Artificial intelligence is transforming commerce, by replacing online searches with digital agents that act as personal shoppers; but in coming years, it will go beyond and literally break e-commerce out of the box.
A lot of consumer brands now have to interact with shopping decisions starting on ChatGPT or Google AI Overviews, explains Rishabh Jain, co-founder and CEO of Fermat Commerce. The company recently launched AI Search Commerce Engine, a solution to drive brand discovery and sales conversions through AI agents.
“That question of what does a brand do, given that most searches are now going to start on an AI agent, is the primary question that we are interacting with,” says Jain.
To adapt to this AI-driven, agentic future, brands need to work through a four-step process, Jain explains. First, they have to monitor how often the brand comes up on a particular query, “just figuring out if your brand shows up or not,” says Jain. Most companies are still working through this first stage, but the next steps become progressively complex and more difficult to achieve, he warns.
Once the brand has visibility about how it shows up on multiple queries by large language models (LLMs), it needs to spot content gaps where it is not showing up due to lacking product detail pages or other content that can be picked up by agents in those queries. Having spotted the gaps, the third step is figuring what content is required to fill those gaps, Jain explains.
“Okay, here’s my visibility across all of these queries: Where do I have content and where do I not have content?,” he explains. Just as important is the question that follows: “Where do I need to produce more content in order to let people know that I can help them with this particular problem, even though the LLMs don’t know that yet about my product?” says Jain.
Addressing the question of how to produce brand content that will be picked up by LLMs is the fourth and toughest stage of the process. As expected, the number of vendor partners available to assist marketers at that stage shrinks in number, Jain says.
“Visibility is the easiest. Content gap identification is somewhat easy, but not that easy. Producing the content is not easy. And then producing the content in a way in which is relevant and hosted by your system means you have to be integrated into all of those systems,” says Jain.
Most marketers have taken steps to address the first step, visibility, but few have moved beyond, says Jain. “I think the pace of iteration there is very low,” he says.
While it’s easier to agree on how to gain visibility with agents and LLMs, a shortage of tools and expertise leads to trouble getting stakeholders on board for the next steps, says Jain. “Now you have to decide how you’re going to do it. And because there’s not an agreed-upon way to do it, it’s tough to make that decision,” he says.
Four more steps to the AI future
However, sales organizations are determined to harness AI, says Jain. He expects by the end of next year, most enterprises at least will have embarked on the journey, he says. That process also is made up of four steps that also grow in complexity with each stage, Jain explains.
The first step is to understand what visibility into AI query results reveals by using first- and third-party data to understand what a consumer would be looking for when querying an AI agent about the brand, its products or similar items.
“You can use your first party data to reasonably understand what someone would look up when they’re going to ChatGPT to do their research,” Jain explains. Third-party data, mainly gleaned from social media, can help sort out how consumers are talking about the product, so the marketers can figure out which queries to monitor.
The second step is to use the data set of queries generated in the previous step and perform a gap analysis – for example, comparing the conversations on social media with the content of blog posts or product detail pages.
The third step, producing relevant content, becomes more complicated, because not only does the content need to align with the brand voice and context, but it also has to connect with the product feed and be relevant, so the LLMs will retrieve it. “This is where your engineering team will have to get involved,” says Jain.
The fourth step, publishing content, product information and product detail pages on the website, will also require an assist from engineering. LLMs want live, verifiable product information, so all those items need to be integrated, Jain explains.
“These last two steps are the most complex from an engineering perspective, because you have to be integrated between content and product feed, to publish the right type of content that the LLMs will actually surface and know it’s fresh and know it’s correct,” he says.
It is unlikely for most brands to have the scope of technology and talents in-house to address all these stages, so many will need to rely on vendors to support their efforts, says Jain. All tech vendors are aware that this is the future, and are developing products to help, he says.
“We are the only company that makes the whole process seamless,” says Jain. Fermat Commerce can handle all four stages, including producing content and hosting in a way that LLMs can find it and interact with it, he explains. “Once we integrate with your back end systems, it is literally a push of a button,” he says.
Good products with good content
Marketing to AI agents is in a way a more pure form of product-driven marketing, because agents are sifting through product information and making informed decisions, says Jain. Marketing to an AI agent comes down to giving it the information it needs in order to make a good decision for the person querying it, which is why this content production pipeline becomes so important, he says.
“It becomes just representing the reality of what the product is capable of, and then allowing for the LLM to make the best decision for the user,” says Jain. “So good products with good content is what is going to win.”
As AI use spreads, it will transform the e-commerce user experience and websites. Shopping sites will be more visual and dynamic, rather than today’s static experience. He likens the transition to the way touchscreens changed user experience in mobile.
“It’s totally not obvious that the future of e-commerce is a grid,” says Jain. “It should evolve as Gen AI makes it possible to do totally different things.”
Suggestion engines have tried to personalize the experience in the past, but “they don’t actually change how the website looks. They just change which product shows up inside of that little box,” he says.
AI will lead to more dynamic websites, he says. Shoppers like to browse, and AI will allow sites to adapt to individual behavior, he explains.
“The store should be changing as the consumer does stuff on the website,” he says. “No more little rectangles.”
