actually know the customer know where they are know the events uh know their behaviors you can actually start to trigger uh real meaningful personalized experiences and and that's the differentiator right there it's it's really the data so uh thank you for sticking with us I
• 20:00 - 20:30 I'll say that thanks for sticking with us the last last session of the day no wine in your hand I understand we'll fix that very soon one of us has maybe wine in our hand water wine water wine apparently um but just wanted to share a few lessons that we've learned uh that I've personally learned in working with some clients and that you know it it really is uh promising so I know that there is a massive hype cycle out there all around AI gen AI but I really do believe there's meaningful changes
• 20:30 - 21:00 coming and it's going to take time like any other technology we're going to go through those phases uh but I do think that geni promises to drive personalization to another level that we've really never seen before um but in order to get there you you do have to start small you have to start small with some of those humanin thee Loop use cases uh I've even seen some of the highly regulated Industries some of you in the audience are are within those uh figure out how do we start to drive organizational change by maybe maybe sing up a tool that is internally facing
• 21:00 - 21:30 right so we don't have to deal with the compliant situation quite yet let's let's train our people with an internal Tool uh and get used to accustomed to working with these tools first before we start to advance and expand that throughout the organization and then it comes down to you know creating that identity spine that we've talked about so much that data Foundation uh getting that right is going to save you so much time in the long term because you're not going to have to go and redo work you're really going to set the foundation from the get-go uh and then understand what
• 21:30 - 22:00 are those more advanced use cases that you can start to roll out over time and close gaps in that data Foundation because every every experience I've had is that no c360 no customer data profile is truly complete uh you can always evolve it you can always Advance it and there's always more that you can gather on your customers to build richer more meaningful relationships and experiences well thank you so much that brings us to the end today and and I I
• 22:00 - 22:30 know we're in the unenviable position of standing in between the uh Amplified content and also a happy hour so thanks everyone so much Logan thank you thank you [Music]