Carrie Tharp from Google Cloud tells INFOSTRAVE what businesses and retailers should know about generative AI.
The Vertex AI platform from Google Cloud wasn’t created in a vacuum. Vice president Carrie Tharp said the company spoke with “everybody from luxury down to everyday goods” businesses to comprehend the difficulties and the areas where generative AI may be most helpful.
In her extended role as the president of strategic industries for Google’s cloud business, Tharp—the former head of retail and consumer at GC—is applying her vast retail experience from previous positions at The Neiman Marcus Group, Bergdorf Goodman, Fossil, and others.
In an interview with INFOSTRAVE, she said, “For me, all of [our] announcements are very important to retail and consumer goods.” Because they will set the standard for the rest of the business, it is important that Google and other large merchants begin integrating these experiences.
Tharp can see the possibilities and problems for the industry given her perspective on the technology and its particular application in retail. In order to better understand her viewpoint and the counsel she would provide businesses at this critical time, as the generative AI movement gets going, INFOSTRAVE chatted with the vice president.
Retailers and brands are considering various methods to incorporate AI into their operations since there appears to be a rush to adopt it, particularly generative AI. What are they looking at, and perhaps even more crucially, what are they failing to consider that they ought to?
This technology is new and in its infancy, says Carrie Tharp. The business had been cautiously integrating AI…Generative AI is not a direct replacement for AI, though. Due to the fact that it is a new discipline, new testing procedures as well as sophisticated ideas regarding responsible AI will be needed.
Consider the fact that earlier chatbots for customer care were coded. Based on the cues you gave them, they stated exactly what you intended them to say. Since generative [AI] is a creative process, you might get results that aren’t entirely factually true. We’re developing tools that can help with that, but you should check that it complies with your brand guidelines, among other things.
The use of generative AI is not a limitless panacea for all of your hard problems. Therefore, it must be developed, just like other technology, and we must learn how to manage the associated change and new disciplines. How will you A-B test generative AI, which differs from the deterministic flow of earlier AI, if you asked an enterprise today? Senior leaders must consider how to pace their organisations through change so that they are moving quickly enough to avoid falling behind or losing market share, but not so quickly that they make mistakes that negatively affect their brand perception or interactions with customers.
INFOSTRAVE: In terms of the data and an organization’s strategy, how do you distinguish between earlier incarnations of AI and new ones like generative AI?
C.T.: The tedious basics of building a data foundation needed to alter. There is no way to skip steps while using generative AI. It enables you to perform previously impossible tasks, such as having it analyse a photo and provide you with product details.
Previously, you either lacked the detail or had to perform the task using humans. As a leader, I never wasted any money constructing that data foundation. You can access unstructured data in a manner that you couldn’t before thanks to [generative AI]. How do you intend to use the data, though?
If I were to convert that into an AI leader’s manual, I would say…I believe you would need to reflect on a few things. What is the problem statement, to start with? What exactly are you trying to solve? Not everyone will use generative AI in the same way. They still need to review the core values that the brand represents. To minimise your cost base, you may employ generative AI to streamline, automate, and simplify back-office procedures. You might, however, also use it to strengthen your differentiators.
How can you utilise an AI assistant in your organisation in a way that complements what the humans and ambassadors do, if you genuinely believe that providing excellent customer service, styling, etc., is your thing? After that, choose your strategy.
INFOSTRAVE: Is it good to stick to back-end procedures and avoid showcasing the technology in front of clients if they’re dubious about it?
You need to be aware of how much you can tolerate in terms of control versus experimenting.
These are all still businesses at the end of the day, so you need to decide if you’re prepared to implement cutting-edge technology aimed at enhancing consumer encounters or if you’re going to start with back-office procedures like marketing optimisation to help your company go down the learning curve. Do you employ individuals in your company that are proficient in prompt engineering, for instance?
You might decide to start in some capability areas first if you look at the risk-versus-reward matrix so you can feel confident that your own organisation is moving up the curve. When you have the processes and standards in place to manage generative AI, you can unleash its power while still managing the brand and all of the experiences you’re producing.
Therefore, before they scale, everything undergoes extensive testing, whether someone is thinking about opening a pop-up shop or developing a new app. These principles also remain constant.
INFOSTRAVE: In the past, executives at department stores and legacy brands have asked me what to do with years’ worth of consumer data. A few years ago, the tech expert advised her to start over when the topic came up on a panel. Simply put, the job was too huge.
C.T. : That issue could seem completely overwhelming. Additionally, because some of the organization’s data sources were unstructured, [brands] occasionally were unable to access them. To transform unstructured data into something structured and useful, they lacked the necessary human resources.
When you think about generative [AI], huge language models come to mind. These models can analyse data, synthesise it, and determine what the insight or takeaway was, as well as how the information should be employed in future contexts. Therefore, it allows you to access all of the material in your corpus of knowledge as a firm and utilise that data efficiently. I hesitate to term it a shortcut. You’ll hear people discussing how data is their new asset, the future of e-commerce, and the centre of everything that merchants are focusing on. You can access [it] now thanks to generative AI.
It resembles the human brain. You are now able to put more of your business knowledge to use than you ever had before thanks to it. I believe some of the enthusiasm and fanfare stem from it. You can use original ideas, social media insights, sentiment, and visuals, which is a significant one.
An image was just an image in the past, and you couldn’t necessarily extract the data and use it. Recommendations are the best illustration.
The prior recommendations engine could only infer a relationship based on the information it had because it only had access to the product qualities you provided. If the most crucial product characteristics of the garments it was looking at were not taken into account, it might have been utterly off the mark. Due to the fact that the AI was only as good as what you told it, you could see people going through this phase of, “I need more information,” Machine learning can only detect trends in the attributions you are providing it with.
You now possess additional instruments to say, “I’ll go get generative AI.” Look at the illustration and make out other details that might explain why that particular item is so well-liked. It could not be the fundamental details that a person previously determined to be significant. Inventory planning and assortment decisions may benefit from these recommendations, or whatever else you’re using them for, in a way that wasn’t previously possible.
In addition to chatbots with more human-like interactions, this new generation of AI also includes AI image production. However, what are the possible risks? One clothing company received criticism for employing AI to depict diversity rather than using varied models. Will similar problems continue to arise?
C.T.: This reinforces what I said earlier, which is to “determine the problem, determine your approach, and then pick the technology.” Genuine is genuine is genuine. If your brand is not holistically focused on diversity and inclusion and you utilise generative AI to simply impose diverse models onto your photos, brand history would indicate that this is probably not going to be successful.
The technological issues are resolved via generative AI…However, you must plan and schedule the remainder of your business process accordingly. When we talk about “responsible AI,” we’re talking about how to use AI in a way that is inclusive, secure, and all of these other things. To get the desired result, technology must be used responsibly. Don’t use it as an excuse to omit a step that would normally have been part of your overall brand communication or general business process. Because I believe that’s where you’ll run into trouble.
The questions: Am I choosing the right platform? must be asked when you return to the tools and technologies. Am I making the proper choice in choosing a partner who has considered change management and the entire process that surrounds us? Because using technology merely for its own sake or as a shortcut could put the company at risk.
What other factors should brands take into account when selecting an AI platform partner?
C.T.: Will the platform or software provider I’m using scale to the size at which I must operate? Exists any safety and defence? For instance, is your data secure? Is your enterprise data being used by a model, whether it be an image- or a huge language-related model, to learn things about your customers or your brand that you don’t want to be made public?
We designed our tools essentially to let you fine-tune using your own data, which won’t be incorporated into the Google model. Therefore, considering all of those issues adds some new perspectives. If you don’t reconsider some of your evaluation criteria, change management strategies, and process controls, we’ll see more stores that struggle or didn’t consider all the potential effects or customer reactions.
Because AI makes it so simple, there is this whole concept of sampling, borrowing, copying, and remixing creative work. For example, this is how “Pulp Fiction” might appear if Wes Anderson directed it. However, isn’t it a complicated intellectual property problem? What prevents people from copying the designs of others?
C.T. : This fits into the category of ethical AI. As photography has transitioned to digital, new tools have emerged. For example, you can look back through the years at the difficulties brought on by technology. In this sense, the entire industry must change jointly.
In my opinion, generative AI is only getting started. We believe it offers up a wide range of possibilities for creative interaction, output, and working methods. However, you did point out some of those dangers.
Google is in favour of regulation, and our attention is on ethical AI. At the end of the day, however, there is still IP, brand voice and copy concerns, product design, all of which will need to be incorporated into how those things are safeguarded in a systematic manner across the available tools and capabilities.
That’s why we’re concentrating on making sure to include those things in our tools as we construct this platform [Vertex AI] so that when you work with Google, you know that you’re not leveraging someone else’s product or creativity in a manner that you shouldn’t or is overly referential. So, in that case, every minute detail of how it functions is practically being defined at the moment. However, I believe that this is the kind of item that will need to quickly climb the learning curve.
I do believe that from our perspective and what we’re sharing with companies, there are ways to assume that position of trusted content throughout this early stage of technological development. As an illustration, we are discussing how to develop a trustworthy platform with media players. So how can someone seeing something know that it isn’t a profound fake? How do you begin validating and endorsing content? Because of the dynamic that will emerge, people will demand the original, not just a rendition, of what they see or hear.