This month, Target will launch its official generative AI chatbot, dubbed the Store Companion,which will be deployed across Target’s 2,000 stores spread across the US, acting as a helper for their store employees.
According to the retailer, the Store Companion will be able to “answer on-the-job process questions, coach new team members, support store operations management and more.”
With the release of the bot, Target has become the first major retailer to provide generative AI capabilities to its service employees. However, the Companion is just one of several generative AI initiatives that the US retailer has undertaken.
With a dedicated team of technologists, data scientists and researchers, 1,700 of which are based in Bengaluru, the retailer has been working on several generative AI initiatives over the last year.
Speaking to AIM, Brad Thompson, Target’s senior vice president of technology, said that the retailer has spent the last year working on as many as 20 generative AI use cases that they’re hoping to scale from proof of concepts, with the Companion as part of this initiative.
“Generative AI presents us with new tools and techniques to create experiences that were not possible before at scale with traditional AI. But it’s very difficult to go from a proof of concept to scaling it to the size of what our operations are. So that’s where we’ve been focused for the last year, scaling them so that it can be used by 400,000 team members on the field or by hundreds of millions of guests in the US,” he said.
Interestingly, the way Thompson explains the adaptability of the Companion and other generative AI integrations seems to reflect on the overall consensus of how retail-oriented chatbots will be dealt with across the board.
How Are Retailers Using Proprietary Models?
The Companion leverages GPT-4o in its functioning, with the retailer having already partnered with Google, Microsoft and OpenAI, as well as others in the AI field. Thanks to this, Thompson said that they’re able to cater to whatever generative AI needs come up.
“We have partnerships with all the big players. We are willing to use Google’s LLMs, OpenAI’s LLMs or open source LLMs that we can host on our infrastructure. They all have different strengths and weaknesses, so depending on the kind of experience we’re trying to build, we will experiment with multiple and pick the right one,” he said.
This also seems to be the case for many retailers and other enterprises, wherein the general consensus seems to be partnerships with multiple AI companies, rather than just one, to leverage their capabilities.
Potential For Leaks?
Thanks to this, cybersecurity is a major focus. Thompson said that while there was no way that the retailer would build their own LLMs, moderation was something that was focused on heavily, especially when it comes to integrating a third-party LLM.
“We will use proprietary LLMs extensively, I would be surprised if we built our own LLMs. We benefit from their investment, but we also add our own moderation services on top.There are very clear guidelines about what a retailer has to do and an LLM provider, like Google or OpenAI, doesn’t care about that. So we create our own moderation services and do lots of testing,” he said.
This involves a cybersecurity team who internally red team their systems, or hire contractors to do so, with Target actively investing in these capabilities. This is similar to AIM’s recent coverage of how jailbreaking is turning into a profitable industry, as well as cybersecurity startups coming up to deal with potential threats due to generative AI integration in enterprises. However, Thompson admitted that there would always be a risk.
“We take that risk very seriously, but that doesn’t mean there won’t ever be a situation. We think it can and will happen. It happened to Amazon, and I’m sure that’s something they’re looking at and learning from. But we will need to invest and test, test, test. The risk is not zero, but we think there are some common sense precautions to reduce those risks,” he concluded.
