AI-powered, voice-capable chatbot helps shoppers make the right choice in stores
Photo: BevMo!

AI-powered, voice-capable chatbot helps shoppers make the right choice in stores

Bryan Wassel

Through a special arrangement, presented here for discussion is a summary of a current article from the Retail TouchPoints website.

While in-home voice tech is becoming more commonplace, BevMo!, a leading wine and liquor chain on the west coast, is piloting an AI-powered, voice-capable whiskey shopping experiences in five stores.

“We chose the whiskey category, with the idea that we would be able to take over 50 bottles from all throughout our store and place them in a single location with some exciting creative engagement, and see if customers really wanted to do what a lot of people say is going to be this new trend,” said Tamara Pattison, chief marketing and information officer at BevMo! in an interview with Retail TouchPoints.

The displays utilize an Amazon Echo smart speaker, custom signage and the SmartAisle system from Mars Agency to power the experience. Shoppers who walk up to the display are guided through several basic questions about their whiskey preferences, including what event they’re shopping for, their taste profile and their price range. The solution filters down the full assortment to three recommended bottles. The chatbot also can answer questions about a specific bottle or brand, recommend whiskeys with similar taste profiles and tell jokes.

AI-powered, voice-capable chatbot helps shoppers make the right choice in stores
Photo: BevMo!

Initially, the kiosks are being focused on two common customer interactions: exploration and gifting.

The technology offers shoppers a low-pressure way to learn more about the products on their own without the need to find an associate. The kiosks only interact after being engaged with first.

Three weeks into the pilot, BevMo!’s shoppers were responding positively to the displays, and its similarity to at-home voice assistants means many customers already know what to expect from the interaction.

“I think the number of successful conclusions, where they’re actually initiating dialogue and then getting to the three products, gets to a pretty high conversion rate,” said Ms. Pattison. “All of those metrics are very positive, but it’s super early, so we’re excited to see how this continues.”

The use of such voice-technology in stores remains relatively rare. Sephora and H&M, among the early adopters of chatbots at retail, both use AI-technology to personalize recommendations for shoppers, but it’s largely only accessible online or through mobile devices.

BrainTrust

"The chatbot is one approach and maybe it can be made into a subtle form of salesmanship using entertainment as the disguise."

Bob Amster

Principal, Retail Technology Group


"replacing low-level often transient associates with first-line technology and fewer high-level but very capable staff, is the path to successful in-store AI..."

Ken Lonyai

Consultant, Strategist, Tech Innovator, UX Evangelist


"It’s a “low-pressure way to learn more about the products on their own without the need to find an associate.” Wait. What?"

Georganne Bender

Principal, KIZER & BENDER Speaking


Discussion Questions

DISCUSSION QUESTIONS: Will chatbots eventually become a common selling tool at the store level? Do you see them as a complement or replacement for sales associates?

Poll

24 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Art Suriano
Member
4 years ago

Although I think chatbots will become more prevalent in stores and all around us, we’re not there yet because the technology is not there. How many times do we have to say “hey Siri” and she doesn’t understand the question? Also, noise is a significant factor. Stores have noise, and that will also be a problem causing the chatbot to not correctly hear the request. However, there’s a bigger problem. When I choose to use Alexa, I’m doing that myself and the device is helping me do something quicker and more comfortable than having to do it myself. However, when I shop in a store or dine in a restaurant, I expect service. The line in the article “so the customer doesn’t have to find an associate” is sad because it implies there aren’t any. How much did this investment cost or would BevMo have been better off spending that money hiring more staff and training them? Tell me how the customer is going to feel when forced to deal with a chatbot rather than a human being. Will the chatbot be capable of smiling at the customer, especially if they’re programmed to thank them for their business?

We’re putting much onus on technology which is excellent, but there has to be a benefit for the customer and not just a savings for the retailer.

David Weinand
Active Member
4 years ago

This will interesting to watch. There are generally knowledgeable associates at places like BevMo and Total Wine but they are hard to find. This is a very targeted application so I believe it will be successful for the company. As the technology matures and becomes more comprehensive and accurate, I think more applications will arise.

Bob Amster
Trusted Member
4 years ago

The chatbot is one approach and maybe it can be made into a subtle form of salesmanship using entertainment as the disguise.

Bob Phibbs
Trusted Member
4 years ago

People go to a store to feel something. I don’t see basically “Alexa tell me what I’d like” doing that. And how does anyone know this is truly what they would like? Unless you’re going to do it side by side with a human, how are these recommendations any better – and more impartial – than a mailer to your home? I appreciate the novelty but hard data would make a more compelling case for AI in-store exceeding customer expectations and delivering real, tangible results. Otherwise, it just feels like trying to replace employees.

Ethan Goodman
Reply to  Bob Phibbs
4 years ago

Hi Bob – I appreciate your feedback and thoughts. Very helpful. For what it’s worth, a few clarifying points:

(1) We generate recommendations based on preferences that shoppers communicate to the system. For example, if they indicate they want “something smooth” (v. bold) under $60 that has a “cool story,” our algorithm will quickly serve up 1-3 products from our database that best meet those criteria. So, while I can’t say for sure that the recommendation is better than a store associate’s, it’s at least as (or more) impartial.
(2) I can’t share specifics, but I can say that in our first two pilots we’ve observed increases in both category sales and customer satisfaction with the in-store shopping experience. Small sample size, but an encouraging start.
(3) Our goal is to complement store associates, not replace them — be it by providing customer assistance when they’re unavailable or with another shopper, answering customers’ initial/simple questions so they’re free to handle more sophisticated queries, or serving as an extension of their product knowledge.

Thanks again.

Steve Montgomery
Steve Montgomery
Member
4 years ago

The advantage of a well-trained salesperson is that they can guide the customer through all the options they have. In this case the item selection has already been narrowed down to allow the chatbot to function.

I’m not a liquor drinker but on my last excursion to buy beer the salesman answered my question about where a product was and in guiding me to its location asked some additional questions. Based on his verbal and physical guidance I was able to buy some great beers that brewers ration to retailers. I would have never found or selected them had I used the type of chatbot process described.

Bottom line, it will be a long time before a chatbot can come close to replacing a knowledgeable salesperson.

Ethan Goodman
Reply to  Steve Montgomery
4 years ago

Thanks for your feedback, Steve. I agree there are many cases when there’s no substitute for a person-to-person interaction with a knowledgeable store associate. But I’ve had many experiences where an associate either isn’t available (maybe because he/she is helping another customer) or isn’t equipped to provide an expert recommendation (for example, because he/she knows a lot about wine but not as much about whiskey). We designed SmartAisle to help bridge the gap in those situations and complement (v. replace) store associates, all in the name of making the in-store shopping experiences as helpful and enjoyable as possible.

Ben Ball
Member
4 years ago

This is a huge winner! Pairing opt-in information technology with a category filled with nuance and choice is a high value customer experience. The logical extension to wine varietals and appellations, craft brewers, etc. is easy to see. This approach will add value to any category with multiple choices, high variation and complexity.

P.S. If you don’t think whiskey has a high degree of nuance and complexity just Google “whiskey” vs “whisky.” Then head on down to Kentucky for the Bourbon Trail experience. When you think you have it figured out, come a little further south and the good folks in Tennessee will be happy to explain to you why there is a legal definition of “Tennessee whiskey” versus “bourbon.”

Ethan Goodman
Reply to  Ben Ball
4 years ago

Thanks, Ben. We appreciate your support. You’ve highlighted the exact pain point SmartAisle was designed to address — finding the right product in categories with large assortments, (relatively) high price points and complex product attributes. Whiskey fit that profile to a T; and the categories you mentioned (wine & beer), plus others like consumer electronics, are on our roadmap.

Ken Lonyai
Member
4 years ago

This is absolutely where retail is headed but in a more useful/sophisticated form than this pilot. The non-believers that color a virtual assistant as a replacement for a human in an either/or scenario don’t see the forest for the trees. They will be complimentary and eventually when brands utilize properly conceived VAs, the benefits of both AI and human interaction will be realized.

BTW – what I’m referring to is not an Alexa skill approach. That’s going to go nowhere. Brands will eventually be building or using a white-label platform that gives them control of their own assistants that live on consumer devices, not store shelves. When the assistants reach their limit, they will hand-off to a human, just as chatbots do now. To be successful though, smart retailers will need to ensure a robust handoff mechanism and staff their store with well-paid, knowledgeable associates that can close the deal when the assistants can’t. Sounds like a fantasy, but replacing low-level often transient associates with first-line technology and fewer high-level but very capable staff, is the path to successful in-store AI and happy consumers, especially digital natives.

Oliver Guy
Member
4 years ago

It will be interesting to see customer take-up. A lot will depend on quality of response not just there but in general dealings with such devices (I would personally avoid any Echo device because my experience at home has been so poor). There is the question of ambient noise and some customers may not take kindly to having responses to their needs spoken loudly by such a device – contrast a sales associate who can be much more empathetic.

Ethan Goodman
Reply to  Oliver Guy
4 years ago

Thanks for sharing your thoughts, Oliver. User engagement has been strong so far, but it’s still early days and there’s of course room for improvement based on what we’re learning at BevMo!. Background noise hasn’t actually been a significant issue, thanks in large part to the quality of Amazon’s and Google’s (we support both platforms) speakers & microphones and their native “voice lock” feature. But we’re constantly optimizing the platform to make our recommendations/answers better and smarter, and our experience more discrete, intuitive and human.

David Naumann
Active Member
4 years ago

For products with extensive details and repeatable sales processes, chatbots are a great extension of the sales associate when they aren’t readily available. Chatbots will likely complement sales associates, as there are times when a chatbot isn’t as effective as a human.

Programming extensive product details and insights from wine and other beverage experts into a chatbot can offer much more comprehensive recommendations than a sales associates. With the right voice prompt questions, the chatbot can narrow down the choices and offer targeted product recommendations that match the customer’s preference. As this technology is perfected, there are a plethora of opportunities to automate process in retail and improve the customer experience.

Ethan Goodman
Reply to  David Naumann
4 years ago

Thanks, David. I think you’re right on — our tool and other virtual shopping assistants like it are meant to be a complement to/extension of the sales associate (v. a replacement), can oftentimes provide more thorough information and recommendations than its human counterparts, and will absolutely get better and smarter over time.

Georganne Bender
Noble Member
4 years ago

It’s a “low-pressure way to learn more about the products on their own without the need to find an associate.” Wait. What? Learning about products on my own is what Google is for. On the sales floor I want to interact with a person – a living human being who knows his/her stuff and can answer every question I have, not just what’s pre-programmed into the device.

Ken Lonyai
Member
Reply to  Georganne Bender
4 years ago

Georganne – great in theory, but in reality that’s a rare associate. Example: we recently went to an appliance store in a very wealthy town that only sells high-end brands. We had the lead associate helping us and I had to correct him on a feature he stated incorrectly on the very first appliance we looked at. He moved around the floor with his laptop on the manufacturer’s site as his info source. Although he clarified things and added his take on customer experience with the brand we were considering, he relied upon what was “pre-programmed” into the brand’s website. AI and machine learning is far more than “pre-programmed” data. Today it’s limited. In 2025 or 2030, it won’t resemble anything we now have. Although retailers are slow to adopt change they certainly will not be the lone vertical to eschew what will be universal technology.

Georganne Bender
Noble Member
Reply to  Ken Lonyai
4 years ago

I don’t agree that it’s a rare associate. It’s unfair to paint every person who works in a store with a broad brushstroke — and it’s done too often. I hear people complain about Best Buy associates all the time, but I have yet to have had a bad experience with any of them in the past year and I am in stores a lot; it’s part of my job. I agree that by 2025 or 2030 AI will be much different than it is now, but that’s not my point. Most technology is meant to be used as a tool to make the experience better or to make a a store associate’s job easier. I will never be quick to discount the people who work in stores.

Ken Lonyai
Member
Reply to  Georganne Bender
4 years ago

Providing answers to simple questions, initial guidance, and self-service opportunities to customers that want it (many do) frees expert human associates to have more time for customers that need in-depth help DOES make a store associate’s job “easier” and/or allows them more time to support customers. I work with AI/conversational technology on a daily basis in an enterprise innovation environment and what I am describing is precisely the model that other verticals are developing. Retail will absolutely be there too.

Ethan Goodman
Reply to  Ken Lonyai
4 years ago

Georgeanne and Ken — great dialog here! I agree with both of you, I don’t think we’ll ever be able to (nor will we want to) replace store associates, and that’s not our intention with SmartAisle. Instead, as Ken points out, our goal (and I imagine the goal more broadly for similar in-store technology) is to complement him/her — i.e., enable the virtual assistant to tackle simple customer queries at the front-end of the sales process, and free the associate to handle more sophisticated customer requests (and other higher-order tasks). We’re also working on a training module that can help associates expand their product knowledge during downtime. I’m with Ken that it’s inevitable solutions like this will soon be the rule v. the exception at retail, as long as they continue to deliver a superior customer experience for shoppers.

Sterling Hawkins
Member
4 years ago

To me it’s a question of where they can add meaningful value, and I think BevMo! has found a use case here with SmartAisle. It’s something that (especially at this stage) isn’t appropriate for all retail verticals or chains, but I can definitely see them complementing more sales associates in the not too distant future.

Ethan Goodman
Reply to  Sterling Hawkins
4 years ago

Thanks, Sterling. That’s exactly right — we’re focused on categories and use cases where we can solve a shopper and/or retailer problem, add value to the experience and complement (rather than replace) store associates. Our solution won’t be appropriate for every category, but we think (and the results bear out so far) that Alc/Bev is particularly well-suited.

Cate Trotter
Member
4 years ago

Definitely a really interesting experiment. I’m still not convinced that a lot of shoppers want voice-powered services in stores though. How many of us actually want to be stood there saying things out loud? To be sharing our preferences in a way that anyone else can hear? This is the reason why you tend to see people typing search queries while out and about, rather than using voice. It’s one thing to talk to an assistant in the comfort of your own home, but another to be doing that in-store. I think the tech also needs to be good enough to cope with all the background sounds of a busy store (like other people talking) and still work well. Perhaps we’ll get there with both the tech and culturally, but in the meantime I’ll be intrigued to see how this idea goes.

Ken Lonyai
Member
Reply to  Cate Trotter
4 years ago

Curious: how is sharing personal preferences verbally to a virtual assistant different than doing the same with a human associate? I’ve heard plenty of personally identifiable information spoken in stores and medical offices and people don’t seem to mind blabbing at all.

Ethan Goodman
Reply to  Cate Trotter
4 years ago

Thanks for your comment, Cate. All of the questions/concerns you’ve raised are completely valid, and candidly are ones we continue to ponder, test and address daily. For example, re: discretion at the shelf, we started with a category like Whiskey where our hypothesis was that the sensitivity of being overheard in aisle would be lower than in a category like sexual health (this has proven to be true so far, based on the quant and qual data we’ve collected). We also designed the experience so that shoppers can give brief answers in lieu of divulging personal or potentially embarrassing information out loud. Re: background noise, this has been much less problematic that we anticipated. Between the quality of the speakers & microphones and the native “voice lock” feature on the Amazon Echo and Google Home devices (we support both), our ability to filter out background noise and hear shoppers’ utterances is actually really good. Still lots of room for improvement, but we’re very excited about the future potential.