Unpacked: Agentic AI is the latest retail buzzword
There’s a new AI buzzword on the block.
In recent months, major retailers like Amazon and Walmart have unveiled plans to move toward agentic AI. Unlike the current generative AI e-commerce assistants, the idea is that agentic AI bots will be able to autonomously complete tasks on their own, without relying on inputs. According to industry experts, agentic AI systems will eventually become the standard when it comes to the online shopping experience. However, these systems are still years away from being able to understand and contextualize shoppers’ search and product discovery habits. Still, retailers are making investments now to eventually make their AI bots more agentic.
According to Forrester’s recent report on agentic AI, these “systems are poised to not only become the backbone of the knowledge economy but will completely redefine how organizations operate and compete.” In recent weeks, Walmart, for example, announced it’s investing in agentic AI as a means to prepare for personal shopping agents in the future.
“While the technology is still in its early days, Walmart is moving rapidly and intentionally to integrate agentic capabilities into existing workflows across the business, creating powerful new capabilities as we build the future of retail,” wrote Hari Vasudev, CTO at Walmart U.S.
This past week, Walmart unveiled Sparky, its generative AI assistant, which can do things like summarize what reviews say about a specific product or offer recommendations. But the hope is that Sparky will eventually move beyond just generating answers and product information and, with the help of agentic AI, be able to choose specific items based on the Walmart customer’s shopping history, whether by booking services or re-ordering items on the customer’s behalf. A similar program was recently announced by Amazon, which is piloting a “buy-for-me” button, a feature that can purchase items from third-party websites for the customer.
The evolution of AI
As AI has become more popular, different subterms have popped up to explain different evolutions in capabilities. Generative AI, a term that became more popular with the rollout of ChatGPT, refers to algorithms that were trained on a particular set of data, and can generate texts, images or other content based on different inputs from a user
When talking about generative AI in e-commerce tools, it’s often referring to AI agents that can generate product suggestions or answer questions about what’s included in a product by, say, scraping reviews. Assistants like Amazon’s Rufus and Walmart’s Sparky fall into this category. By contrast, agentic AI, a term that started to become more popular in the past few months, means these assistants will be able to act autonomously without follow-up questions and user intervention.
Amazon’s “buy-for-me” button, which claims to use agentic AI, will — if a customer searches for a product that’s not found on Amazon’s website — automatically generate a link for the customer to buy that product on another site. If a customer decides they want to buy that product, they click on a “buy for me” button. Then, “using agentic AI capabilities, Amazon makes the purchase by securely providing the customer’s encrypted name, address and payment details to complete the checkout process on the brand’s website,” according to a press release explaining the feature.
While retailers and brands continue to embrace internal and consumer-facing generative AI, iterations like agentic AI can have a muddled definition.
Emily Pfeiffer, principal analyst at Forrester, said that current generative AI models can help by scouring the web and summarizing information and question-and-answer tasks. In contrast, she said,
agentic AI systems are autonomous.” What’s more, Pfeiffer said there is some confusion around the definition because some retailers announce tools that conflate agentic AI with actual AI assistants, also known as AI agents.
Pfeiffer explained that the main differentiator of agentic AI-based e-commerce is that it will be able to perform e-commerce tasks with minimal human prompts or intervention.
Still, as interest in agentic AI grows, solution providers are trying to simplify how retailers think about investing in it.
A.J. Ghergich is the vp of consulting services at Botify, an AI SEO company that works with brands like Crate & Barrel, Macy’s and Sephora. Ghergich said clients are increasingly interested in understanding and building agentic AI tools to power their search results on new platforms like ChatGPT and Perplexity.
Ghergich said much of the challenge of getting further adoption of agentic AI will come down to customers’ trust. Shoppers would have to be willing to rely on a self-acting agentic AI assistant to find and purchase the right item on their behalf.
According to an April report by Forrester, only 23% of online adults in the U.S. are comfortable giving up personal information to generative AI tools. This report also emphasizes the importance of the customer’s trust and comfort in AI adoption, which is still in its infancy. As an example, Ghergich said, agentic AI would eventually have enough insights into a customer’s health history to suggest and buy the right running shoes for them. “It should already know I have flat feet or had knee surgery six months ago,” he added.
Ghergich said, for retailers selling these products, the key is to push out accurate information for assistants’ live retrieval, especially as ChatGPT and Perplexity search adoption grows.
“We’re telling them to do some old school things like focus on feeding Bing with up-to-date product information,” said Ghergich, adding that Bing’s database helps feed OpenAI’s ChatGPT through the Microsoft partnership.
As for using buzzy terms like “agentic AI,” Ghergich said that retail companies using marketing jargon is to be expected. “I don’t get too worked up about it,” explained Ghergich.
“It’s an alphabet soup of acronyms, but we’re all just trying to be visible to customers,” he said.