In a recent interaction with Express Computer, Steve Conine, Co-Founder, Wayfair, and Fiona Tan, CTO, Wayfair, share how the company leverages AI and emerging technologies in retail. The conversation discusses Wayfair’s AI-driven personalisation strategies, innovations in customer experience with generative AI, and the balance between automation and human interaction in customer service. They also speak about the challenges of integrating AI while maintaining security and privacy, and the evolving role of robotics, IoT, and automation in their supply chain.
AI is the current buzzword across all industries. How are you leveraging the benefits of AI?
Steve Conine: At a high level, we’ve encouraged our entire team to explore and use AI tools. AI is new to everyone, and we don’t want just one team specialising in it while others miss out. By getting everyone involved, we increase our chances of leveraging AI across all areas of the business. In retail, which thrives on efficiency, AI has the potential to be a significant driver. For instance, our tech team is already using AI tools to improve code efficiency, and on the customer side, we’re utilising AI for visual inspiration, such as with our product Decorify. This tool uses stable diffusion to create millions of inspirational images, transforming our traditional image library. AI is also helping with tasks like improving product classification and search results, ensuring a smoother shopping experience. Ultimately, AI is enhancing our operations while saving labour and improving overall outcomes.
Fiona Tan: We’ve been using traditional predictive machine learning and AI for some time, so both our tech and business teams are well-equipped to identify areas where AI can accelerate progress. With generative AI, we can now personalise content much more specifically for individual customers, moving beyond the old model of segmenting audiences. For example, Decorify allows customers to upload photos of their homes and explore potential remodels, significantly increasing engagement. Additionally, we’ve made sure to provide everyone, including non-tech teams, access to AI tools like coding assistants. Encouraging widespread use of AI, even for simple tasks, allows our teams to identify opportunities for automation and innovation in their workflows. This accessibility of AI makes it a powerful tool for the entire organisation.
Initially, many enterprises were reluctant to incorporate GenAI into their workflows due to cybersecurity concerns. You encourage your employees to use GenAI for idea generation. How do you ensure that the originality of these ideas remains unique to your business?
Steve Conine: We’ve been very clear with our employees on when to use public versus private models. We have enterprise licenses for ChatGPT and Gemini, and our understanding from both Google and OpenAI is that these are private models. This means any data we input stays within the company and doesn’t go public. We educate employees to use enterprise versions for work, while it’s fine to experiment with public models in personal life. For example, I recently made a fun video speaking Kannada and Hindi, even though I don’t know the languages, to demonstrate the creative uses of AI. It’s important for employees to explore these tools in harmless ways, but we’re strict about not pasting sensitive information like our code into public AI models. Security and licensing are key, so we ensure everything runs on enterprise versions.
Fiona Tan: Yes, we also have a GenAI Council that formed early on. It includes one of our lawyers, science leaders, and product managers who carefully review different models and capabilities. Initially, we didn’t allow the use of tools like ChatGPT because we were waiting for enterprise-level protections and assurances. Once we got solid indemnifications, enterprise licenses, and data protection, we opened it up for broader use. The council vets everything, and we negotiate with vendors to ensure our data is protected before adopting new models.
You predominantly operate in the North American and European markets, using the same platform for both. How challenging is it to comply with the varying security and privacy regulations and how do you manage these complexities?
Fiona Tan: GDPR has been in place for a while, so we had already implemented the necessary compliance systems and data storage methods for our European customers. Their laws are stricter, and we were complying with those long before generative AI came into the picture. Now, with AI, it’s similar. For instance, if I’m working with UK customers, I ensure compliance with UK-specific regulations. For tasks like tagging colours, which aren’t restricted by GDPR, we can still leverage generative AI to benefit all customers. However, anything involving customer data follows different, region-specific rules.
I believe you have that feature as well, like AR, VR feature, where customers can take a snap of the room and then can place the product.
In the context of your augmented reality features, how do you use customer data to train your models? Additionally, what measures do you take to ensure data security and privacy while leveraging this technology?
Fiona Tan: With our augmented reality feature, we use 3D models of the products. When you take a picture, it only analyses your space in real-time; we don’t store any data from your room. However, if you choose to upload a photo to clarify something, we explain the process to you and ensure it’s done transparently.
Steve Conine: Yes, that involves a different set of terms and conditions. You have to agree to a specific agreement before using these features on our platform.
How are you incorporating robotics, IoT or other automation technologies into the supply chain? You had mentioned automation earlier, but how are you incorporating robotics and IoT to improve efficiency and reduce costs?
Fiona Tan: In our category, we use some robotics, but not as extensively as others. This is mainly due to the nature and size of our products, which don’t allow for full automation like, say, Amazon’s robots. So, while we do have automation, it’s not at the same level.
Steve Conine: We do employ robotic automation, especially in our return centres, but it’s more about simplifying transportation. For example, we use tugs, which are like automated trains that move goods. We have a fair amount of that. As for IoT, we’ve focused on damage detection. Damage is a big issue in our industry, so we’ve implemented devices that track potential damage throughout the supply chain. These tools help us and our partners ship products with lower damage rates, which has made a significant impact.
Fiona Tan: Additionally, in our fulfilment centres, we’re using computer vision with models from Google. This helps us detect package damage early, sometimes even predicting internal product damage before it reaches the customer. This saves time and money, as we can catch the issue before shipping it out. Initially, we were using the technology just to scan barcodes for sorting, but now it’s advanced enough to identify potential damage too. So, it’s been really interesting to see how we’ve adapted existing tools to improve efficiency and reduce costs.
Most e-commerce companies use GenAI for customer service. Do you also have customer service bots?
Fiona Tan: Yes, we’re using GenAI in two ways. Initially, it was to assist our agents with generating responses and making it easier for them to search for relevant information through a conversational interface. Now, we’re incorporating more advanced conversational interfaces into our virtual agent bots, which were previously more template-based. We’re focusing on creating smarter, more interactive experiences, allowing customers to self-serve for tasks like checking orders or processing returns. However, we still ensure that customers who prefer to speak with us can easily do so, striking a balance between automation and personalised support.
Are your chatbots capable of supporting other languages for different European countries? How challenging is it to manage multilingual support?
Fiona Tan: We support both French and English. Additionally, some of our suppliers are located in different countries, like Poland, where we have many Polish suppliers. We have had machine translation capabilities for quite some time. Now, we are evaluating and potentially transitioning to GenAI versions versus traditional machine translation. We’re assessing these options and making changes as we see fit.
You currently don’t have a business presence in India but have a GCC. Do you believe there is a talent gap given the rapid development of emerging technologies and AI skills? Is there a skills gap in the current workforce?
Steve Conine: What I see happening now, which I’ve witnessed a few times before, is that when a new technology like AI emerges, people often feel inadequate or experience imposter syndrome. This happened with computers too. Right now, many feel unsure or scared to engage with AI. But there’s a group that embraces it, learns, and eventually becomes the experts simply because they experimented with it. We’re still in the early stages globally, and it’s crucial that everyone engages with AI to avoid falling behind. We encourage our teams to experiment, make mistakes, and learn. If you wait for someone else to figure it out, you lose the opportunity to innovate. There’s plenty of good talent, especially in our India centre, and we want them to explore AI both personally and professionally. We provide tools and encourage usage, tracking who’s engaging with it to ensure that everyone is taking advantage of these resources. Change can be uncomfortable, but it’s necessary to stay competitive.
Fiona Tan: This is also a strong selling point when we recruit, especially in Bengaluru. We can tell candidates that we’re working on some of the toughest challenges, like selling home goods online, and we provide access to the latest technologies. We’re a founder-led company with an entrepreneurial spirit, and when you join, you’re not just an employee—you’re an owner. We expect and encourage you to use the latest tools, and that’s exciting for technologists.
What is there in your roadmap at Wayfair in the next one year?
Fiona Tan: Over the next year at Wayfair, we’re really focused on how shopping will evolve with generative AI. Since we specialise in home, we’re working on improving customer experiences that are particularly challenging in this space. One key area is helping customers find what they’re looking for, even if they can’t fully articulate it. We’re leveraging conversational interfaces and analysing browsing history to suggest items like a mid-century modern green sofa, based on previous interactions, even if you didn’t explicitly search for it.
We’re also exploring the use of text-to-image generation to inspire customers. If you’re unsure of what you want, AI-generated images can spark ideas, offering far more options than what’s available in reality. Another exciting initiative is our program to audit and verify products we believe stand out, giving customers confidence in their choices.
Steve Conine: It’s all about dramatic improvements in personalisation and inspiration. We’ve even been testing AI tools like Gemini, where you upload an image and it describes it for you. This helps customers who can’t quite put into words what they’re looking for. They can refine the description and get new image suggestions based on that feedback—leading to even more personalised shopping experiences.
You started your business two decades ago, and since then, the e-commerce landscape has changed significantly. How do you envision this landscape will evolve in the next four or five years? What future trends and developments do you foresee for e-commerce?
Steve Conine: It’s interesting because when we started this business in 2002, Amazon and Walmart were already big players, and we were worried about them then. In some ways, the core retail model may not change much, but with AI, things could evolve. For example, with advanced AI, if someone asks for help shopping for a living room, it should recommend Wayfair, especially if it’s suggesting the best retailers. Currently, we pay for ad placements, but with smarter AI, why should we pay if we’re already the best option?
This could lead to significant changes in the marketing landscape over the next few years, particularly in how traffic is driven to e-commerce sites. Companies like ours, which are operationally strong and have a physical layer, will likely continue to thrive. The businesses that can adapt quickly to changes in marketing channels will succeed, while those that lag will face the same fate as those that fell behind during the early days of the internet.
The biggest change in e-commerce will likely be in marketing, and the companies that can stay ahead of that curve will be the ones that win.