Using AI to beat Amazon
An interview with Kerry Liu, CEO and co-founder of Rubikloud, talked to FierceRetail about AI, machine learning and the elastic cloud
Amazon is currently the frontrunner in the world of artificial intelligence and engineering. It’s accomplished a great load of “firsts” in the retail industry, including its work with Alexa. Kerry Liu, CEO and co-founder of Rubikloud, says that Alexa’s role in Amazon’s success should not be understated and that it’s a primary factor in the e-commerce company’s valuation and contribution to the progression of AI.
FierceRetail talked to Rubikloud’s Kerry Liu to learn how other retailers can compete with the e-commerce giant by using AI, machine learning and the elastic cloud.
FierceRetail: How would you describe the role of AI in the retail world today? Is that role changing?
Kerry Liu: Beyond Amazon, retail has historically experienced a technology deficit with old legacy software driving operations, but we’re now at a tipping point for the enterprise. Machine learning promises to help all retailers optimize their profits and customer base with predictive intelligence and machine-learning generated campaigns. Basically, AI is the X-factor. At Rubikloud, we see that the companies that are adopting AI right now are getting in front of their slower-to-adopt competitors and staying competitive with Amazon and Walmart.
We expect machine learning and AI to be par for the course in enterprise software in five years.
FR: How does AI differ from what Amazon and Alexa are doing?
KL: Think of AI in two form factors: The first is AI solving problems humans find very easy: vision, voice, distance control. Systems like Alexa are creating solutions by providing a consumer entry point into a larger community. But there is this whole second world in AI: solving problems that a human can’t solve, such as optimizing loyalty program for millions of distinct customers, or making pricing decisions on hundreds of thousands of SKUs. Humans can’t make these decisions, they need a machine. We live in that world at Rubikloud.
One of the issues with Alexa is that there’s a tradeoff between voice and visualization. For the consumer, speech recognition is a welcome novelty. Soon we anticipate Alexa will ask us for our weekly grocery order or holiday shopping list and then deliver our purchases to our homes automatically.What we sacrifice is the chance to incorporate visualization into unsupervised machine learning. To do that, retailers need a toolkit, comprised of deep learning, regression, a classifier and a data platform for starters. Until now, there’s been no packaged software solution for retailers to access.
FR: What are some of the challenges with AI?
KL: Quite simply, AI is overhyped and still a very new technology. Anyone developing “AI for everything” is probably not as successful as they’d have you believe. Even AI experts are learning something new every day. As AI technology grows and progresses, it takes time to educate the market on AI’s value. To some, “AI” still feels like something out of a sci-fi movie, so educating people that AI has, in fact, already arrived is a challenge the whole industry is facing.
Another challenge is highly concentrated knowledge among a select few. The biggest disconnect is that the skill set to do this is very different than the skill set to build an algorithm. There are a finite number of AI PhDs on the market and most of them work for the large internet giants. I am skeptical of a lot of companies that tack AI onto what they are doing, simply because the talent to accomplish the things they claim is monopolized by the big guys.
We haven’t yet reached success with generalized AI, either. AI is at the stage where it needs to be specific to work. I believe that vertical AI has seen successful applications but beyond that, there’s no panacea. The reason for this healthy dose of reality is that we’re in the beginning stages of a complete paradigm shift in computing—one that requires an entirely different ecosystem to support it. A lot of tech executives don’t even recognize the massive computing power needed to run it. You can’t run it on your own infrastructure because it won’t scale.
FR: Which retailers do you think could benefit most from AI? Are there any you don’t suggest getting on board with this technology?
KL: Retail is under siege right now. This past earnings season underscored the trouble the majority of retailers are facing and the pessimism on Wall Street. Omnichannel retailers should explore machine learning to help with their business and campaign decisions. However, the technology might not make the most sense for independent, small businesses that are native to the internet and don’t need to compile and predict data from multiple channels on- and offline.
My criteria for good candidates: First, retailers that are in a highly competitive vertical with a high volume of products and SKUs and frequency of purchase for their products would benefit a great deal from operational efficiencies. Second, if retailers have strong loyalty programs, AI and machine learning can unlock revenue they were not previously able to tap. Finally, and most importantly, they have a senior management team and CEO supportive of innovation through AI and willing to invest in it.
FR: Can you give some examples of successful AI-generated campaigns that you helped to run?
KL: Sure, here are the results from three campaigns where our packaged software helped retailers:
- A health and beauty brand with 1,000 stores increased forecasting accuracy by 24% and operational efficiency by 50%, reducing excess inventory by 30% and stock-outs by 24%.
- A beauty retailer generated $88 million in sales and migrated 21% of low loyalty customers to high loyalty with ML-generated promotions.
- A global conglomerate used RubiCore to build a cloud-based data lake 1.75 years faster and four to eight times cheaper than quoted by Oracle and systems admins.
FR: What is the elastic cloud and how is it being used by your clients?
KL: An elastic cloud and borderless infrastructure provide the greatest flexibility because retailers operate on thin margins and need to access cloud resources on an as-needed basis. In the second example above, a fixed infrastructure could have crashed under $88 million in sales generated in a campaign. Conversely, it would have been extremely expensive to pay for that amount of infrastructure when it wasn’t needed. Personally, I think the elastic cloud is table stakes—every retailer should live in this environment.
FR: What advantages and disadvantages are there with the elastic cloud?
KL: There’s more upside than downside. Microsoft Azure, Google and AWS are lightyears ahead of the rest of the market. Right now, Microsoft appears to be the most flexible and have the most aggressive pricing so we recommend them to our enterprise clients, but Google’s platform is excellent too.
FR: What does the future of harnessing data look like for retailers?
KL: The future survival of retailers is probably predicated on technology for most of the industry. Sadly, most retailers are behind the times and are feeling it. They need to crunch their customer and campaign data to optimize every business decision and transaction. For those at the forefront, machine learning not only has the authority to make campaign decisions, but also to execute them.
To do this, first you have to take control of harnessing your historical data from existing legacy applications—promotions, CRM, loyalty, POS, e-commerce and inventory. We do that with RubiCore. Then you can focus on your future data sources, such as in-store associates, your (future) mobile experience, in-store analytics, social shopping influences and port them into an application in an ecosystem that can combine the two together for AI. Once that is done, the machine learning will learn from every transaction, anticipate the best results and execute tailored campaigns on your behalf.
FR: For smaller retailers where the cost for these systems can be an issue, what is your response?
KL: The beauty of elastic cloud and the new software coming out makes it more affordable than ever for retailers with razor-thin margins. The prices of the new systems are usually pegged to the size of the retailer, amount of revenue, number of channels and physical locations, so a small retailer will pay proportionally less.
FR: What else can you tell retailers about AI/machine learning and the elastic cloud?
KL: Not using machine learning is simply leaving money on the table. E-commerce moves quicker than brick and mortar, and ignoring your competitors’ tech advances will come back to bite you. Customer loyalty is getting stronger each day, so losing customers to your competitors simply because they adopted machine learning before you will be a tough pill to swallow.
The ORIGINAL ARTICLE