🤔 Are you curious about how to enhance customer experiences and boost sales through AI-driven chatbots? Are you wondering about their role in building a successful e-commerce brand? Look no further! Our guest speaker, Blair Beckwith, is here to share invaluable insights!

🌟 Introducing Blair Beckwith: With a remarkable track record, Blair has been the driving force behind major e-commerce achievements. As the former head of the Shopify App Store, Blair elevated the platform from a team of one to over 15, achieving astounding growth from $30k MRR to an impressive $1m+ MRR! 

💼 Currently Head of Ecommerce at Tidio, Blair collaborates cross-functionally to empower online sellers in delivering top-tier customer experiences. As Founder of Railspur Partners, Blair has also guided numerous impactful companies in their journey to success.

‍🗣️: What strategies can we employ to ensure that AI chatbots contribute to reducing cart abandonment rates?

💬 Blair: You know, cart abandonment is like the digital equivalent of walking into a store, picking up a few items, and then just sort of... setting them down and walking out. One strategy is to use the chatbot to send a friendly reminder to customers who have left items in their cart without completing the purchase. You don't want to come across as pushy, but a simple "Hey, you forgot something!" can work wonders. Plus, chatbots can offer special discounts or incentives right then and there to sweeten the deal. Maybe offer free shipping or a 10% discount if they complete the purchase within the next hour.

‍

🗣️: Can AI chatbots effectively handle complex customer inquiries that require in-depth product knowledge or technical expertise?

💬 Blair: This is actually where chatbots shine the most. We’ve been able to automate the really simple questions from customers – think things like “where’s my order” for a long time. With the introduction of these large language models, we’ve been able to train these bots on all of the product knowledge you have internally and externally on your website. If the information that is required to answer a question from a customer exists within the training data, the AI chatbot should be able to leverage it to answer the question.

‍

🗣️: Can AI chatbots be utilized to guide customers through the sales process and improve conversion rates? If so, how?

💬 Blair: Absolutely – think about the real in-store experience and how customers are usually guided through the journey from entering the store to checking out. This type of experience has been hard to implement online today due to volume and timing of visitors, but chatbots make it possible.

I think there is some hesitation among customers today to engage with live chat for less serious questions, because they might feel like they’re bothering someone on the other end of the live chat. Once customer understanding of chatbots grows, I think we’ll see more customer engagement with these tools in a more casual, conversational way like you might experience in a real store.

‍

🗣️: How can AI chatbots assist in personalized product recommendations and upselling/cross-selling?

💬 Blair: Today, most AI chatbots are triggered by a direct question from the customer, so personalized recommendations usually only come as a result of a question from the customer where a recommendation would be a logical answer. This is going to change though as bots become more powerful and proactive.

Going back to my earlier point, though – this is where adopting a platform that does more than just AI chatbots becomes super helpful. Upselling and cross-selling can be handled well by a more traditional rules-based experience today, and so being able to leverage both AI and non-AI powered tools is super helpful.

‍

🗣️: What are the key factors to consider when designing an AI chatbot strategy for an e-commerce business?

💬 Blair: The most important thing to consider is what metrics you’re trying to improve by adopting a chatbot, and whether that’s a good fit for the product you’re selling. There are basically two reasons you’d adopt a chatbot today: a reduction in customer service queries and an increase in sales. Which of these is most important to you will answer most other questions about how you want to approach the implementation, and what types of training data you’ll need on hand to tailor the experience.

It’s also important to understand where AI chatbots start and end in your live chat experience. Not all interactions are best served by a conversational AI; some should be served by traditional point-and-click chatbots, and some should be sent directly to a human. Using a platform that handles all of these seamlessly lets you really tailor the experience.

‍

🗣️: What emerging trends do you see in the field of AI chatbots that could significantly impact e-commerce businesses in the near future?

💬 Blair: Honestly, I think this question is so hard to answer and most people who have a confident answer are just making something up. This field is moving so fast – if you had asked anyone two years ago what was possible in the next two years, I don’t think many people would have been able to guess at the progress we’ve made.

That being said – my hope is that we see a better integration of voice in the whole experience. I think the ineffectiveness of Siri and Alexa has left a bad taste in a lot of people’s mouths, but there is a ton to gain from having a similar style of interaction we see with text-based bots today with our voice. And the trends to enable this are there – speed of generation, accuracy of text-to-speech and speech recognition, etc. are all improving at incredible rates and will enable this type of interaction in real time soon.

‍

🗣️: How do AI chatbots enhance customer support efficiency, and what types of inquiries can they effectively handle? What's the balance between AI-driven support and human agents, especially for more complex customer issues?

💬 Blair: One of my core beliefs is that AI customer service will never – or rather, should never – replace human agents. I like to think of customer support questions as being in 3 big buckets: the trivially automated, the real relationship building moments, and then the messy middle. We have always been able to automate the really simple stuff, and we never want to automate the relationship building moments, so AI chatbots allow us to solve for that messy middle and dedicate even more time and energy towards the opportunities we have to really wow and connect with our customers.

‍

🗣️: Can AI chatbots aid in educating customers about the features, benefits, and usage of complex products, contributing to a smoother onboarding process?

💬 Blair: This is totally where chatbots shine – the information is almost certainly already available to customers (and therefore your chatbot), but we all know that customers don’t always have the time to go digging for answers. Chatbots have perfect recall for this sort of information.

‍

🗣️: What metrics should we track to measure the success and impact of our AI chatbot on our e-commerce brand?

💬 Blair: Depending on your goals, you’re probably going to want to focus on one of two areas: increased revenue, or decreased customer support load. Fo revenue, you’ll want to adopt a platform that can track revenue across chatbot-assisted and non-assisted visitors. For support load, you’ll want to track what % of interactions are dealt with by a human, as well as the accuracy of those interactions handled by the bot.

‍

🗣️: How do we continuously refine and improve our AI chatbot's performance and effectiveness over time? 

💬 Blair: AI chatbots can seem like magic, but it’s important to remember that they don’t actually know anything that you don’t tell them. In the context of a customer service bot, think of the intelligence in “Artificial Intelligence” more like memory than reasoning. These chatbots will today, in general, not know something that isn’t included in the training data. So when you launch your chatbot, you may feed in 20 FAQs that are on your website already, and it may be able to answer 30% of customer support questions. As questions get diverted to human agents, it’s important to make note of them and then add these human-solved questions to the training data to make the % of AI-solved questions increase.

‍

🗣️: Can AI chatbots learn from past interactions and adapt to changing customer behaviors?

💬 Blair: This will definitely vary from provider to provider – this field is so nascent, and every company delivering AI chatbots today is at a different part of the journey and focusing on different areas. Today, we focus the training of chatbots on explicitly given training data in the form of question-and-answer pairs like you might find on an FAQ page, although more in-depth. This generally allows for a higher quality of training data, at the expense of taking more time to set up.

‍

🗣️: What technical challenges might arise during AI chatbot implementation, and how can we overcome them?

💬 Blair: Technical challenges should be few and far between if you choose one of the more user-focused platforms. AI chatbots by their nature shouldn’t require an in-depth technical implementation – all the hard technical problems should be solved at the platform level. 

Just as one example, Tidio’s chatbot, Lyro, can be installed without touching code at all on some platforms, and by putting one line of code in to your site on others.

‍

🗣️: In times of high website traffic or during a crisis, how can AI chatbots assist in managing customer inquiries and maintaining a positive user experience?

💬 Blair: This is of course an area where support debt can build up super easily, and is therefore an area where chatbots can really shine. A crisis in particular usually results in a lot of the same questions being asked, which is really a perfect situation for a chatbot to step in and handle the bulk of those queries once they’ve been trained on the appropriate answers. A non-AI rules-based chatbot can also quickly be configured to deliver the right information to the right people before they even have a chance to ask.

‍

🗣️: With the rise of voice assistants and visual searches, how can AI chatbots adapt to these interfaces to provide a seamless shopping experience?

💬 Blair: All of the tools are there for this to be really huge, it’s really just a matter of pulling them together in a compelling way. You have speech recognition API’s like Whisper from OpenAI and a bunch of generative voice models. 

Nobody has really pulled these together in an ecommerce context and there are technical barriers to overcome around latency, especially at scale, but this seems so obviously ‘the future’, and as with all things AI, the future is closer than we think.

‍

🗣️: What strategies can be employed to build trust between customers and AI chatbots, especially among those who are skeptical of automated interactions?

💬 Blair: I think by far the most important thing we can do to increase trust is provide as close to an error-free experience as possible. Chatbots sometimes get things wrong or hallucinate – the best way as a brand to avoid this is to increase the training data available to the model. I keep repeating myself throughout here, but it really is that simple: the bot doesn’t know anything other than what you tell it. The best way to increase trust is to train our bots better.

Another key thing is making transition to an actual human agent as seamless as possible, whether right at the beginning because someone refuses to interact with the bot, or when an error happens. The worst way to build trust is to force a customer to use the bot when it doesn’t fit their paradigm.