Is Generative AI the Answer to All Questions?
A common refrain for retailers is, “I know generative AI [artificial intelligence] is going to impact our industry, but I’m not sure where to place my bets.” It’s clear the technology has far-ranging capabilities and potential, but where should retailers invest? Of course, the answer will vary for each retailer as each situation is unique, but following are suggestions for inspiring some ideas.
Questions retailers are asking about generative AI
There’s a ton of hype around generative AI right now, and as such, expectations are inflated. For example, we hear retailers ask, “Does generative AI replace all my machine learning [ML] solutions?”
The short answer is, probably not.
The following image illustrates how I think about generative AI compared with ML in a simple old school analogy.
Of course, generative AI can also predict things, but not necessarily better than a ML solution. Two real-world situations that often come up are whether to use existing forecasting and personalization ML models, or move to solutions using generative AI. One could use an off-the-shelf large language model (LLM), then fine-tune for forecasting or personalization. However, that’s most likely an expensive option with results which may not be any better than using ML.
This is where “agents,” like Agents for Amazon Bedrock, can help create multi-step solutions. Amazon Bedrock gives builders access to a library of foundation models along with a broad set of capabilities—so they can lower the lift when creating chatbots, summarizing texts, writing essays, and more. One might use an existing personalization model, such as Amazon Personalize, to determine product recommendations then pass those to a LLM to write a compelling email containing those for a customer.
ML excels at finding patterns and discerning rules from vast amounts of data which it can then use to make predictions. Typical use cases are demand forecasts, image (or even facial) recognition, and fraud detection. After “seeing” patterns over and over, it can then make predictions with a level of certainty. Because it’s less “creative,” the results are more transparent and understandable.
We also hear questions like, “will I be able to reduce labor using generative AI?” Think of generative AI as a better tool for many things that will increase efficiency, but it shouldn’t be relied on completely. If you’ve got 10 editors creating product content, then perhaps generative AI will reduce that number (allowing them to work on something else), but it will never replace them completely.
So, in summary, both generative AI and ML solutions will play a part in your retail enterprise. This leads us to ask, what are the best retail use cases for generative AI?
Generative AI Use Cases
My favorite thought experiment for finding generative AI use cases is to think about what could be done with a thousand interns working for free. You wouldn’t hire a bunch of editors to write personalized emails to your customers, but you might assign a thousand free interns to the task. “Thousand” implies scale, “free” is low cost, and “interns” are smart but not experts—a good measuring stick for generative AI use cases.
What could you do with a thousand free interns?
Here are some other jobs for those “interns”:
- Write better product descriptions for the website
- Collect trends from all the product reviews
- Summarize the sales reports
- Answer customer questions online
- Make product recommendations based on search terms
One way to look at the landscape of use cases is to consider the three main constituents of retail: employees, products, and customers as depicted in the following graphic.
Generative AI can be used to design better products and then market those products. It takes a lot of time to photograph and describe products on a website, something that can be automated as we’ve seen from AWS Partners like ContentStack, AI21 Labs, Anthropic, Amplience, and many others.
Making employees as productive as possible by minimizing rote tasks allows retailers to redeploy resources to higher impact activities. For example, adidas, the multi-national sports brand, used Amazon Bedrock to build a knowledge management solution.
“Using Bedrock, we have developed a generative AI solution that gives the community of adidas engineers the ability to find information and answers from our knowledge base through a single conversational interface, covering everything from getting started to highly technical questions,” said Daniel Eichten, VP of Enterprise Architecture, adidas.
One of the most important capabilities of Amazon Bedrock is how easy it is to customize a model. Customers simply point Amazon Bedrock at a few labeled examples in Amazon Simple Storage Service (Amazon S3), and the service can fine-tune the model for a particular task without having to annotate large volumes of data (as few as 20 examples is enough).
Imagine a content marketing manager who works at a leading fashion retailer and needs to develop fresh, targeted ad and campaign copy for an upcoming new line of handbags. To do this, they provide Amazon Bedrock a few labeled examples of their best performing taglines from past campaigns, along with the associated product descriptions. Amazon Bedrock makes a separate copy of the base foundational model that is accessible only to the customer and trains this private copy of the model.
After training, Amazon Bedrock will automatically start generating effective social media, display ad, and web copy for the new handbags. None of the customer’s data is used to train the original base models, and since all data is encrypted and does not leave a customer’s virtual private cloud (VPC), customers can trust that their data will remain private and confidential.
Perhaps the most important and visible area is with customers, where generative AI can have many impacts.
Before continuing, I want to point out that our goal here is not to “fool” the customer into thinking they’re dealing with a human. Rather, generative AI is used to improve the relationship with the customer, making information more accessible and relevant.
When dealing with customers, the contact center is a critical touchpoint where the natural language abilities of generative AI can help both employees and customers get the answers they need quickly. AWS Partners like Talkdesk and Genesys are taking this to heart with their contact center solutions.
“To make it easier for them to tap into the potential of generative AI, we’re enabling our users with access to a variety of large language models, such as Genesys-developed models and multiple third-party foundational models through Amazon Bedrock, including Anthropic’s Claude, AI21Lab’s Jurrassic-2, and Amazon Titan,” said Glenn Nethercutt, Chief Technology Officer, Genesys.
Similarly, premier travel media company Lonely Planet is using generative AI to help build travel itineraries for its customers. “By building with Claude 2 on Amazon Bedrock, we reduced itinerary generation costs by nearly 80% percent when we quickly created a scalable, secure AI platform that can organize our book content in minutes to deliver cohesive, highly accurate travel recommendations,” said Chris Whyde, Senior Vice President of Engineering and Data Science, Lonely Planet.
Generative AI is better at creating, while ML is better at predicting. Essentially, generative AI is just another tool in the toolbox, not some magic bullet. Both generative AI and ML solutions have an important place in the retail enterprise. With enterprise-grade security and privacy, access to industry-leading foundation models, and generative AI-powered applications, AWS makes it easier to build and scale generative AI—built for your data, your use cases and your customers.